Why every email has two jobs, and the industry only measures one

Marketing has spent twenty years perfecting the question of what to put in the next email. It has never built an instrument for the harder question underneath it: whether anyone will be waiting to open it.

1

The Blind Spot

Take the last hundred emails your brand sent. Not a representative sample — the actual hundred. Lay them out and ask one question of each: was this email designed to make the next one more likely to be opened?

For most programmes the answer is none of them. That is not a criticism of the people who built them. Every one of those hundred emails was designed with care to do something today: convert a browse, recover a cart, confirm a shipment, announce a launch, move a slow-selling line. Each had an owner, a brief, a target and a report. The reports exist and they are read. What does not exist anywhere in that stack is a line for what the hundred emails did to the audience that received them.

This is the asymmetry the essay is about. Revenue from an email is attributed with real precision. The platform can tell you which segment opened, which link was clicked, which order followed, and what the whole thing produced against a target, often to two decimal places. The attention that the same email consumed is attributed to nothing at all. It is spent on every send, it is finite, it belongs to the brand as much as any other asset on the balance sheet, and it appears on no statement anywhere in the organisation.

Every email programme knows what its emails earned. None knows what they cost.

The cost is real and it is observable — but only at a level of aggregation where nobody is accountable for it. It shows up as the engaged base shrinking a little each quarter. It shows up as the list growing while the number of people who reliably open it does not. It shows up, eventually, as a rising share of media spend going to reach customers the brand already has, because the owned channel can no longer reach them. By the time it surfaces, the campaign that caused it was declared a success eighteen months earlier, and the person who ran it has been promoted on the strength of the number that was recorded.

The industry recognises the problem without naming it this way, and has built a full set of instruments in response. Frequency caps reduce pressure. Fatigue suppression holds back customers who have seen too much. Sunset rules remove people whose engagement has already decayed. Send-time and send-list optimisation try to place fewer messages into bad moments. Re-engagement journeys attempt repair once the decay is visible. Every one of these is useful and every one of them is a brake: it reduces the rate at which attention is spent. Not one of them creates a systematic reason for future attention to grow. The programme has an elaborate braking system and no accelerator.

Two properties keep this invisible. The first is delay. Attention does not fall off a cliff after a bad email; it erodes across dozens of them, over quarters, in a curve too gentle to trigger anything. The second is diffusion. No single email causes the erosion, so no single email can be charged for it. The damage is real in aggregate and unassignable in particular, which is the precise shape of a cost that no management system will ever capture on its own.

It would be convenient at this point to reach for an open-rate statistic — the familiar line that marketing emails are opened by only a small fraction of the people they are sent to, and that this proves email has failed. That argument should be resisted, and not out of politeness. Reported open rates today are contaminated by privacy-driven pre-fetching, which registers opens that no human performed. The number is not measuring what it appears to measure, and any thesis built on it can be dismantled by anyone who knows that. The argument does not need it. It needs only the observation that the responsiveness of a base changes over time as a result of how it is treated, and almost nobody measures the change.

There is an existing instrument for this, and it is already in the vocabulary: Real Reach — the share of the base reliably reachable today — and CRR, the rate at which that share decays. Those two numbers are the missing page of the statement. They are not new metrics invented for this essay; they are the existing ones, pointed at the question nobody asks.

That question is what the rest of this essay is about, and it can be put in a single line. Not: was this email good? But: what did this email do to the probability of the next one?

2

The Two Jobs

Here is the doctrine the blind spot points to, and it is deliberately plain.

Every email has two jobs: deliver value now, and increase the probability of future attention.

Or, compressed to something a team can hold in its head on a Tuesday afternoon: do today’s job, earn tomorrow’s attention.

The first job is the one the industry has spent twenty-five years learning to do. The second is the one it has never been paid to do, never been measured on, and therefore never systematically built. Call them Track 1 and Track 2 — but the naming matters less than getting the relationship between them right, and this is where the idea is easy to break.

The tempting move is to divide email types between the two tracks: promotional mail on one side, relationship mail on the other. That division is wrong, and the rest of this essay depends on not making it. The four jobs an email can perform — Sell, Engage, Notify, Digest — describe what the email is doing. The two tracks describe what it is optimising. These are different axes. Any of the four jobs can sit on either track, and in practice a well-built email sits on both at once.

Figure 1. The job and the track are separate axes. A Digest leans towards Track 2 and a Sell email towards Track 1, but neither is confined to one column.

A Sell email can do superb Track 2 work if it leaves the buyer holding something that has not resolved. A Notify email — as the fourth section will argue at length — is often the best Track 2 email a brand sends, without anybody having intended it. A Digest, which looks like the natural Track 2 format, can consume attention rather than build it if it is dull, and a dull Digest is a more expensive mistake than a dull promotion, because it was supposed to be the deposit.

So Track 2 is not the name of a category of email. It is a property that any email can have more or less of, and the useful diagnostic is not which emails are Track 2? but how much Track 2 is in the emails we already send? For most programmes the answer is close to none, spread evenly across all four jobs — which is a considerably more interesting finding than a missing format, because it means the fix does not require a new email. It requires a change to the ones already going out.

The mental model, stated as an equation rather than as arithmetic:

Email Value = Value Now + Attention Later

Most programmes measure the first term with precision and assume the second is zero. It is never zero. It is positive on the emails people are glad to receive and negative on the emails they tolerate, and over a few hundred sends the second term dominates the first — because the first term is collected once and the second compounds.

For readers who have followed this series, this is the message-level view of something already argued at portfolio level: attention behaves like an account, with deposits and withdrawals and a balance that can be read. The two jobs are what that account looks like from inside a single email.

One consequence is worth stating before moving on, because it changes what a marketing team is optimising. If every email has two jobs, then an email can succeed completely and still be a mistake. It can hit its revenue target, clear its cost, satisfy its brief, and leave the base measurably less willing to hear from the brand. Under a one-job model that email is a success and there is no vocabulary available to say otherwise. Under a two-job model it has a recorded gain and an unrecorded loss, and the only open question is which was larger.

3

Track 1 Is Not Broken

It is important to be accurate about the state of the art here, because the argument does not depend on Track 1 being bad. It depends on Track 1 being excellent and insufficient, which is a harder and more interesting claim.

Sell has become very good. A decade ago a promotional email was a broadcast with a segment attached. Today the better programmes decide what to show at the level of the individual: next-best-action decisioning, live inventory, propensity models that know which of forty categories this person is currently in the market for, and increasingly agents that assemble the offer rather than selecting it from a list somebody built in advance. The gap between a competent Sell programme in 2016 and one in 2026 is not incremental. It is a different discipline.

Engage has become disciplined. Welcome, browse, cart, pre-purchase, post-purchase, replenishment, renewal, milestone — the journey library is now standard equipment, triggered by real events rather than a calendar, with timing that answers a customer moment instead of a marketing one. When these programmes are built well they are the highest-performing thing in the stack, and they perform because they are responsive to what the customer is doing.

Notify has become precise and trusted. Confirmations, dispatch, delivery, status, statements. High-integrity, first-party, near-universally welcomed. It is the part of the email programme nobody complains about.

All of this is real, and AI is about to make it substantially better again. The decision of what to put in front of a person at a given moment is exactly the kind of problem these systems are suited to, and within a few years the marginal return on further improvement to today’s job will be close to exhausted. Track 1 is not the frontier. It is close to solved.

And yet the base keeps drifting. The reason is structural, and it survives any amount of improvement to the decisioning.

Track 1 email is episodic. Each message has to win attention on whatever happens to be true at the moment it lands.

An episodic message is complete in itself. It arrives, it makes its case, it is acted on or ignored, and it closes. Improving it means making a better case — a sharper offer, a more relevant product, a better-timed trigger. That improves the odds of this message. It does very little for the next one, because the next one will arrive into the same conditions and have to make its own case from scratch.

That statement needs one qualification, because in its unguarded form it is false. Track 1 does leave things behind. It leaves brand familiarity, sender reputation, the accumulated sense that this company is competent and does not waste your time, and a general willingness to give the next message a moment’s benefit of the doubt. That residue is real and it matters. But it is implicit memory — held loosely in the reader’s impression of the brand, unmeasurable, unaddressable, and impossible to build on deliberately.

What Track 1 does not leave behind is an explicit object: something specific that exists after the email is closed, that the reader knows they hold, whose state has changed as a result of this interaction and will change again as a result of the next one. Implicit memory makes the reader slightly more receptive. An explicit object gives the reader a reason.

Figure 2. Episodic and serial email. In the first, each message argues for itself and nothing survives the close. In the second, an explicit object is carried across, and yesterday’s interaction is still doing work today.

So the honest formulation is this: today’s brand email is largely stateless, and the Next Open requires state. Not a better argument each time — a thing that persists between the arguments. The distinction sounds small. It is the difference between a programme that must be re-justified with every send and one that has something working on its behalf in the gap.

Which sets up the problem the rest of this essay addresses. If AI makes each individual email close to perfect and every email still has to earn its attention from scratch, the programme will have improved the quality of isolated moments without building any continuity between them. That is a real and expensive kind of progress, and it has a ceiling that better decisioning cannot raise.

4

Notify Already Knows How

Before proposing a mechanism, it is worth noticing that every brand already runs one, at scale, successfully, and draws no lesson from it whatsoever.

Ask why anyone opens an order-shipped email. It is not the subject line, which is usually four words and has had no creative attention paid to it in years. It is not the offer, because there is not one. It is not timing in any sophisticated sense. It is opened because the reader already holds state: I ordered something, and this is about the thing I ordered.

The chain that follows is the clearest demonstration in commercial email of a mechanism nobody has named:

Figure 3. The transaction supplies the persistence. One unresolved object is carried across four messages, and no message in the chain has to argue for attention on its own.

Order placed. Shipped. Out for delivery. Delivered. Four messages, each of which is opened at rates a promotional campaign would be delighted with, and not one of them is persuasive. They are not trying to be. Each simply advances an object the reader already knows about and cares about, and each one moves it closer to a resolution the reader is waiting for. The reader is not being convinced to care four separate times. They cared once, at the point of purchase, and the state has carried that caring forward across every message since.

The pattern generalises immediately, and once seen it is everywhere in the transactional stack. A booking creates state, and the flight-change notice advances it. A loan application creates state, and every status update advances it. A payment creates state; a refund request creates state; a service ticket creates state. In each case the customer is following an unfolding object, and the emails are instalments in something already underway rather than fresh appeals for attention.

Transactional email already knows how to create anticipation. The transaction carries state across messages. The problem is that relationship email has no equivalent.

This is the most useful fact in the essay, for three reasons.

First, it removes the burden of proof from the mechanism. Nobody has to be persuaded that held state produces opens, because every brand has the evidence in its own reporting. The behavioural claim is not a hypothesis about what customers might do. It is a description of what they already do, in the one part of the programme where the condition happens to be satisfied.

Second, it explains why the highest-opening mail a brand sends is the mail it thinks about least. Notify earns its opens structurally rather than editorially. No creative team is assigned to it. No subject-line testing is done on it. It is filed under operations rather than marketing, which is exactly why the lesson has never travelled — the people responsible for attention have never had reason to look closely at the emails that hold it best.

Third, and most usefully, it reframes what is being proposed. The rest of this essay is not an argument for gamifying email, which is what any proposal involving points and collections sounds like on first hearing. It is an argument for engineering into the relationship the same property the transaction supplies for free. The transaction produces held state as a by-product of commerce. Relationship email has to produce it deliberately, because nothing about the relationship generates it automatically.

It also explains why more personalisation does not solve this, which is the objection most likely to be raised by a team that has invested heavily in its data stack. Personalisation improves the probability that today’s content matches today’s need. Persistence changes the starting condition for tomorrow. The first says: we know something relevant about you now. The second says: something you already care about is still in motion. Both are worth having, and only one of them carries a reason across the gap between messages. A programme can be perfectly personalised and completely stateless, and most sophisticated programmes are exactly that.

There is one more thing the Notify chain teaches, and it is a constraint rather than an encouragement. The state has to be real to the customer, not merely tracked by the brand. The backend has always known everything: purchase history, browse behaviour, propensity scores, lifecycle stage. None of that produces a single open, because the customer cannot see it and does not hold it. What makes the shipping chain work is that the object lives on the customer’s side of the relationship. They know they have an order out. Any mechanism that hopes to reproduce the effect has to put something on the customer’s side too.

5

Three Ways to Earn the Next Open

If the object is to raise the probability of the next open, there are exactly three mechanisms available. They are not equally good, they are not equally expensive, and only one of them has been built into email infrastructure.

Figure 4. Three mechanisms, distinguished by where the reason for tomorrow’s open is held.

Promise. The reader knows what is coming and roughly when. The morning brief at seven. Friday’s five ideas. The market read before the open. This is the mechanism publishers perfected, and it works — a good newsletter has a standing appointment in the reader’s day that no campaign can buy. The distinction that makes it work is ownership: a promise is accepted by the reader; a calendar slot is owned by the sender. That is why a strong publication can send daily without feeling like a daily campaign, and why a brand sending daily usually does. But the promise sits entirely on the sender. Somebody must be good every single day, at a fixed hour, indefinitely, and the moment quality dips the appointment quietly lapses.

It is worth being fair about why brands have so few Digests, because the usual explanation — that marketers only think about selling — is both unkind and wrong. Brands ran the arithmetic. Daily editorial requires a standing production capability, judged against a campaign calendar where every slot has a revenue number attached and a Digest has none. The CRM team has an offer calendar. It does not have a newsroom. Under those economics the decision not to build one was correct, and it will stay correct until the production cost of a daily useful email falls far enough to change the sum. That is a real shift now underway, and it is why the Digest question is live again after twenty years of being settled.

Reciprocity. Yesterday’s email was useful, so today’s gets the benefit of the doubt. This is ordinary good relationship-building and it is real. Its weakness is that it has weak memory: goodwill exists, but there is no explicit object carrying it forward, nothing that accumulates, nothing the reader can point at. It raises the floor and it does not compound. A programme built only on reciprocity has to keep re-earning the same ground.

Persistence. Something from today survives into tomorrow, and it is typically incomplete or accumulating. Four of six cards. Seventeen of twenty-four collected. A prediction placed this morning that resolves tomorrow evening. A progression begun and not finished. The useful shorthand is unfinished state — the reader closes the email still holding something — but the concept underneath is persistence, and the distinction matters because the examples are not frustrating loose ends. A collection at seventeen of twenty-four is not an abandoned task. It is an asset that grows more valuable as it progresses.

That distinction is worth defending, because unfinished carries an unhelpful connotation of something broken or nagging, and a mechanism that makes people feel they owe the brand a task will fail quickly. What is meant is closer to pleasurable incompletion: a collection not yet complete, a series with more to discover, an outcome that will resolve later. Today’s interaction still has to deliver its own value and feel finished in itself. What it leaves behind is a larger object still in motion — not a debt, and not a chore.

Set side by side, the three sort themselves cleanly by what they ask of whom:

Promise is an expectation about future content. Reciprocity is a memory of past value. Persistence is an object from today that survives into tomorrow. And that produces the distinction that decides which one an email programme should reach for:

Promise asks the sender to create another reason. Persistence lets yesterday’s reason survive.

The economics of that difference are severe. Under Promise, the cost of tomorrow’s open is a fresh act of creation, paid every day, forever, with the quality bar rising as the reader’s expectation settles. Under Persistence, tomorrow’s open is substantially pre-paid by work already done — the state exists, the reader holds it, and the email’s job is to advance it rather than to justify itself from nothing. One model has a recurring cost that scales with frequency. The other has a set-up cost and a maintenance cost, and gets cheaper per open as the state deepens.

None of the three is exclusive, and the strongest programmes will run all three at once — a standing promise, delivered usefully, carrying persistent state. But if a programme can build only one thing, it should build the one that survives a bad week, and only persistence does. A promise broken twice is a promise withdrawn. A collection ignored for a fortnight is a collection waiting.

Which reframes the operating question a marketing team asks itself every week. The standard version is what should we send next? — and every planning meeting, editorial calendar and agent in the stack is pointed at answering it well. The Next Open asks a different one:

Every email programme asks: what should we send next? The Next Open asks: what did we leave them holding?

6

The Persistence Breakthrough

So what would persistence look like in an inbox that has never had it? The answer that has emerged from this work is a structure with three levels: a Card, which sits inside a Set, which sits inside an Album. One Album per person. Sets within it. Cards within those.

Described that way it sounds like a mechanic, and the instinct is to file it as gamification. That instinct should be resisted long enough to notice what the structure supplies, which is the thing the previous section said email had never had: persistent owned state between messages. Today you hold four cards. You close the email. Tomorrow you still hold four. Nothing in the interval reset you. And the fifth card is worth more than the fourth was, because of the four already held — which is the opposite of how a campaign behaves, where the tenth promotion is worth less than the first.

That is compounding, in the strict sense. Each interaction increases the value of the next one rather than depleting the reader’s willingness to have it.

One detail about how a card arrives decides whether any of this is worth opening, and it took a long time to find. The obvious design has the card arrive as the reward: do the interaction, receive the object. The better one inverts the order. The card arrives face down, and the interaction is what opens it — a single retrieval attempt before anything is revealed, with a right answer, pitched so that roughly a quarter of readers get it first time, and the answer shown immediately in the same view whether they were right or not.

A wrong answer still opens the card. It earns less, and that is the whole of the difference. The reason to build it this way is not fairness; it is that nobody taps to receive a picture and everybody taps to find out what is under a seal. It also collapses three things the programme would otherwise have to deliver in sequence — a reason to open, an interaction, and an object — into one moment.

The card is not the reward for the interaction. The card is what the interaction opens.

The design decision underneath deserves stating, because it is where most attempts at this fail. The obvious way to create persistence in email is a streak — consecutive days, visible counter, loss on a miss. Streaks work, and they work by loss aversion, which is why they are brittle. One missed day destroys the accumulated investment, and the reader who breaks a fourteen-day run frequently does not return at all, because what they lost was not the fifteenth day but the previous fourteen. A mechanism that punishes absence is a poor foundation for a mass-market inbox habit, where absence is normal and unavoidable.

Figure 5. Two ways to accumulate. The streak has a cliff; the collection has a pause.

A streak makes yesterday fragile. A collection makes yesterday valuable.

A collection has no cliff. Miss a day and nothing is lost; the four cards are still four cards, and the gap to six is still motivating tomorrow. Progress pauses rather than resets. This lets persistence compound without turning participation into an obligation, which is the only version of the mechanism that survives contact with people who have jobs and children and weeks when the inbox goes unread.

The same discipline extends past the streak, and it is the place where the marketing instinct will push hardest in the other direction. Nothing in a set is ever withdrawn. A brand can stop issuing new cards, concentrate a fortnight on one part of a set, or close a set and open another. What it should not do is put a closing date on a card, because a rarity that expires manufactures regret in everybody who arrives afterwards — and in any programme that works, almost everybody arrives afterwards. Where scarcity is wanted it belongs in the particular copy: where it was earned, at what level of recall, and whether it came from another person. Those are renewable. A window is not.

Three components do the work, and it is worth being precise about which does what, because they are frequently collapsed into one another. The Magnet earns today’s interaction — the small interactive unit inside the email that gives the reader something to do in under a minute. Mu records what was retained, not what was received; a balance moves when a reader answers for something weeks later, never when a card lands. The Album carries the persistent state across messages. Magnet earns, Mu records recall, Album persists. Confusing the three produces a system that looks complete and does not compound, because the recording layer and the persistence layer are doing different jobs.

That rule about Mu is worth holding on to, because the alternative is available and it fails quietly. If the balance moved on receipt, a brand could lift any reader’s number by sending more, which turns the currency into a measure of how much mail somebody has been given. Accruing on recall makes it a record of what stuck, which is the only version of the number that can be shown back to the reader in a subject line and mean anything to them.

The state also reads better as more than one number. Hold is what the reader has, and it moves daily. Finish is how many sets are complete, and it moves weekly. Know is how much can still be answered for, and it moves slowly, and downwards when attention lapses. A single progress figure eventually discourages everybody; three on three clocks means something is always nearly moving.

There is a fourth element that belongs to this structure without belonging to this essay’s argument. A Circle is best understood as a group in the way a WhatsApp group is a group: a person belongs to several, each with its own membership and its own life. Circles matter here for one specific reason — cards move between people, and movement is what turns a collection from a solitary accumulation into something social. There are three ways it happens and they are not interchangeable: a **trade**, reciprocal and card for card; a **gift**, one way with nothing expected back; and a **hand-over**, agreed in person between two people in a room and settled afterwards. Only the last of those carries the fact that somebody was there. A card you cannot trade is an inventory item; a card you can trade has a market, and markets generate reasons to return that no sender has to manufacture.

The boundary should be drawn plainly. Circles are largely a consumer property. On the brand side their relevance is confined to trading, and the essay’s argument does not rest on them: persistence works for a single reader with no Circle at all. Social obligation — returning because other people expect you to — is a fourth mechanism, distinct from the three in the previous section, and it deserves its own treatment rather than a paragraph here.

The general principle is larger than the implementation, and it is the sentence worth carrying out of this section:

Messages stop being independent objects and become viewports into persistent experiences.

Read literally, that sentence changes what an email is for. A message need not be the experience. It can be the viewport into an experience that persists elsewhere, in state that outlives any individual send. This matters because inbox surfaces have always been transient — the message is seen, acted on or ignored, and closed — and a transient surface cannot hold a relationship on its own. Give it a durable object to look into and the transience stops being a problem. The email is the window; the state is what makes returning to the window worthwhile.

Cards, Sets and Albums are one way to build that, and they are not the only way. A financial services brand could let a customer accumulate a learning path that advances across messages and is visibly incomplete. A travel brand could build destination sets around places explored, saved and visited. A health or wellness brand could carry a progression through a habit or learning sequence without making a missed day destructive. A publisher could attach collectible structure to recurring knowledge, so that reading builds something rather than merely passing time. The form varies widely by category. The underlying property does not: something from the previous interaction has to remain valuable in the next one.

Whatever the form, it has to satisfy the condition the Notify chain established: the object must be real to the customer, held on their side, and changed by the interaction rather than merely reported to them. A progress bar the brand maintains internally and displays occasionally is not persistence. It is reporting with a graphic.

7

From One Brand to a Daily Habit

Nothing so far requires an outside party. It is worth conceding that fully and early, because the alternative — an argument that quietly arrives at and therefore you need a vendor — is the reason most essays of this kind are not believed.

A brand can build this alone. Nike could run a Nike collection across Nike emails and it would work. A bank could build a financial-learning progression that advances week by week. An airline could maintain destination sets that fill as a member travels. There is no technical barrier, no proprietary component, and no architectural inevitability that makes this someone else’s job. Persistence is a design pattern, and design patterns are available to everybody.

The constraint is not possibility. It is cadence.

Persistence converts into habit only at habit frequency. A collection that advances twice a week is a pleasant novelty; a collection that advances daily is a routine. The gap between those two states is where every attempt at inbox habit-building has historically died, and it is not a design problem. It is an arithmetic one. Habit formation requires repetition at a rhythm the reader’s day can hold, and a brand’s ability to supply that rhythm is capped by how often it has an honest reason to write.

Figure 6. The cadence problem. One brand writing only when it has reason to reaches a fraction of the frequency a habit requires; aggregated progression reaches it without any brand writing more often.

Three constraints arrive together, and they compound. Frequency is capped by how often the brand has an honest reason to appear. Variety is capped by how many recurring themes a single brand can keep useful before the experience becomes repetitive — a coffee company can be interesting about coffee roughly twice a week and then begins to strain. Production economics are capped by what it costs to generate and operate that material at a quality the reader will keep returning to. A brand that already struggles to justify a weekly Digest will not build a daily newsroom plus a daily persistent experience because the mechanics have become available.

Some categories are exempt. A trading platform, a news service, a marketplace, a high-frequency grocery or quick-commerce business has legitimate daily reason to be in the inbox, and for those brands single-brand persistence is entirely sufficient — they should build it and stop reading here. But most brands do not have that. A fashion label writing daily is not building a habit; it is depleting one. Most brands cannot independently sustain the frequency, variety and production economics required to turn persistent state into a daily habit — and the right thing to do with that sentence is to treat it as an operating constraint rather than a law of nature.

Which raises the escalation, and it is a question rather than a requirement. Can persistence become portable across brand relationships — so that the state a reader holds advances through the day regardless of which sender happens to be writing?

The arithmetic changes completely if it can. A reader who hears from twenty-five brands is receiving several messages a day already. No brand needs to write more often. If the same object advances across all of them, the progression is daily without any individual sender behaving unnaturally, and the habit that no single brand could economically supply becomes available to all of them at once. This is the structure that makes airline alliances more useful than any single carrier’s programme: the value comes from the aggregation, not from any one participant’s generosity.

That is the argument for a shared substrate operated above the brand rather than inside it — a recurring carrier for the interactions, and a ledger that belongs to the reader across every sender rather than to any one of them. It is an escalation of the thesis rather than a precondition for it. The Next Open holds perfectly well inside one brand. Portability is what makes it a daily habit rather than a weekly one.

It also changes what the durable asset is, and the answer is not the one a brand would guess. The valuable thing is not any particular set of cards. It is the press: the card chassis, the schema every card is built to, the album, the rail along which cards move between people, and the delivery that puts one in front of somebody every morning. Sets are printed on the press and any of them can fail without the press being worth less. A brand that wants in is renting the press rather than building one, and it sponsors a set rather than authoring it — which is the arrangement that keeps the album neutral enough for the next brand to join.

One practical constraint applies whichever route a brand takes, and it is the thing that kills interactive email pilots more reliably than anything else. Interactive and composed-at-open experiences are not universally supported across mail clients. Persistent state must therefore survive graceful static fallback — the reader on an unsupported client has to see where they stand and be able to act, even if the interaction happens a click away. A persistence programme that works only in its interactive form is a demonstration, not a product.

There is a sequencing consequence underneath that, and it is better heard before a launch is planned than after. Sending the interactive version at all requires registration with each mailbox provider, and registration requires a demonstrated record of low complaints — which a new sending identity does not have on the day it starts. So the first months run the static version by default and move the interaction into the inbox once the reputation exists. The architecture does not change; the order of delivery does.

And whether a brand builds this itself or has someone run it is an operating decision, not a doctrinal one. The claim here is narrower and easier to defend: this is a specialised job, it is not most brands’ core competence, it has meaningful set-up costs and real ongoing production demands, and jobs shaped like that tend to get industrialised. That is an argument about efficiency. It is not an argument about necessity, and it should not be dressed up as one.

8

What Would Prove This Wrong

The first section left a cheque uncashed, and it is time to cash it. The blind spot is not merely an unmeasured quantity. It produces behaviour that is locally rational and globally destructive, and that is a considerably more serious problem than a gap in reporting.

Consider a promotion that produces $100,000 in attributed revenue against a few thousand in cost. Under any current reporting standard that is a success, and it will be repeated, because the system is designed to repeat what worked. Now suppose that the same treatment, applied repeatedly to that cohort, leaves it measurably less responsive ninety days later — slower to open, quicker to ignore, a larger share of it drifting out of reach. The campaign P&L still shows the $100,000. It will never show the other half.

Figure 7. The recorded result and the unrecorded one. The campaign P&L has no line for attention consumed, and the cohort’s responsiveness ninety days on is where the missing line would have been.

This is a time-horizon problem, and time-horizon problems are not solved by trying harder. They are solved by changing what gets measured, because a system optimising a metric that omits a cost will reliably drive that cost to its maximum. Every marketing organisation running one-job reporting is doing exactly that, competently, at scale, with quarterly targets attached.

So the test has to measure both terms of the equation, and it has to be declared before the pilot rather than assembled afterwards from whatever the data supports.

The instrument. A concurrent randomised holdout — control and treatment running at the same time, against the brand’s current best effort rather than a prior-period baseline, with the control group receiving a pre-agreed normal treatment rather than nothing. Ninety days. Two questions, answered separately: did the intervention create value now, and what happened to that cohort’s future responsiveness relative to control? The second is measured on Real Reach and CRR, not on engagement uplift, not on opens, and not on any soft attention score constructed for the purpose.

The three questions the result has to survive. A single directional reading is not enough, because three different failures all look like success at first. Does the effect persist past novelty — a collection that produces excitement for a fortnight and then flattens is not persistence, it is a new toy, and the ninety-day window exists precisely to separate the two. Does Value Now hold — Track 2 is not exempt from the standard Track 1 is held to, and a mechanism that buys future attention by suppressing this quarter’s revenue has moved the problem rather than solved it. And does it hold against the right comparison — not against doing nothing, but against the brand’s current best effort.

The same design answers the more useful question of which mechanism to build. Promise can be tested against no recurring edition. Persistence can be tested against content that is useful and stateless, which is the honest control for the central claim of this essay. Collection can be tested against streak, which settles the design argument in section six with data rather than reasoning. Cadences can be compared against each other. The object is not to prove the architecture once and declare it settled; it is to make future attention an experimental outcome that sits alongside revenue in the same report.

One leading indicator is worth watching long before ninety days are up, because it isolates the claim more cleanly than anything else available. Count the opens that happen on a day when there is nothing new to collect — no card, no arrival, no reason supplied by the sender. Those opens can only be caused by something the reader is already holding. If that number is near zero, the programme has built a delivery habit rather than a persistent one, and the ninety-day result will confirm it slowly and expensively.

The kill condition. If the treated base does not hold or grow its engaged share against control over ninety days, the thesis is wrong. Not under-executed, not early — wrong, and it should be abandoned rather than extended for another quarter while somebody looks for a segment where it worked. Persistence is a falsifiable claim about human behaviour in an inbox, and a claim that cannot fail is not worth making.

It is also worth being precise about what is being claimed as new, because overclaiming here would be easy and would deserve everything it got. The phrase is not new; variants of earn the next open have been in circulation in this field for years. The idea that good email produces future opens is not new either — every thoughtful practitioner has known it, and frequency caps, fatigue suppression and sunset policies are the industry’s existing response to it. But as the first section noted, those are all brakes. They reduce the rate at which attention is spent. None of them builds any.

Email has no standard operating architecture for measuring and deliberately compounding future attention across messages. That is the gap. The phrase is not new; the architecture is.

The architecture has four parts, and they have all appeared above: the doctrine — every email has two jobs; the axis — the job an email performs is separate from the track it optimises; the mechanism — persistence, an explicit object held by the customer that survives the close; and the measurement — Real Reach and CRR on the treated cohort against a concurrent control. Take away any one of the four and what remains is either a metric with nothing attached to it or a mechanic with nothing to prove it.

Marketing has spent twenty years learning how to decide what to put in the next email, and AI is about to make that decision extraordinarily good. It answers the wrong question first. The harder question is whether anyone will be waiting to open it.

The next era of email will not be won by the programme that best decides what to send next. It will be won by the one that best decides what to leave behind.

Do today’s job. Leave a reason to come back. Earn the Next Open.

One programme, seen from both ends

A companion to The Attention Architecture

Architecture arguments are easy to nod along with and hard to picture. So here are two people who never meet: one Tuesday in detail, then the ninety days after it, then the seventh month.

1

Two people who will never meet

Meera is thirty-one and runs product at a mid-sized software company. She changed jobs two years ago and has not thought seriously about changing again since. Eight months ago she stopped opening email from the careers site she once used every week. Nothing happened. There was no unsubscribe, no irritation, no decision. She simply stopped, the way people stop.

Rohan is thirty-eight and runs retention at that careers site. He has a database of just under two million registered users and a number he does not put on slides, which is that around a third of them have not opened anything in six months. He has been told, more than once, that this is normal and that everyone’s file looks like this.

Neither of them will ever know the other exists. But over the next seven months each will be the reason the other’s number moves.

What follows runs in three acts, because no single one of them shows the whole thing. A day gives the texture — what it is like to be either of them for an hour. A quarter is the shortest span in which a habit can be seen forming or a holdout can be read. And the seventh month shows what the first two produced, which is the part neither of them was aiming at.

It is illustrative. The mechanics are real, the sequence is the one the architecture implies, and the figures exist to make the shape visible rather than to forecast anything. Nobody should quote them.

Figure 1. Three acts. A day for the texture, a quarter for the habit and the proof, seven months for the asset.

2

07:12 — The one message her software leaves alone

Meera, a Tuesday in March

Meera clears her inbox on the train, which takes about four minutes and is mostly deletion. Her phone has already reduced the overnight arrivals to a short brief, and the brief is accurate. A grocery delivery notice. A bank statement. A retailer telling her that forty-eight hours remain on something she was not going to buy. A fare alert for a route she saved in March.

None of these need her. Two of them her phone has already handled. In another year or two most of this will not reach her at all, and she will not miss it.

Figure 2. Four messages her software could handle. One that is addressed to her rather than to her wallet.

One message has not been reduced to a line in the brief. Its subject line opens with a number — twelve Mu — and then says something odd: card three of eight is waiting, and it does not say which one.

The number is hers. She has no memory of earning it, which is the point; it dates from a period two years ago when she used the site properly. What the rest of the line does is smaller and works harder: it tells her something is there and withholds what it is. What it does in that moment is not persuade her of anything. It tells her that this message knows her, and that something inside belongs to her already. That is a different proposition from every other line on the screen, all of which are asking.

She opens it, which she has not done since February.

Figure 3. What is inside: a signal, a card face down, a minute of work, and a balance that has not moved yet.

Inside there is one thing, and above it a card lying face down. Three questions about salary bands for her function in her city, and a line saying that answering turns the card over. Her mailbox supports the interactive version, so the questions open in place; had it not, the same minute would have cost her one extra tap onto a page — and she would have taken it, because by then she wants to know what is under the card rather than what is in the email.

She answers the first correctly. She gets the second wrong, and the correction is worth having — the function that moved most in the last year is not the one she would have guessed, and she will mention it to two people that week. The third asks her to place her own band against the median, and the answer is mildly annoying in a way that makes her think.

Sixty seconds. The card turns over. Salary Bands, her city. Add it to your album. Getting the second one wrong cost her a little of what she would otherwise have earned, and it did not cost her the card — which she does not notice, and which is the reason she does not close the message in irritation.

The number moves later rather than now. Twelve Mu is still twelve Mu at the end of the minute; it becomes fifteen a fortnight later, on the morning she is asked which function moved most and gets it right without being shown the answer first. She will not register the delay. What it means is that the figure in her subject line is a record of what she has retained rather than a count of what she has been sent, which is the only version of it that could survive contact with a marketing department.

The button is not pre-ticked, and nothing has been created behind her back. She taps yes because the card is a decent-looking object with something on it she now knows, not because she has understood anything about what sits behind it. And at the bottom of the message, in small type, a line that does more work than everything above it.

One of eight cards in this set.

Meera did not want a card when she woke up. She wants seven cards now.

Nothing was sold to her. No form appeared. Nobody asked her to consider a career move. The careers site spent that Tuesday making her marginally better informed about her own market, and in exchange it got the only thing it needed, which was to be someone she opens.

Key points

• The pre-open signal works because it reports her own state rather than making a claim.
• The Magnet leaves her knowing something she will repeat to other people that week.
• Where interactive rendering is unavailable, the same minute costs one extra tap. The intent does not change.
• The card is face down. Answering opens it, a wrong answer opens it too, and that is why she read to the end.
• The set is what creates the return: one card is a nice thing, one of eight is an itch.
• Mu moves on recall, not receipt. No brand can lift her number by sending her more.

3

09:05 — A database with two clocks

Rohan, the same Tuesday

Rohan’s dashboard used to be organised around campaigns. Sends, delivery, clicks, what is scheduled for the week. Those numbers still exist and he still needs them, but they no longer decide anything. The first view now splits the base by attention state rather than by campaign, which sounds like a reporting change and is in fact the whole programme.

The dormant cohort has been on his list of problems for two years, and it has never once been at the top. It is hard to argue for budget against a segment defined by the fact that it ignores you. Every proposal he has drafted has died at the same question, which is a fair one: how would we know it worked?

Because the honest answer has always been that he would not. A share of dormant users return on their own every quarter — a redundancy, a bad review, a friend who mentions something. If he sends a campaign and some of them come back, he cannot separate his campaign from their lives. He has watched three colleagues at other companies claim reactivation numbers he did not believe, and he did not want to become the fourth.

So the programme starts with a decision that has nothing to do with creative. He takes everyone who has not engaged in six months, three hundred and forty thousand people, and splits them at random into two halves. One half receives the new programme. The other receives exactly what they receive today, which is the same weekly broadcast that goes to everybody and that none of them open.

The holdout is not a control in the polite sense of the word. It is a group of a hundred and seventy thousand people that Rohan is deliberately choosing not to help, for ninety days, so that at the end he can say something true.

He is not measuring against last quarter. He is measuring against what his own company would otherwise have done, in the same weeks, to the same kind of person.

Then the second decision, which is the one his instincts fight. The programme going to the treated half is smaller than what he sends today. One interactive unit. No offers. No third-party placements. No banner strip. And it goes out weekly rather than daily, because a domain that repeatedly pushes into addresses that never respond is a domain that stops arriving in inboxes at all, and the engaged programme depends on the same reputation.

There is also an exit rule, written down before the first send, which he expects to be the least popular line in the plan. After a set number of unanswered sends, the address comes out. He is building a programme designed to give up on people, on purpose, in order to keep working for everyone else.

At 09:17 an analyst points at the first movement: a small set of previously dormant addresses opened, completed the interaction and claimed a card. Rohan does not celebrate. One morning is not a habit, and a claim is not revenue. He asks for the same figure from the half he held back, and for the curve on second opens when there is one to look at.

That is the shift in what his job is. The old question was whether the campaign worked. The new questions are whether attention was re-earned, whether the person did something worth doing, whether the action left anything behind, and whether they came back when nothing made them.

Key points

• The dormant segment fails on measurement, not on creative — that is why it never gets funded.
• A concurrent randomised holdout replaces the argument with an answer.
• The recovery programme is deliberately smaller and less frequent than the standard one.
• The exit rule protects the sending reputation the engaged programme relies on.

4

12:30 — The awkward meeting

Rohan

Rohan’s hardest conversation that day has nothing to do with subject lines, creative or deliverability. It is about ownership, and it is the meeting every version of this programme eventually has.

The brand director asks first, and asks well. If we are issuing the card, why do we not own the album? Legal follows with the adjacent question, which is the sharper one. Are we helping somebody else build a relationship with our customers?

Rohan puts one page on the screen.

Figure 4. The three boundaries, and the question underneath them.

Everything up to the interaction belongs to the brand. The brand chooses the content, chooses what it is willing to issue, and owns the relationship the email sits inside. The crossing after that is voluntary and visible: a reader can enjoy the minute and stop there, and only an affirmative tap creates anything on the other side. No list is copied anywhere. Invite, never enrol.

Then the point about the balance, which is the one that takes longest to land because everyone in the room has spent a career thinking in loyalty schemes. The Mu count is not something the brand issues and then owes. There is one balance and it belongs to the person, accumulated wherever she acts. The brand is showing her own number back to her, which is why it reads as recognition rather than as points. It also means she can never hand it to anybody else, which keeps it a record of what she did rather than something with a market in it.

And then the trade, stated without softening, because softening it now only moves the argument to a worse moment later. The album is not the brand’s.

The brand gives up ownership of the container in exchange for greater persistence of its relationship inside the container.

The reason to accept that is not generosity. It is that a campaign ends when its budget ends, and a card in an album that someone values is still there afterwards — looked at, missing from a set, mentioned in a group, remembered long after the email that issued it has gone. The brand’s relationship gains durability precisely because the brand does not control the whole environment.

The finance director asks the practical question, which turns out to be the one that decides the meeting. Do we need the album for the economics to work?

No. The recurring programme has to justify itself as a relationship surface on its own terms. The recovery programme has to beat a holdout on its own terms. Monetisation only arrives after recurring attention exists, and never in recovery at all. The consumer side is upside, and if it never materialises the programme still stands.

That answer is what changes the room. Nobody is being asked to subsidise a speculative network. They are being asked to make their own email worth opening, and then to let the people who value the object keep it somewhere durable.

The programme is approved with three boundaries written into it: invite, never enrol; the album is not ours; earn before monetising.

Key points

• The ownership question is not a legal detail — it is the meeting, and it should be had early.
• One balance belonging to the person, shown back to her. Not a scheme the brand issues and owes.
• The deciding question is whether the Digest pays for itself without the album. It has to, and it does.

5

20:55 — A five-minute appointment

Meera, the same evening

At ten to nine that night a second message arrives, and it is not another card. It says that a five-minute recall round opens in five minutes, and that the questions will come from cards she already holds.

She opens it at 20:55 because the appointment means something, which is a sentence that would have been untrue about this brand twelve hours earlier.

There are three questions. The first comes from the card she earned that morning. The second comes from a travel set she started a fortnight ago with an entirely different company, and she is briefly surprised that it knows about that, until she remembers that the album is hers rather than anyone else’s and that this is the obvious consequence. The third asks her to choose between two answers and say why.

This is where her balance moves, and it moves because she remembered something rather than because she received something. Holding a card pays nothing. Answering for it weeks later pays. That is a small rule with a large consequence: no brand can lift her number by sending her more, and the figure in tomorrow’s subject line is worth reading because it was earned somewhere she was not being marketed to.

The whole thing takes under four minutes. It is small, and the smallness is the point. Nothing here needed to arrive as a destination with a hundred features and a marketing launch. It grew out of things she already had: cards owned, gaps remaining, a balance, and now a reason to be somewhere at a particular time.

That is what a fixed appointment does inside this architecture. It is a cadence, not another product. Tomorrow it could be a recall round, next week something cooperative, later something longer. Each of them inherits the same collection and the same balance instead of asking her to sign up again and start from nothing.

Old email contains the content and hopes the reader clicks. This one revealed the next state of something she already had.

By 20:59 it is over and she has put the phone down. What survives is not the four minutes. It is that tomorrow morning’s message now has context before it arrives.

Key points

• A fixed-time appointment is a cadence inside the habit, not a separate product beside it.
• The questions draw on cards from more than one brand, because the album is hers.
• Small is correct. New experiences inherit the collection and the balance rather than rebuilding them.

6

Day 9 to Day 41 — A habit, forming

Both

The second card arrives the following week. She notices it faster, because the shape of the subject line is familiar now and because she remembers the number is hers. This one is about which functions are hiring in her city this quarter. Her album reads two of eight, and six blanks sit in a neat grid doing more to bring her back than either card she owns.

This is the part that is difficult to design and impossible to fake. Nobody returns for a message. People return for a thing that is unfinished. Collectors have known this for a century, which is why the album has always been sold with the spaces printed in.

Rohan sees almost nothing on day nine. Two sends is not a trend, the treated half is barely distinguishable from the half he held back, and this is the point in most programmes where somebody senior asks whether it is working. He has one advantage, which is that he agreed the ninety days in advance and wrote down that he would not look before day twenty-three. Deciding when to look, before you have any results, is most of what makes a holdout honest.

By day twenty-three Meera has opened four of four. On the system’s definition, which is a plain behavioural one rather than anything clever, she is no longer dormant, and she moves. From her side one thing changes: the email starts arriving daily rather than weekly, and it is fuller. Two interactive units now instead of one, and she picks whichever she prefers. A short piece of writing that repays the ninety seconds. And, for the first time, something being advertised — a professional course, placed carefully, which she does not click and does not resent.

That is the sequence working. She would have resented it on day one. On day twenty-three she has been given four useful things in a row, and a well-chosen advertisement inside a message she values reads as a normal part of the world rather than as the reason the message exists.

Day twenty-three is also the first time Rohan looks, and the first time in his career he has looked at a reactivation number without a knot in his stomach. The treated half has pulled away from the half he held back. Not dramatically — a few percentage points — but cleanly, and in the same direction, week after week. What matters is not the size of the gap. It is that the gap is the whole of the effect, with no attribution model underneath it for anyone in finance to dismantle.

By day forty-one Meera is at six of eight and has two copies of one card, which is mildly irritating in a way she recognises from being nine years old. She mentions it in a group of five people from her old team. One of them has been collecting the same set, has the card she is missing, and needs the one she holds twice.

Nothing is settled between them. He sends her his spare because she asked, and eleven days later she sends him hers, by which point neither of them is thinking about the first exchange. The system does not broker a trade and does not price one card against another, which sounds like a missing feature and is the reason the group is still talking a fortnight later. A settled swap closes the matter. Two gifts leave it open.

Two things happened there that no referral programme has ever managed. Nobody was asked to promote anything — she was not offered a discount for bringing a friend, and the careers site was not mentioned in the conversation at all. The transmission happened because her position was incomplete without another person, which is a far more reliable engine than goodwill. And a third person in that group asked what they were both talking about, and now collects too. That is what acquisition looks like when it costs nothing.

It also matters that this happened in a group about work. Meera is in several such groups — family, a running club, three people she argues with about cricket — and her sets belong to different rooms. If everything had to route through a single group, most trades would fail, because the person holding the card she needs is usually not in the room she happens to be standing in.

Key points

• The blanks pull harder than the cards. The unfinished set is the return mechanic.
• Agreeing when to look, before there are results, is most of what makes a holdout honest.
• Movement between programmes is behavioural and automatic, not a campaign decision.
• Monetisation lands acceptably on day twenty-three precisely because it was absent on day one.
• The sharing trigger is about the collection, not the brand, which is why it travels.
• Two gifts, not one swap. Nothing was settled, which is why the group was still talking a fortnight later.

7

Day 90 — The report

Rohan

Rohan’s report is one page and one chart, and it is the first thing he has produced in this job that he does not have to defend.

Figure 5. The whole argument on one chart. Two halves of one population, ninety days, one difference between them.

In the treated half, a little under eleven per cent of people who had been silent for six months now have some engagement in the trailing thirty days. In the half he held back, that figure is one point six per cent — the rate at which dormant people wander home by themselves, which is exactly the number the argument has always been missing.

The difference between those two lines is not an estimate. It is not a model output. It is what happened.

And it is not perfectly clean, which he says on the page rather than waiting to be asked. Some of the treated group were also reached by paid media in the same period. Some came back for reasons that had nothing to do with him. Contamination exists and always will. But it applies to both halves, which is the entire reason for splitting them, and no other line in his budget can say as much.

There is a second number he did not expect to be the interesting one. The people who came back are opening at a rate materially better than his existing engaged base — not because they are better customers, but because they arrived through a programme that had to earn every open from a standing start and never got to lean on habit or goodwill. The dormant half, treated properly, produced a better relationship than the half that never left.

He can also now answer a sequence of questions rather than hide behind a single metric. Did the recovery programme re-earn attention. Did the interaction produce something worth doing. Did claiming a card raise the odds of a later return. Did anyone choose to keep it. Did any of this damage the sending reputation or the engaged base. Each of those has an answer, and the answers point in the same direction.

The finance conversation is short, which is not something he is used to. He is not asked what the attribution window was, whether the uplift is incremental, or how much would have happened anyway. The design answered those before the programme started.

The only argument left is whether to run it on the other hundred and seventy thousand.

Key points

• The holdout rate is the number the reactivation argument has always been missing.
• State the contamination on the page; it applies to both halves, which is the point of splitting them.
• Recovered users can outperform the never-lapsed base, because nothing about their attention was assumed.

8

Month seven — The album

Meera

Seven months later Meera has finished the careers set and mostly stopped thinking about the careers site, which sends her something twice a week that she opens perhaps half the time. By the standards of any brand she is an excellent customer. By the standards of her own attention it is a small habit that costs her nothing.

But the album now holds four sets, and only one of them came from that first brand.

It reads on three numbers rather than one, which is why it has not turned into a scoreboard she ignores. She holds twenty-two cards. She has finished two of the four sets. And she can still answer for fourteen of the twenty-two — a figure that went down in the fortnight she was travelling and came back up afterwards, because the album noticed and put the slipping cards back in front of her. The first number is the one any brand can move. The third is the one nobody can move on her behalf.

Figure 6. Four brands, one album, one balance, three numbers. The gaps are what bring her back.

A coffee company issues cards about origins and roasts, and she is at five of nine. A travel brand issues cities, and she is at three of twelve and unlikely to finish, which does not bother her. A financial app issues a set about how various things work, which she completed in a fortnight because it was better than the explainers she had been searching for. One of those cards she sent to her father, who had been asking the same question for a year.

The travel set is the interesting one, because she started it four months late and it made no difference. Nothing had been withdrawn, no first printing had closed, and the nine cards she had missed were reachable at the same cost as the three she held. Had the brand run the obvious promotion instead — a limited edition, gone for good after the first month — she would have opened the message, discovered she was too late to complete anything, and never opened another. Almost everybody arrives after the beginning.

None of those four companies knows about the other three. Each sees a customer who opens their email, which is all any of them wanted.

The fourth one had an easier start than the first, and nobody at that company will ever know why. When its first card arrived, Meera was not a cold user being asked to understand a new mechanic. She already had a balance, already knew what those sixty seconds were, and already had somewhere for the card to go. The cost of teaching her was paid once, months earlier, by a careers site she had ignored for eight months. Every brand after that inherited it.

What none of them can build alone is the thing she ends up with: one album with four sets in it, some finished and some not, a handful of duplicates she is trying to move, and a running conversation in two different groups about who needs what. The careers set on its own would have been a nice campaign that ended. Sitting alongside three others in a container she owns, it is still working seven months after the email that issued it stopped mattering.

That is the whole trade, seen from her end. She was never asked to join anything, never told about an architecture, and never gave any brand more than it earned. She answered three questions on a Tuesday because the subject line said something was hers, and everything after followed from the fact that the card did not disappear when she closed the message.

Key points

• One album, four sets, four brands that know nothing about one another.
• The fourth brand met a warm user on its first send. The first brand paid for that and never knew.
• A campaign card ends with the campaign; a card in an album is still working seven months later.
• She joined the travel set four months late and it cost her nothing. Almost everybody arrives after the beginning.
• She joined nothing and understood nothing about the architecture. That is the design working.

9

What the two stories are showing

Read the seven months from Meera’s side and it is a small pleasant habit that arrived without being asked for. Read them from Rohan’s side and it is the first defensible number he has produced against a segment everybody had written off. Both are true, and neither could have happened without the other.

That is the point worth taking from the pair of them. The consumer experience is not a wrapper around the commercial mechanism, and the commercial mechanism is not a tax on the consumer experience. Meera returns because the set is incomplete. Rohan can prove it because the half he held back did not return. The same fact serves both ends.

Five things in the story would break if they were done in the obvious order instead.

If the first message had carried an advertisement, Meera would have closed it and Rohan’s chart would be flat. Earn first, monetise second is not a principle about taste. It is the difference between the two lines separating and the two lines staying together.

If the card had lived on the careers site’s own page, month seven would not exist. The set would have ended when the campaign ended, and the four brands in her album would be four separate logins she never uses. The brand gives up the container to get the seven months.

If the balance had belonged to the brand rather than to Meera, the subject line on that first Tuesday would have been a loyalty number and she would have deleted it with the rest. And the fourth brand would have had to start from zero, the way the first one did.

If the card had been the reward for the questions rather than the thing the questions opened, the first Tuesday would have been a survey with a prize at the end, and Meera has ignored a great many of those. The seal is why she read to the bottom of a message from a company she had not opened since February.

And if Rohan had run the programme without holding anyone back, he would have a reactivation number today and no way to know whether it meant anything — which is exactly the position every one of his predecessors was in, and the reason the dormant third of the database sat untouched for two years.

She came back for the set. He could prove it because of the holdout. Neither of those is the interesting part on its own.

The architecture essay argued that each link in the chain exists because the one before it was insufficient. This is what that looks like when it is happening to somebody. A message worth opening. A minute worth spending. An object that survives the minute. A container that survives the brand. A group that makes the container social. And underneath all of it, a number in a subject line that told a woman on a train that something in there was already hers.

Key points

• The consumer mechanism and the measurement mechanism are the same mechanism seen from two ends.
• Monetise early and the chart flattens; own the container and the relationship ends with the campaign.
• A balance that belongs to the brand is a loyalty scheme. One that belongs to the person is a reason to open.
• Without a holdout there is a number, and no way to know what it means.

How brand email builds the consumer network

The consumer network does not begin by acquiring consumers. It begins with brands sending better email to people they already know.

1

The inbox is being sorted, not shrunk

The easy conclusion about agents reading our mail is that they will reduce the attention we give to email. The more interesting conclusion is that they will change what deserves it.

Think about what fills a typical inbox. Order confirmations. Statements. Delivery notices. Offers. Coupons. Password alerts. Product recommendations. Nearly all of it is informational, and a capable agent can read it, summarise it, compare it and act on it. In most of those cases the human was always the least efficient part of the loop.

That does not make the inbox irrelevant. It makes the inbox more selective.

An agent can understand a promotion. It cannot care about a collection. It can compare prices. It cannot want the card it is missing. It can summarise a newsletter perfectly well, and it cannot feel the small satisfaction of completing a set, keeping a run going, remembering yesterday’s answer, or having the spare that a friend needs.

So the inbox divides rather than empties. On one side is mail that is informational — useful for what it tells us, or for what software can do with it on our behalf. On the other is mail that is experiential — useful because a person wants to see it, play with it, collect from it, learn from it, or come back to it. Most brands today live almost entirely in the first lane. The opportunity is to build the second.

Figure 1. The agentic inbox creates two lanes. Software takes the left. Almost nothing is being sent into the right.

This is a thesis about direction rather than a description of a settled market. But it becomes urgent because of a second pressure arriving at the same time.

Brands are losing their owned relationships. Lists keep growing while the share of the list that still pays attention keeps shrinking. Someone who once opened, clicked and bought becomes an address that remains in the database while the attention behind it has gone. Eventually the brand reaches for rented media and pays again to reach a person it already knows. That is AdWaste in its simplest form: re-buying attention the brand once owned. The striking thing about it is not the cost but that it is voluntary. It became normal only because the owned channel stopped being worth opening.

And the alternatives are getting less comfortable. Messaging showed brands the power of a high-attention push channel, and it also demonstrated the economics of depending on a channel somebody else controls. The owner sets the rules, the access and the price. Every brand that built its retention there is a tenant, and the rent is decided elsewhere.

Email is structurally different. No single company owns it. A brand holds the address and the permission, chooses its provider, changes its technology, and goes on reaching the same person. In a world increasingly built on rented surfaces, that independence gets more valuable.

But independence alone does not create attention. Email is not destined to win because nobody owns it. It wins only where there is something inside the message that a human wants enough to come back for.

That is the real problem, and it is not the one the industry has been working on. Two decades of email optimisation have gone into deliverability, personalisation, subject lines, send-time, frequency, templates and journeys. All of it matters. None of it answers the question underneath: why would a person form a habit around a brand’s email?

A habit requires anticipation. The reader has to believe that opening tomorrow’s email gives them something tomorrow that they cannot get by ignoring it today. That something cannot always be a discount. Discounts consume attention. They do not build it.

The objective is not a better campaign. It is inbox attention as a habit.

A brand email has to become a small appointment: useful enough to open, interactive enough to take part in, persistent enough that yesterday affects today, and rewarding enough that tomorrow matters.

Once that is the goal, the pieces connect. The Digest earns the open. The Magnet earns participation. The Card lets the moment persist. The Album gives the Card a home. Mu gives participation a memory. Small groups turn solitary progress into something social. And a wider world gives all of it somewhere to grow.

The chain matters because each element exists to solve the insufficiency of the one before it. Anyone who disagrees with the argument should be able to say exactly which link fails.

The consumer network is not the starting point. The first problem is much simpler than that. Make the next brand email worth opening.

Key points

• Agents will absorb informational mail rather than empty the inbox; attention concentrates on what software cannot enjoy on our behalf.
• Brands already hold identity and permission, and are paying rented media to reach people they already know.
• Messaging is a rented channel with a price set by its owner. Email is the one nobody owns.
• Independence does not create attention. Discounts consume attention rather than build it.
• A habit needs anticipation — a reason to believe tomorrow’s email gives something today’s neglect cannot.

2

SEND: four kinds of email, one of them missing

Every brand has a database, and every database divides the same way. There are people who still respond, and there are people who have stopped. Most companies know both numbers, and most companies have quietly written off the second group — not by deleting them, but by continuing to send them the same thing that stopped working and hoping for a different result.

That division is where this architecture starts, because the two halves need different email, not different subject lines.

Brand email has four jobs. I call them SEND: Sell, Engage, Notify, Digest.

Sell is the commercial email every marketer knows — offers, promotions, recommendations, renewals, conversion nudges. It asks the reader to buy. Engage is tied to a moment: welcome, onboarding, pre-purchase help, post-purchase guidance, renewal preparation. The customer does something and the brand responds. Notify is functional: a transaction happened, an account changed, an order moved. These carry trust because the reader expects them and often needs them.

Digest is the different one. It is recurring rather than triggered, editorial rather than transactional. Its purpose is not to report what happened or push what the brand wants sold. Its job is to make the relationship itself worth revisiting.

Almost every brand runs the first three well. Very few have built a fourth worth opening repeatedly. And that gap gets more expensive in an agent-mediated inbox, because Sell and Notify sit naturally in the agent lane, Engage sits on either side depending on the moment, and the Digest is the only one that unambiguously belongs to the human.

A good Digest asks a different question from a campaign brief. Not what do we want to say this week, but what could we send regularly that leaves this person better off even if they buy nothing?

For a jobs brand that might be a salary puzzle, an interview judgement call, or a skill card. For a financial brand, a concept a week, a market explainer, a decision exercise. For travel, a destination clue, a language card, a cultural fact. For a retailer, a materials guide, a styling challenge, a piece of product knowledge worth having. The brand is still present in all of these. It is simply present as editor, coach or guide rather than as seller.

It is worth separating SEND from a second four that sits alongside it. EARN — Email, Act, Run, Network — describes the ladder of accountability a provider climbs, from delivering the message to being answerable for the outcome. SEND describes what the brand emits; EARN describes what someone can be held to. Two different views of the same business. No reader should go looking for a correspondence between them.

The Digest has two configurations, because the two halves of the database do not start from the same place.

Standard goes to the engaged half. These people still open, still click, still interact. Standard can carry the fuller experience: two Magnets, useful brand content blocks, and, once attention has been earned, carefully governed monetisation. It runs at whatever daily or near-daily cadence the brand can sustain, because that is what an appointment requires.

Mini goes to the dormant half. It is narrower, not weaker. One Magnet — the single strongest reason that individual has to come back. No third-party monetisation. No clutter. No attempt to extract value from the first recovered moment.

Mini is also mini in days, not only in length. A dormant address cannot absorb a daily programme, and sending one is precisely how a sending domain gets damaged: mailbox providers watch for repeated delivery to people who never engage, and they act on it. So Mini runs at a lower frequency, and it runs with an exit condition — after a defined number of unanswered sends, the address stops receiving it. Restraint here is not politeness. It protects the reputation that the Standard programme depends on.

The two run in parallel. The brand does not choose one or the other; attention state chooses the configuration, and people move between the two as their behaviour changes. Dormant, recovered, engaged — and, when attention decays again, back.

Figure 2. Standard and Mini are the same architecture applied to two different attention states.

Mini also creates an unusually useful measurement environment, and this is the part a finance director will care about most.

Dormant audiences are where conventional marketing has already reduced effort or stopped trying altogether. That means the current best effort against them can be very low, and is sometimes no active treatment at all. Split the cohort at random, treat one half, hold the other back concurrently, and the comparison is against what the brand would otherwise have done, in the same period, under the same conditions. Never against a prior period, which measures the season as much as the intervention.

That will not be perfectly clean. Customers can be exposed elsewhere, return organically, or be reached through paid channels, and an honest programme measures with that in view. But the experiment starts from a far less ambiguous place than most marketing measurement ever does. Reactivation claims are usually buried under attribution argument. Mini makes them testable.

Inside both configurations sits the same repeating unit, the attention processing unit: the Mu count in the subject line, the Magnet, and the Mu ledger.

The subject-line count does a subtle job. Before anything is opened, it tells the reader that something of theirs exists inside. This is not another promotional subject line competing on adjectives; it carries state. In an agent-mediated inbox it picks up a second role, because visible personal state is a signal that the message was built for the human rather than for the software reading on their behalf.

Then the Magnet, which earns the next action. And underneath, the ledger, which remembers what the reader has earned and done.

One point about that count matters more than it first appears. It is not a balance the brand issues and the brand owes. It is the reader’s own, accumulated across everything they do, and the brand is showing it back to them rather than granting it. That is why it reads as recognition instead of as a loyalty scheme, and it is also why the number is already there on the day a brand starts.

Mu does not need to appear in everything a brand sends, and it does not belong automatically inside Sell or Notify. The claim is narrower than that. The Digest has a recurring architecture that can make attention cumulative rather than disposable, and almost nothing else a brand sends can.

Key points

• Every database splits into a responding half and a dormant half. They need different email, not different subject lines.
• SEND = Sell · Engage · Notify · Digest. Only the Digest creates its own appointment.
• Standard serves the engaged half at sustainable daily cadence; Mini serves the dormant half less often, with an exit rule.
• Mini’s restraint protects sending reputation, which the Standard programme depends on.
• Dormant reactivation admits a concurrent randomised holdout — far cleaner than conventional attribution, though never perfectly clean.
• The attention processing unit is the Mu count in the subject, the Magnet, and the Mu ledger.

3

The Magnet, and the line it may cross

A Digest can be useful and still be passive. The reader opens, reads, perhaps learns something, and leaves. That is better email. It is not yet a different architecture.

The Magnet is what changes it. A Magnet is a short interaction — thirty to sixty seconds — that leaves the reader better off than they were before they opened. It might test recall, reveal a useful fact, ask for a judgement, offer a puzzle, benchmark someone against their peers, or ask for a prediction. It is not a banner. It is not a survey wearing engagement as a disguise. And it does not exist to collect data for the brand. The reader gets something first.

Consider someone who stopped engaging with a jobs brand eight months ago. Nothing dramatic happened; they simply stopped opening, and the brand stopped expecting them to. Another discount or another vacancy list will not restart that. But a Mini arrives on a Tuesday with a small number in the subject line — their own Mu balance, which they had forgotten they had — and a line saying that card three of eight is waiting, face down.

They open it, which they have not done since February. Inside is one thing: three questions about salary bands in their function and their city. They answer, get one wrong, learn something they will repeat at work that week, and the card turns over.

No application was demanded. No form was presented. No transaction was required. The brand has achieved the only thing that mattered on that Tuesday, which is that attention was re-earned.

And then the architecture hits its next limit. A Magnet is a moment, and moments disappear. The reader may enjoy Tuesday, but if nothing survives it, Wednesday starts from zero. Habit requires consequence: yesterday has to matter today.

That is why the Magnet has to be able to issue an object. The object is the Card, and the order in which the two arrive turns out to matter more than it looks.

The obvious sequence is Magnet then Card: answer three questions, receive a card as the reward. That works, and there is a better version. The card arrives face down, and answering is what unseals it. What the reader can see before they act is the set, the position in the set, and nothing else; what they cannot see is which card it is. A wrong answer still opens it, with less credit attached.

Two things follow. The reader now has a reason to act that has nothing to do with the brand’s generosity — nobody taps to receive a picture, and everybody taps to find out what is under a seal — which is also why the fallback click gets stronger rather than weaker where the inbox cannot render the interaction in place. And the three separate things the architecture was asking a brand to deliver in sequence become one moment: a reason to open, an interaction, and an object.

The card is not the reward for the Magnet. The card is what the Magnet opens.

Figure 3. The Magnet does not earn the card. It opens one — a reason to open, an interaction and an object in a single moment.

This does not change what the brand supplies. It still chooses the subject, still writes the interaction, still decides what it is willing to issue. It changes only the order in which the reader meets them, and the Card still creates the most important boundary in the whole architecture.

The brand issues the Card; the consumer owns the Album.

Cards move between people. Mu never does.

Those two sentences are the hinge. Everything before them belongs to the brand. Everything after them belongs to the consumer.

The Magnet crosses that line because it is a format rather than a possession. The same kind of interaction appears on the brand’s side and again on the consumer’s side, so what travels is a behaviour, not an account or a permission asset.

Mu is different again, and the distinction is worth being exact about because it is easy to get wrong. Mu is not a brand’s currency held in a brand’s scheme. There is one balance and it belongs to the person, earned wherever they act — in one brand’s Digest on Tuesday, in another’s on Thursday, in their own collection at the weekend. It is a record of what they have done, not a liability any brand carries.

There is a second rule about Mu that decides whether the number means anything, and it is easy to get backwards. Mu accrues on recall, not on receipt. Holding a card earns nothing; being able to answer for it weeks later earns something. That is what stops the balance becoming a measure of how much mail somebody has been sent, and it is why a brand cannot inflate a reader’s number by issuing more cards. The only way the figure moves is that a person remembered something without being shown it first.

Which is precisely why it never moves between people. A card can be given away; that is what makes a collection social. A balance cannot, because a balance that can be handed over stops being a record of anything and becomes something else entirely — purchasable, poolable, and a regulated instrument in most places worth operating in. Keeping Mu attached to the person who earned it is what keeps it a score rather than a token, and the boundary that matters is the one between people, not the one between the brand and the consumer.

The same discipline governs the relationship itself. A brand can invite. It cannot silently enrol. The reader who unlocks a Card chooses whether to keep it and continue, and that affirmative act is what creates anything on the consumer side. Nothing in the architecture requires a brand’s list to move anywhere.

This matters because the commercial objection is obvious and entirely reasonable. A brand that has spent years and real money acquiring customers will not willingly become an acquisition channel for somebody else’s database. The architecture survives only because the crossing is built on consumer choice rather than audience transfer. The brand issues the object. The consumer decides whether to keep it.

Figure 4. The brand issues, the consumer owns, and one balance runs under both. Cards move between people; Mu never does.

There is a second change, and it is about timing. A conventional email is decided at send. A card-based experience has to know what the reader holds now: which cards are already there, which are missing, whether a challenge has been completed, what balance exists, what the next useful object should be. That has to be composed at the moment of opening, for that person, against their own history.

The components for dynamic email existed long before anyone found this interesting, so the honest answer to why now is not that it became possible. It is that it became affordable. What has changed is the economics of composing and operating highly individualised experiences at the moment of attention — deciding far more of the message per person, at open, without a human team pre-building every branch. Doing this for one customer was never hard. Doing it for several million at a cost that leaves the programme worth running is recent.

The other constraint is more mundane and should be stated rather than buried. Interactive rendering inside the inbox is supported by some mailbox providers and not others. Where the mailbox supports it, the Magnet is completed in place. Where it does not, the same interaction falls back to a hosted surface reached by a click.

There is a sequencing consequence that a brand should hear before it plans a launch rather than after. Sending the interactive version at all requires registering with each mailbox provider, and registration requires a demonstrated record of low complaints — which a new sending identity does not have on the day it starts. So the first months of a programme run the hosted version by default and move the interaction into the inbox once the reputation exists. The architecture does not change. The order of delivery does, and planning for it is cheaper than discovering it.

That fallback is not merely a compromise. It can be the stronger consent moment, because the reader has deliberately stepped out of the message to claim something persistent. The architecture is inbox-native where it can be and gracefully portable where it cannot. The inbox owns the moment of return; the state survives beyond it.

Key points

• A Magnet turns passive reading into participation, and must leave the reader better off rather than extract from them.
• A Magnet alone is ephemeral, which is why it has to issue something that persists.
• The card arrives sealed and the Magnet is what opens it — one moment instead of three.
• The hinge: the brand issues the Card, the consumer owns the Album. Cards move between people; Mu never does.
• Mu is one balance belonging to the person, earned across every brand — a score, not a brand’s currency.
• Mu accrues on recall, not on receipt. A brand cannot inflate it by sending more.
• A brand can invite but cannot silently enrol. The crossing is consumer choice, not audience transfer.
• Composing per person at the moment of attention became affordable, not possible — that is the why-now.
• Inbox-native where the mailbox allows; a hosted claim surface where it does not, which is a stronger consent signal.

4

The Card, the Set and the Album

You unlocked today’s Card. Add it to your Album.

That sentence matters because the object does the explaining. The consumer does not need to understand attention architecture, incentive design or cross-brand networks. They understand a card.

Cards carry a deep cultural advantage: we know what they mean without instruction. They can be collected, completed, compared, remembered, traded and missed. A card creates scarcity without requiring money, and progress without requiring a leaderboard. It gives a small interaction an afterlife.

That is why the Card is the smallest unit of attention that can be accumulated — small enough to be issued daily, durable enough to still be there next month.

The Card stays linked to the brand that issued it, and that constraint is not negotiable. If a brand’s Digest starts handing out unrelated third-party collectibles, the Card is simply advertising inventory wearing a costume, and the brand has no reason to support it. A brand-linked Card is different: it is the brand’s own contribution to the consumer’s collection, drawn from what the brand knows better than anyone.

A jobs brand issues careers, skills, industries, interview judgement. A travel brand issues destinations, monuments, foods, languages. A financial brand issues concepts, principles, historical market events. A retailer issues materials, craft, style, product knowledge.

Choosing what to issue is a harder problem than it sounds, and it is where most collection products fail before a single card is drawn. Three tests do most of the work. Can the reader name three members of the set instantly and then stall — because that gap between recognition and recall is the set, and a list anybody can complete from memory has nothing to offer. Does every card teach something of a different kind, or does card three teach the same shape as card nineteen, in which case the set dies about a third of the way through however good it looks. And do the members relate to one another at all — can they be compared, ordered, ranked, combined — because a set whose items have no relationship can only be collected, and collecting alone runs out.

A set also has to have a visible edge. Eight cards, or twelve, or twenty-two, stated at the outset. An open-ended series of cards is a newsletter with pictures on it, and it produces none of the behaviour this section is about, because there is nothing to be missing from.

The strongest cards do more than entertain. They leave behind knowledge, memory or judgement — something the consumer ends up proud to know rather than merely proud to possess. That is also why this works in categories where promotional creativity is tightly constrained. Explainers, definitions and factual series are not a lesser version of the idea. They may be the strongest version, because what the consumer gains compounds.

Cards need not live only in email. Someone might meet one on a brand’s website or inside its app, and they should. But the surfaces do different jobs, and confusing them is how collection products usually fail. Email owns daily progression — it is where the next card arrives without being asked for. The web owns archive and provenance — it is where a collection is browsed, sorted and shown. Group messaging owns invitation and social coordination — it is where people tell each other what to look at.

Then the Album, which is where the architecture changes category.

A brand can issue Cards. It should not own the Album. The analogy that holds all the way down is the stamp album: countries issue stamps, the collector owns the album, and a stamp does not vanish because a country changes its policy or stops printing a series.

That gives the structure its shape. A person has one Album. Inside it are Sets — one from each brand that issues, sometimes more than one from a single brand. Inside each Set are Cards. One album, many sets, many cards. There is no per-brand album and no separate container above the album, because the album already is the container. A collector does not manage albums; they have one, and they fill it.

Figure 5. One album per person. Sets inside it. Cards inside those. The gaps are the point, and the album reads on three clocks.

From that structure comes the grammar of the whole thing: Got, Need, Give. Got is what I hold. Need is what is missing. Give is what I hold twice.

The album reads that grammar back as three numbers rather than one, and the reason is that a single progress figure eventually discourages everybody. Hold is what arrived, and it moves daily. Finish is how many sets are complete, and it moves weekly. Know is how much can still be answered for, and it moves slowly, permanently, and downwards when somebody stops paying attention. Three satisfactions on three clocks. A brand looking at its own set sees the first two; the third is the one that tells anybody whether the thing worked.

Hold moves daily. Finish moves weekly. Know moves slowly, and it can fall.

A single card is content. A set of cards is a collection. A collection with gaps creates anticipation. A collection with duplicates creates trade. None of that works inside a gallery page owned by one brand, because a set nobody else is collecting has nothing to complete and nothing to exchange. The value of the container comes precisely from its persistence across relationships.

Figure 6. Brands issue the Cards. The Album belongs to the person holding it.

Which produces the most uncomfortable commercial fact in the architecture: the Album is not the brand’s.

That is not an implementation detail to be softened in a later meeting. It is the bargain, and it should be put to a brand plainly at the start.

The brand gives up ownership of the container in exchange for greater persistence of its relationship inside the container.

A campaign disappears when the campaign ends. A card sitting in the consumer’s album continues to exist. It gets seen again, completed around, discussed, compared, traded and remembered long after the email that issued it has gone. The brand’s relationship gains durability precisely because the brand does not control the whole environment.

One discipline protects that durability and it is worth writing into the programme early, because the marketing instinct runs against it. Nothing in a set is ever withdrawn. A brand can stop issuing new cards, run a fortnight that concentrates on one part of a set, or close a set and start another. What it should not do is put a shutter on a card, because a rarity that expires manufactures regret in everybody who arrives afterwards — and almost everybody arrives afterwards. Scarcity, where a brand wants it, belongs in the particular copy: where it was earned, at what level of recall, and whether it came from another person. Those are renewable. A closing date is not.

That is a hard trade for conventional marketing thinking, which assumes value rises with ownership. Here, partial ownership produces more persistence than total control. A brand-controlled loyalty page contains the brand’s objects, and the consumer visits when they happen to remember the brand. A consumer-owned album reverses the direction: the consumer visits because the album matters to them, and meets the brand again because the brand’s card is part of something larger they value.

Key points

• Cards turn ephemeral attention into persistent objects, and need no explanation to be understood.
• Cards stay brand-linked, or they are advertising inventory the brand did not sell.
• The strongest cards leave knowledge behind; constrained categories get the full version, not a degraded one.
• Email owns daily progression, the web owns archive and provenance, group messaging owns invitation.
• One Album per person. Sets inside it. Cards inside those. No per-brand album, no container above the album.
• Got · Need · Give turns a collection into anticipation and exchange.
• The album reads Hold · Finish · Know — three numbers on three clocks, and the third can fall.
• Nothing in a set is ever withdrawn. Scarcity lives in the copy, never in a closing date.

5

From collection to habit

An album makes yesterday matter today. But collecting, at first, is solitary, and solitary habits are fragile. They run on interest alone, and interest fluctuates.

The next layer changes the motivation from I want to complete this to we are doing this together. Small persistent groups create that shift, and the principle matters far more here than the mechanics.

A group gives an individual collection a social context. Someone else may hold the card I need. I may hold the spare they want. We can compare progress, help each other finish a set, recognise who knows the subject, or coordinate around a shared challenge. The important behaviour is not share this promotional message. It is a conversation about an object both people value.

That changes distribution in a way referral mechanics never manage. Most referral schemes ask a user to interrupt their friends on a company’s behalf and pay them a coupon for the trouble. A collection produces a more natural prompt, because the social state is incomplete without the other person. Did you get today’s card. I have got one you need. Our group is missing this one. Check your inbox.

Groups also create obligation, and obligation is what makes a daily habit durable. A private streak is easy to abandon. A shared ritual is much harder, because missing a day now costs somebody else something. This is the same force that sustains book clubs, fantasy leagues, study groups and most multiplayer games: the individual action acquires a social consequence.

A person should be able to belong to several such groups rather than one. The sets someone collects cut across different parts of their life — work, family, the friends they argue with about sport — and the people holding the cards they need are not all in the same room. A single group would make most trades fail. What matters is that each group is small enough for members to matter to one another, persistent enough for reciprocity and reputation to form, and connected enough to turn private progression into conversation.

Beyond that, the mechanics should stay unspecified, and deliberately so. How groups form, how they grow, who governs them, what happens when one goes quiet — those are real decisions and they are not settled. Publishing a specification now would freeze choices that deserve to be made against evidence rather than against a diagram.

Mu runs through this layer too. It signals participation, records that a person did something, and provides a common measure of progress across everything they collect. But Mu is not the reason anybody stays. If the underlying experience is weak, no points system rescues it. The Card creates the object, the Album creates persistence, the group creates social meaning, and Mu records the movement.

Once that exists across many collections, another insufficiency appears: why stop at cards? A persistent consumer identity with collections, earned progress and real social relationships can carry many kinds of experience — knowledge challenges, recall games, predictions, cooperative quests, timed events, longer journeys towards mastery.

Those should not become a scattering of disconnected products, each with its own audience to acquire and its own social graph to rebuild. They belong inside the same world. The album is the first vehicle into it because collecting is easy to understand and naturally persistent, but the album is not the destination. A fixed-time daily challenge becomes a cadence inside that world rather than a separate product beside it.

Which finally clarifies what email is doing here. The inbox does not need to contain the whole world. It is the daily window into it. Something has changed. Something is waiting. A card is available. A challenge has opened. A group needs you. A set is one away from complete.

Old email contains the content and hopes the reader clicks. The new email reveals the next state of something the reader already cares about.

Key points

• Solitary collecting is fragile; obligation to a person is not.
• The sharing trigger is about the collection, not about promoting the brand — which is why it travels.
• People belong to several small groups, not one, because their sets cut across different parts of their life.
• The mechanics stay unspecified here on purpose; they are not settled and should be decided against evidence.
• Mu records movement but never substitutes for an experience worth returning to.
• The inbox is the daily window into a state that persists between messages.

6

What the brand gets

At this point a CMO has an entirely fair question. Why should a brand fund a Digest, issue Cards, and help build an Album it does not own?

The answer cannot depend on the consumer world eventually becoming large. The Digest has to justify itself before any of that happens. It does.

The first benefit is recurring attention the brand does not have today. Most brand email spends attention rather than building it. Promotions ask for money. Notifications report something that already happened. Triggered journeys attach to moments that end. The Digest creates a recurring relationship that is independent of an immediate transaction, and a brand that is useful between purchases stays mentally available for the next one.

The second is reactivation with a measurable counterfactual. Mini gives a dormant customer one strong reason to come back, and because the audience starts from low engagement the experiment can run against a concurrent randomised holdout. The brand finds out whether the intervention restored behaviour instead of assembling an attribution story afterwards. Standard strengthens a live relationship; Mini tests whether a dead one can be restarted. Very few things in a marketing budget can be tested that cleanly.

The third is economics, and the sequence is the whole of it. Once a Standard Digest has earned recurring attention, that attention can carry monetisation — selected partner value, action-led placements, transactions. Reverse the order and the Digest becomes another advertising vehicle pushed into a database the brand already owns, and the attention collapses before the economics arrive. Earn first. Monetise second. Never in Mini at all.

The ambition there is not merely to reduce the cost of the email. It is to change the economics of the channel, so that the message funds its own delivery rather than sitting on the budget as a line item paid per send.

The fourth is survival in the agentic inbox. If informational mail is increasingly handled by software, a brand needs a class of email that humans still choose to experience. That is what the Digest is for. The Mu count signals state before the open. The Magnet creates participation. The Card creates anticipation. The Album gives today’s interaction consequences tomorrow. Software can summarise the message perfectly well. It cannot complete the collection on anybody’s behalf.

The limits should be stated as plainly as the benefits, because they are design principles rather than hidden weaknesses.

Interactive rendering is uneven, so the experience is inbox-native where the mailbox permits and degrades to a hosted surface where it does not. A meaningful share of any base will meet the crossing as a click rather than a tap, and the programme should be planned on that basis rather than surprised by it.

And the Album is not the brand’s. That is the second limit and it is also the source of the largest upside, which is why it should be put on the table in the first conversation rather than the fifth. The brand trades control for durability. Not every brand will accept that, and the first programmes will show which categories, which collections and which kinds of value make the trade worth making.

There are no numbers in this essay, and that is deliberate. How many open, how many complete a Magnet, how many claim a Card, how many return the following week — these are knowable only from a programme that has run. Quoting them earlier would be quoting an assumption. They come after the first one works, not before.

Key points

• Recurring attention independent of a transaction, which almost no brand has today.
• Reactivation measured against a concurrent randomised holdout rather than an attribution story.
• Earn attention first, then monetise, so the message can fund its own delivery.
• A class of email that survives agentic filtering because a human chooses to read it.
• The limits are real and stated up front: uneven rendering, and an Album the brand does not own.

7

What would have to be true

The architecture does not need speculative performance claims to deserve a test. It needs a chain that can be falsified, and this one can be, link by link.

Does the Digest earn repeat human attention? Does the Magnet improve the open by giving the reader something worth doing? Does issuing a Card increase the chance of tomorrow’s return? Does an Album create persistence beyond a single brand? Do small groups produce natural distribution? Does the wider world make the inbox a recurring window rather than another feed?

One of those links is weaker than the others and it should be named rather than buried in the list. The chain assumes somebody wants the set. Tests can reject the obviously flat candidates before anything is built, and a filter is a way of failing less often rather than evidence that the survivors are wanted. The first set a brand issues will be opened because the format is unfamiliar. Whether the third one is opened is the question that decides whether any of this is a programme or a novelty.

Each of those is a real question with a real answer, obtainable in a quarter rather than a decade. If the chain breaks, the architecture says exactly where. That is a more useful property than confidence.

EARN is the business architecture. This is the attention architecture.

And if it holds, the thing at the end is not an email template, a points balance, or a game. It is a cross-brand collection held by the consumer.

No single brand can build that alone, and the reason is structural rather than competitive. One brand issues a Card. Another issues another. Each contributes a small object drawn from its own expertise and its own relationship. The consumer keeps them together because the whole collection is more useful, more interesting and more complete than any brand-specific version could ever be. A set that only one company is issuing has nothing to trade against and nothing to complete beyond itself.

The same is true of the balance underneath it. No single brand can build a currency that means anything across a person’s whole commercial life, because a points balance confined to one company is a loyalty scheme and everybody already knows what those are worth. But it works in the other direction too, and this is the part that compounds. When the second brand issues its first Card, it does not meet a cold user. It meets somebody who already carries a balance, already knows what a Magnet is, and already has an album for the card to go into. The cost of starting is paid once, by the first brand, and every brand after that inherits it.

It follows that whoever ends up holding that container will not be any one brand. It also follows that it does not get built by asking brands to be more interesting, which the industry has tried for a decade without success.

It gets built one Digest at a time, by brands solving a problem they already have: a dormant half of the database worth recovering, an engaged half worth keeping, and a lane in the inbox that software is not going to read on their behalf.

Figure 7. The chain of insufficiencies. Each link exists because the one before it was not enough on its own.

The network does not begin by acquiring consumers and then selling access to brands. It begins with brands sending better email to people they already know.

One habit at a time. One card at a time. One album at a time. That is how a consumer network gets built.

Key points

• Every link in the chain is falsifiable in a quarter: open, participation, return, persistence, distribution.
• If the chain breaks, the architecture identifies where — which is more useful than confidence.
• The end asset is a cross-brand collection held by the consumer, which no single brand can build alone.
• It gets built one Digest at a time, by brands solving a problem they already have.

Agents will run the instances. Humans will own the outcomes. Customers will bring agents of their own. And the advantage will move from software to context, trust and accountability. Eight predictions, written down so that they can be scored — and so that they can embarrass me.

It is 8.02 on a Tuesday morning in March 2027, and Maya has opened her laptop.

There is no campaign dashboard waiting for her. Overnight, her system has moved 18,400 customers into weakening attention and 3,100 into lost. It has found a replenishment play running eleven per cent ahead of its control group and widened it. It has flagged that paid reacquisition of customers already in the database rose yesterday for the third day running. And it has escalated two decisions to her, because both of them breach the pricing guardrail she set in January.

She does not ask what campaigns are going out today.

She asks: where are we losing money?

To understand why her morning looks like that, here are eight things that get settled in 2027.

0  

Where we stand, August 2026.

Martech today has agents in it. Almost every serious platform now ships something described as agentic, and some of it is real. Salesforce is describing collaborative AI marketing teams where a marketer sets goals, budgets, guardrails and autonomy limits. Adobe has an orchestration layer coordinating purpose-built agents across customer-experience workflows. Braze is pushing continuous one-to-one decisioning across offer, channel, timing, frequency and creative rather than executing predetermined journeys. Shopify is syndicating merchant catalogues into AI shopping surfaces and reporting that AI-driven traffic to stores has grown many times over year on year. OpenAI has published a commerce protocol.

And yet almost all marketing still runs the way it ran in 2019. It runs on campaigns, on static segments, on journeys built once and left alone, on input pricing, and on rented attention. The agents have arrived; the operating model has not moved to meet them. That gap is what 2027 closes, at the frontier.

A caveat that buys the right to extrapolate. This is not a prediction that every marketing department will work this way in 2027. It is a picture of the frontier — capabilities already visible in 2026, assembled into the operating model they point towards. Maya is a leading enterprise, not the median brand.

The brake is worth applying properly, because the noise around agentic commerce is far ahead of the behaviour. Most current agentic experiences are still conversational rather than transactional, and humans still control the overwhelming majority of purchases. Consumer trust in AI remains low — recent research puts the share of consumers who completely trust it in the low teens. Anyone forecasting a wholesale transfer of purchasing to machines by next December is selling something.

So the useful question for a year this near is not what becomes possible. It is what becomes indefensible — which habits a competent CMO will find hard to justify out loud, in a budget meeting, in front of a CFO who has read the same articles.

None of these becomes illegal in 2027. Each becomes awkward to defend, which is a stronger force.

That is the frame for everything below. Not a wave of arrival. A slow withdrawal of excuses.

1  

Eight things that get settled in 2027.

What follows is not a trends list. A trends list cannot be wrong, which is why nobody ever revisits one. Each of these is written as a claim that could fail, with the test that would establish it stated alongside. I will mark them in public in December 2027, on this blog, against these words.

Resolution 1 · The campaign stops being the primary unit of planning.

The claim is not that campaigns disappear. Campaigns will still exist in 2027, in enormous numbers. The claim is narrower and more consequential: at leading brands the objective becomes the thing a human specifies, and the campaign becomes a generated execution artefact — something the system assembles on the way to the goal, rather than something a person writes and approves in advance.

The same business and the same quarter, briefed twice.

In the first panel the human has chosen the audience, the trigger and the timing, and implied the message. In the second the human has chosen the outcome, the economics and the limits, and left the rest to be determined per person. Less specifying, more governing.

The analogy is manufacturing. A factory still has production runs, but the production run is not the intelligence — the intelligence sits in the system deciding what to make, in what sequence, under what constraints. The campaign becomes the production run.

This is underway rather than speculative. Salesforce’s goal-driven marketing agent already lets a marketer define a goal, a budget, guardrails and an autonomy limit and then determines audience, content, channel and timing within them. What 2027 settles is how much of the estate moves.

How it gets scored. By the end of 2027, can a CMO at a leading brand set a commercial goal plus guardrails and have the system determine a meaningful share of audiences, treatments, timing and channels — with the campaign artefacts generated rather than authored? If the answer is still no anywhere outside a pilot, I was wrong.

Resolution 2 · The segment loses to the customer decision.

Segmentation was never a philosophy. It was a compression algorithm — the cheapest workable approximation to a problem nobody could afford to solve properly. One team could not make five million separate decisions, so it grouped people who were not alike, sent each group the average, and called the compromise a strategy.

That constraint is what has changed, so the unit of execution moves down: from segment, to customer, to individual decision. Segments survive as a lens — for reporting, governance, strategy and explaining a business to a board. They stop being the instruction.

And the most important thing now decided per person is the least glamorous one on the list: whether to contact them at all. Old personalisation meant producing more variants. Agentic personalisation can mean producing fewer interventions, because the system can see that some customers need no push. That single capability changes the economics more than any content model.

How it gets scored. What proportion of customer interventions at a leading brand are selected dynamically by a decisioning system, rather than assigned by a human-built segment or journey? If that proportion is still in single digits at the end of 2027, this resolution failed.

Resolution 3 · The CDP does not die. It disappears below the waterline.

The tempting version of this prediction is that the customer data platform category collapses. That is too strong for a single year and probably wrong in any timeframe. None of that work becomes less important. It becomes less visible. The CDP stops being the interface and becomes the substrate — indispensable, expensive to get right, and mostly invisible to the user, like the database under a modern application.

The deeper change is from profile to context. A profile tells you what is known; context tells you what matters now. Not only that she bought three months ago, but whether the item is replenishable. Not only that an offer was sent, but why it was sent, what alternative a human rejected, and what happened next.

So the buying question changes. Not which model do you use — the models will be the same models, available to everyone, at falling prices. It becomes: what does the system remember, and can the agent use it? That is a hypothesis rather than a moat anyone can declare. The same model is not the same marketer if it remembers a different history. Salesforce’s 2026 research on Indian marketers already found disconnected data limiting how far marketers trust AI to act; the bottleneck has moved from model capability to what the model is allowed to know.

How it gets scored. Do enterprise martech RFPs in late 2027 contain material sections on memory, decision traces and agent data access — as distinct from data ingestion, identity resolution and segmentation? If the RFP still reads like 2024, this was wrong.

Resolution 4 · AI-mediated demand becomes a measured channel.

This is the most important of the eight, and the one that cuts hardest against my own writing.

Every model of marketing any of us has ever drawn has two ends: the brand at one, a human at the other. The whole apparatus — persuasion, creative, subject lines, timing, relationship — assumes a person is at the far end, capable of being interested, flattered, reminded or moved.

The second diagram is the one nobody’s marketing plan is written for.

A consumer tells her assistant what she wants, in her own terms — running shoes under a certain price, comfort mattering more than weight, nothing from brands with poor returns policies. The agent discovers, compares, filters and shortlists. The brand may never get the chance to target her in the old sense at all.

Be careful about how far to push this for 2027. Direct agent-to-agent negotiation is still largely developmental; Shopify says as much itself. The defensible frontier prediction is one step earlier in the chain: that leading commerce CMOs will explicitly measure AI-mediated discovery, recommendation and transaction as a separate source of demand, with its own reporting line, rather than letting it hide inside referral or direct traffic.

The strategic consequence, though, arrives well before the volume does. If part of your audience is a machine, then part of marketing becomes making the brand legible and trustworthy to machines. Product facts. Availability. Price. Reviews. Returns policy. Reputation. Machine-readable context. None of that is advertising, and most of it is not owned by the marketing department at all.

For twenty years marketers competed for human attention. In 2027 they also compete for machine selection.

Now the part that runs against my own doctrine, and specifically against the piece of it I like most.

I have argued for a year that the way to fix email is to earn the open on the day you are not selling — the daily digest, the useful note, the message carrying no offer whose only job is to rebuild the habit of opening. That argument rests on human attachment. It assumes a person who can be pleased, who forms a habit, who comes to expect something.

My first instinct was that a machine reader kills it. The agent forms no habits and feels no warmth; it compares your product facts against five competitors in a millisecond and moves on. That instinct is too binary, and it is worth correcting carefully, because it is the difference between a real argument and a scare.

The machine has no affection. Its principal does.

An agent representing Priya knows that she prefers certain brands, trusts a particular bank, will not buy fast fashion, reads one publisher and has had excellent service from a retailer twice. Brand attachment does not evaporate because an agent mediates the decision. It becomes an input into the utility function the agent is optimising — arguably a more durable input than it was, because the agent will apply it consistently where a distracted human might not.

So the tension is not human relationship against machine facts. It is persuading the human against becoming legible to the machine that represents the human’s preferences. Those are two jobs, not a replacement of one by the other.

Which produces a more useful conclusion than my first one. Relate still matters, because it shapes the preference the agent will eventually represent. But when the purchase moment arrives, sentiment will not rescue bad price, bad stock data, bad service history or an unverifiable returns promise. Affection gets you into the consideration set. Operational truth gets you selected out of it.

That also makes marketing to agents a great deal more interesting than search engine optimisation for machines, which is how most of the industry is currently reading it.

The second gate.

There is a mechanical consequence of all this that I underplayed when I first wrote it down, and it is more immediate than the volume of agent-mediated purchasing. Today a message passes through one gate. The mailbox decides whether to accept it, on sender reputation, authentication and complaint rate — a gate every competent brand already manages, with known levers. A triaging agent adds a second gate on top, and the rule is different in kind.

The second gate is decided by the reader’s own history with that sender. Did she open the last twelve? Did she act on any of them? Has she ever replied? A brand whose base has quietly decayed does not merely get ignored more often. It stops being surfaced, which is a different and worse condition, because ignoring leaves the message on the screen and surfacing does not.

Which changes what the retention metrics are for. Click retention rate and Real Reach have been diagnostics — numbers a careful team watches to understand whether its base is eroding. In an agent-triaged inbox they become an access right. The engagement history is the thing that decides whether the next message is shown at all, and it cannot be bought at the point of need, because by then the history either exists or it does not.

Two gates, and the brand only sets the rule on the first one.

The same shift removes something the industry has never had to name, because it was free. Every programme has quietly collected attention nobody earned — the glance while deleting, the subject line read on the lock screen, the half-second before the swipe. An agent removes that residue entirely. What remains is voluntary, all of it, which raises the value of voluntary attention and takes away the floor underneath everything else.

Agents do not reduce attention. They remove the attention nobody earned.

That is also the structural reason the relationship email survives, and it is a better reason than the one I have been giving. A Sell message contains a completable task, and so does a Notify: compare, decide, confirm, track. Those are precisely what an agent exists to finish, and once it finishes them the human never needs the message. A relationship message contains no completable task. There is nothing in it for an agent to resolve on somebody’s behalf, so it either passes through to the person or it does not exist at all. Being undelegable turns out to be the property that matters, and very little brand email has it.

What follows from both gates is a design requirement rather than a positioning one. A message now has two readers and needs two payloads: something machine-actionable, so the agent can verify the price, the stock, the returns window and the provenance without guessing; and something human-experienceable, so that when it is passed through there is a reason for a person to be glad it was. Most brand email today has neither in any structured form. It has prose that assumes a human and metadata that assumes nobody.

The blur nobody has priced yet.

There is one implication hiding inside this that deserves stating, because it touches the map directly.

If customers carry agents, acquisition and retention start to blur. A customer’s agent may hold the memory of a prior relationship even when the customer does not actively recall it — that she bought from this brand before, that returns were straightforward, that size eight fitted, that delivery took two days. Your CRM has her in the lost column and has written her off. Her agent has not.

Which is a new state, and the grid has no cell for it: the brand may have lost the attention without losing the consideration. What that is worth, and whether it decays, nobody knows yet. It is the first thing I would want measured once AI-mediated demand is reported as a channel at all.

How it gets scored. By the end of 2027, do leading commerce brands report AI-mediated demand as a named channel in their own reporting — with a share of sessions or orders attached? If it is still lumped in with direct traffic, this was wrong.

A second test on the same resolution. Do any leading brands begin treating their own engagement history as a placement input rather than a retention report — suppressing sends to protect surfacing rights, or reporting Real Reach alongside deliverability? If nobody connects the two by the end of 2027, the second gate arrived later than I thought.

Resolution 5 · Outcome pricing splits into the real thing and a counterfeit.

The counterfeit will arrive first, and it will arrive in volume.

It will be marketed with exactly the vocabulary I have been using for a year — outcomes, accountability, skin in the game, pay for performance. And underneath the vocabulary the vendor will define the outcome, own the attribution model, and invoice against its own scoreboard. There will be no control group anywhere in the arrangement, because a control group is the one thing that would make the invoice smaller.

The vocabulary will be identical. The mechanism will not be.

Naming the counterfeit early is the only way to protect the real thing. Once a category has been sold a fake version at scale, the real version has to spend years arguing that it is different, and it usually loses that argument to whoever got there first with a bigger sales team.

If the vendor controls both the treatment and the counterfactual, it is not outcome pricing. It is performance-labelled attribution.

The test is that simple, and a CMO can apply it in a single meeting. Ask who defines the outcome. Ask who holds the control group. Ask whether the vendor can see and influence the counterfactual. Three questions, and every honest vendor will have prepared for them.

How it gets scored. By the end of 2027, is there a visible split in the market — with some vendors selling outcome-linked contracts backed by concurrent randomised controls and others selling attribution-model pricing under the same language? And has at least one credible buyer publicly rejected the second on methodology grounds?

Resolution 6 · Governance becomes runtime infrastructure, not a committee.

The theatrical version of this prediction is that some brand suffers a public agentic failure and explainability becomes a board-level topic overnight. That may happen. It is also unnecessary to the argument, and predicting disasters is a cheap way to sound serious.

The stronger claim is structural. By 2027, any serious autonomous marketing system needs identity, permissions, decision traces, escalation rules and a human veto as things the system executes at runtime — not as a document in a compliance folder. Adobe is already positioning its agentic architecture explicitly around governance and auditable workflows, which is a reasonable indicator of where the buying pressure is coming from.

The question a governance review asks changes accordingly. Not: what is our AI policy? But: which agent is acting and for which objective; what data, customers, channels and actions may it touch; what budget, discount, frequency and margin limits apply; which decisions need a person; where is the veto; and how do we reconstruct afterwards what it knew, what it chose, what it rejected and what followed?

Every one of those is a runtime question. None of them can be answered by a document.

That last clause is the load-bearing one. Explainability is not a compliance nicety and it is not about trust in the abstract. It is the mechanism by which autonomy gets earned. A system whose reasoning cannot be inspected does not get given more scope, and should not be. The progression — recommend, then propose, then execute with approval, then operate inside broader guardrails — is the same one every technology we have learnt to trust has walked.

How it gets scored. By the end of 2027, do enterprise agentic marketing deployments ship with runtime permissioning, decision traces and escalation as standard product surface — or is governance still a slide in the security review?

Resolution 7 · Opens lose their last claim to being an economic currency.

The temptation here is to predict that some universal verified-human-engagement standard emerges. That is too specific for a single year and I am not going to claim it.

The narrower claim, and it needs stating precisely because the sloppy version is wrong: machine-contaminated opens lose economic currency. Verified human attention remains diagnostically useful. Verified human action becomes the commercial currency.

Those are three different things and the industry keeps collapsing them into one. Privacy proxies, image pre-fetching and security scanners have made the raw open unreliable as evidence a person saw anything. But an open confirmed as human is still worth knowing — deliverability teams need it, publishers need it, and attention has value before conversion. What it cannot do is carry an economic claim.

Which produces a hierarchy rather than a single metric.

Six rungs. Real Reach and CRR live on the second; invoices belong on the sixth.

The rule that falls out of it is simple: the higher the economic claim, the higher the proof standard it has to clear. A deliverability report can rest on renders. A retention diagnostic can rest on verified human action. A supplier payout cannot rest on anything below incremental value.

There is a second reason to care about the rungs, and it did not exist when the ladder was first drawn. If a triaging agent decides what gets surfaced on the basis of a reader’s history with a sender, then the second rung stops being purely diagnostic. Verified human action is what the access right is made of. A brand that has spent a decade generating renders and calling them engagement will discover it has no record of the only thing the gate reads.

This also matters for what a decisioning system learns. Reward the machine for opens and it will get extremely good at generating opens — a failure mode with no human equivalent, because no human team was ever fast enough to fully exploit a bad metric. Reward it for a useful action, a declared preference, a state movement or an incremental transaction, and it optimises something worth having.

How it gets scored. By the end of 2027, have leading brands moved their headline engagement reporting off opens and onto verified actions? A partial test: does the open rate still appear in the board pack?

Resolution 8 · The vendor begins carrying the outcome.

The frontier martech contract moves from pay me to operate towards pay me partly for what moves.

Not all software becomes outcome-priced. That will not happen in 2027 or in 2037, and predicting it would be silly — plenty of software should be sold on access, because plenty of software delivers value that cannot be cleanly isolated. The claim is about the frontier: buyers increasingly distinguish between software that claims value and partners willing to expose some of their own economics to proof.

A fixed baseline will often remain, because delivery and infrastructure cost something before any lift exists. What enters the relationship alongside it is a share of the measured difference. The strategic change is not the percentage. It is the transfer of risk: traditional licensing hands operating responsibility to the buyer and pays the vendor whether the system was used brilliantly, badly or barely at all. An outcome-linked contract hands some execution and measurement risk back.

How it gets scored. How many major martech deals contain measurable outcome-linked components, shared baselines or incrementality clauses by the end of 2027? If the answer is a handful of pilots and nothing structural, this resolution was premature rather than wrong — but it was still wrong for 2027.

The scorecard. Eight claims, eight tests. Marked in public in December 2027.

Resolution	The test in December 2027
1 · Campaigns	Can a CMO set a goal plus guardrails and have the system determine a meaningful share of audiences, treatments, timing and channels?
2 · Segments	What share of interventions are selected dynamically rather than assigned by a human-built segment or journey?
3 · The CDP	Do enterprise RFPs contain material sections on memory, decision traces and agent data access?
4 · Machine demand	Do leading commerce brands report AI-mediated demand as a named channel, with a share attached?
5 · Outcome pricing	Is there a visible split between control-backed contracts and attribution-model pricing under the same language?
6 · Governance	Do agentic deployments ship with runtime permissioning, traces and escalation as product — not policy?
7 · Opens	Has headline engagement reporting moved off opens and onto verified actions? Is the open rate still in the board pack?
8 · The vendor	How many major deals contain outcome-linked components, shared baselines or incrementality clauses?

2  

Maya’s Tuesday.

The best way to see what changed is not to walk through a CMO’s day hour by hour. It is to look at the meetings that vanished from her calendar, and what replaced each one.

The meeting Maya had in 2026	The meeting Maya has in 2027
Campaign calendar review	Outcome portfolio review
Segment approval	Guardrail approval
Creative variants meeting	Brand judgement meeting
Channel performance review	Route-tax review
Attribution argument	Holdout readout

Each swap is small on its own. Together they change what a chief marketing officer is.

8.30 a.m. — The campaign calendar review that vanished.

The meeting used to take forty-five minutes every Tuesday. Merchandising brought the priorities, CRM brought the audience, creative brought the assets, and the channel owners argued about cadence. Someone worried that email and WhatsApp were hitting the same people. Someone asked whether the sale needed another reminder. The team left with a calendar.

That meeting no longer exists. What replaced it is an outcome portfolio review, and the difference is visible in what is on screen. Five numbers across the top: second-purchase rate, repeat margin, Real Reach, reacquisition share — a count of customers her paid channels reported as new who were already in the database, not a split of spend — and route tax. Underneath, the grid — where customers are moving between strong, weakening and lost attention, and which pools are leaking the most money.

Maya does not approve a send plan. She changes priorities. Protect the high-value customers whose attention is weakening. Grow the one-time buyers most likely to reach a second purchase. Reduce paid reacquisition of customers already known to the brand. Hold total discount cost flat. Suppress anyone likely to transact without help.

Those are management decisions. The system turns them into instances.

10.00 a.m. — Segment approval becomes guardrail approval.

Maya used to approve segments, because segmentation was where risk entered the system. Was the cohort too large? Did it include people it should not? Was it fair to give one group a benefit another would not get?

Now she approves the boundaries instead. The agent may choose from a very large action space, but it cannot invent its own economics: the margin floor, the frequency caps, the contact exclusions, the product eligibility, the protected cohorts and the actions requiring human review are all set by her team.

One decision is waiting. The system has found a group with high purchase intent and proposes a twelve per cent incentive. The model expects it to lift conversion, and the model is probably right. The expected incremental margin after the discount sits below Maya’s threshold.

She rejects it. This is the cleanest illustration of the new division of labour I can give: the model is better than she is at predicting the response. She remains accountable for deciding whether the response is worth buying.

11.00 a.m. — The brand meeting, which got longer.

This is the part that gets underplayed in every account of agentic marketing, including some of mine. Freed from operating the machine, Maya spends more time on product, pricing, story, customer experience and competitive position — the things no model can decide for her because they are not prediction problems, they are taste and strategy problems.

The pattern is familiar from every previous automation of a professional craft. When the mechanical part of a job becomes cheap, the judgement part becomes more valuable, not less. AI makes human judgement scarcer relative to everything else, and scarcity is where value goes.

1.00 p.m. — The CFO meeting, where nobody argues about attribution.

This is the meeting that has changed most, and it is the one that would most surprise a 2026 marketer watching.

There is no argument, because there is a control group. Maya and the CFO look at the baseline, the measured alpha, the carry paid on it, reacquisition share, Real Reach, customer-state movements and incremental margin. Nobody defends a model. Nobody claims a conversion that a platform also claimed. The conversation is short, because the disagreements that used to fill it were disagreements about measurement, and the measurement is no longer in dispute.

A second programme on the same screen is more interesting, because it produced no measurable alpha at all. In 2026 the team would have presented its attributed revenue and called it a win. The holdout shows that most of those customers would have come back anyway.

The CFO’s response is not disappointment. It is relief — because they have just found spend they can stop. That may be the deepest cultural change in Maya’s organisation, and it takes a year of holdouts to earn: doing nothing can now be reported as a successful decision.

Marketing finally has a language the finance function already speaks. That is worth more to the marketing department than any agent in the stack.

3.00 p.m. — The agent problem, which is not a marketing problem.

Traffic from AI shopping assistants is rising, but the brand is being selected less often than two competitors on a popular line. The product is well reviewed. Pricing is competitive. Stock is good.

The problem turns out to be returns. One channel says fourteen days, another says thirty, the marketplace copy is ambiguous, and the product feed exposes no machine-readable exception for sale items. A customer’s agent cannot establish a reliable returns promise, so it prefers a competitor whose policy it can verify.

In 2026 Maya would have asked for better ads. In 2027 the marketing fix is an operations fix: standardise the returns promise, structure it, and push it into every feed. Her job that afternoon is explaining to two departments that owe her nothing why a data inconsistency has become a demand problem.

This scene is the one to hold on to, because it is where marketing visibly stops being communications. When part of your audience is a machine that reads facts rather than claims, the truth of the operation becomes the marketing. There is no gap left between what you say and what you do, because the machine checks.

5.30 p.m. — She teaches the system.

Maya reviews five consequential decisions the system made today. A customer was suppressed because predicted organic conversion was high — approved. A recovery message used an urgency phrase that passed the brand rules and still felt manipulative — rejected, with two lines explaining why. A high-value customer was routed to a human rather than another automated offer — approved. A promotion was paused because inventory risk had changed — approved.

The fifth is the one worth watching. An agent proposed raising contact frequency for a cohort whose short-term conversion was climbing while its click retention rate was falling. Maya tells the system to protect the attention rather than maximise this week’s response — which is a judgement about the future that no reward function was going to make on its own.

Each decision and each explanation becomes part of the system’s memory.

This is her new managerial act, and it is worth naming precisely, because it looks like nothing. She is not building tomorrow’s campaign.

She is teaching tomorrow’s marketer.

3

Priya, Rahul and Ananya.

A day in a CMO’s life is only half the picture, and it is the flattering half. The other half is what all of this feels like to the person on the receiving end — which, in most accounts of the agentic future, goes conspicuously unexamined.

So: same brand, same week, three customers, three different correct answers. Two of them end with the brand doing less.

If the only illustration a doctrine can offer is a delighted customer, it is a brochure.

Priya — best, strong, and increasingly left alone.

Priya buys regularly and pays attention. In 2026 martech, that reads as high propensity, and high propensity reads as send more. She is the most contactable person in the database and therefore the most contacted.

In 2027 the system reads the same signal and reaches the opposite conclusion. She is not short of reasons to buy. Every additional message is a withdrawal from an attention account that is already full, and the model can see the withdrawal in her declining response curve long before she unsubscribes.

So her experience gets quieter. Service messages when something needs her. A composed-at-open availability alert she explicitly asked for, which is accurate at the second she reads it because it was assembled then. One useful note a week. Nothing else.

Her lifetime value rises while her message volume falls, and she would not describe any of it as marketing. She would describe it as the brand being unusually good at its job. That inversion — better marketing looking like less marketing — is the single hardest thing to sell internally and the easiest thing to prove.

Rahul — lapsed, recoverable, and the reason the holdout exists.

Rahul bought once, fourteen months ago, and has gone quiet. In today’s operating model the sequence is predictable: CRM tries, CRM gives up, paid media eventually finds him, he comes back through a rented route, and the dashboard books it as a win.

In 2027 the brand’s system knows exactly where he sits — one transaction, attention lost — and the recovery capability gets the mandate before any money goes to a platform. That capability is the one I call Team 6: the team that owns the lost column and nothing else. A brand can run it in-house, or use the outsourced version, Progency, operated by MarTech Growth Engineers working with agents. Either way it is paid for measurable improvement rather than activity, which is the whole reason the next paragraph happens.

The model believes Rahul is recoverable. His category has a reasonable repeat pattern and similar customers have come back after a year. The economics look attractive.

And because a belief is not a measurement, the cohort is randomised. Rahul lands in the control group. He gets nothing at all beyond the brand’s ordinary behaviour.

In April he comes back on his own. He needed the product again, remembered where he bought it, and returned without any prompting whatsoever.

Ninety days after randomisation. The treatment arm did not beat the control.

The treatment arm performed no better than the control. The measured lift sat inside the confidence interval. Team 6 earned no alpha and raised no invoice, and a quarter of work produced no revenue that would not have arrived anyway.

On any conventional reading, that quarter failed. Read properly, it is the most valuable thing that happened all year, because the brand has just discovered that a category of spend it was about to industrialise does not work. Every rupee it would have poured into that cohort for the next three years has been saved by one properly constructed experiment.

The system failed commercially and succeeded epistemically. In marketing, the second is rarer and worth more.

This is why prediction and incrementality must never be allowed to share a currency. The model was not wrong about Rahul — he did come back. It was wrong about its own contribution, which is a different thing, and the only instrument that can tell them apart is a control group.

Ananya — who says no.

A short note on the customer nobody puts in the deck. Ananya opens a message that is a little too knowing. Nothing in it is inaccurate; that is the problem. She narrows what the brand may use — no location, no cross-device history, no inference from what she browsed but did not buy. The system complies immediately, does not negotiate, does not offer an incentive to reconsider, and continues working with less.

Her results get slightly worse. The brand accepts that, because the alternative — treating accumulated context as a licence rather than a loan — is how a company ends up on the wrong side of both a regulator and its own customers.

Knowing more about someone grants no permission to do more. Context is not entitlement. An agentic system that cannot be told to use less of what it knows is not sophisticated; it is simply not under control.

4

The supply side’s P&L.

Everything above describes what changes for marketers and for customers. An essay that stopped there would be dodging its own hardest question, because the entire argument is a claim about how suppliers must change — and I run one.

So here is the uncomfortable version, in structural terms.

Input pricing comes under pressure, because agents reduce the human usage that seat-based and screen-based pricing was implicitly measuring. If nobody logs in, what exactly is the seat for? Value migrates from seats, messages and records towards decisions and outcomes — and the vendors who resist that migration will find their pricing model quietly detaching from the value it used to proxy.

Real outcome pricing creates working-capital exposure. The supplier does the work before the alpha is known, funds delivery upfront and collects in arrears. That caps how many engagements can run at once, for the vendor and for the client’s patience alike. Any model that does not name this constraint is not being straight with you.

The vendor also starts carrying measurement risk on top of execution risk, and these are different things. Execution risk is performing badly and earning less, which is fair and easy to explain to a board. Measurement risk is performing well and discovering the intervention was not incremental — the audience would have converted anyway, the brand’s existing programme was already strong, the customer specified a stricter control than expected. A supplier has to be able to survive quarters where the honest answer is that the work was competent and the lift was zero.

Which changes which deals a rational supplier should accept. We can send this is not enough. We can probably improve this is not enough. What is needed is a declared leakage pool, an agreed current-best-effort baseline, a credible counterfactual and enough expected spread to fund the work before any of it is proved. Outcome businesses have to learn underwriting, and most software companies have never employed anyone who knows how.

Which produces a competitive dynamic worth predicting on its own: a vendor who refuses a concurrent holdout increasingly looks like a vendor unwilling to test its own claim. Not dishonest, necessarily. Just unwilling. And in a room where one competitor has offered a control group, unwilling is not a survivable position.

Two models, optimised for two different kinds of certainty.

SaaS was designed to maximise revenue certainty for the vendor. Outcome software is designed to maximise value certainty for the customer. You cannot fully optimise both.

That is the real trade-off of 2027, and it deserves its own name: ARR quality against alpha quality. A pure licence business has beautiful revenue characteristics — recurring, predictable, recognised on access, scaling at the pace of signatures. An outcome business cannot always scale at that pace, because proof has a clock: data integration, a baseline period, a holdout, an intervention window, outcome maturation. And a disciplined one will sometimes refuse revenue outright, because the leakage pool is too thin or the measurement design is too weak to support a claim.

So the honest formulation is not that outcome businesses grow slowly. An excellent one could grow very fast. What it accepts is less predictable growth and greater working-capital intensity, in exchange for stronger proof of value. That is a real trade, made deliberately, and every supplier heading in that direction should be able to say out loud that they have made it.

The likely resolution is hybrid by design: a predictable base priced for the substrate, and a variable layer wherever the causal chain can be measured credibly. Which means the interesting question about any 2027 vendor is not whether it offers outcome pricing. It is which parts of its business it is willing to expose, and why those parts and not others.

And then the part that neither I nor anyone else selling this future writes down often enough.

What happens to the people who currently operate the campaigns.

Campaign operations is the largest pool of human work in marketing. Somebody builds the segment, briefs the content, configures the journey, checks the render, schedules the send, pulls the report. Multiply that by every brand and every agency retainer and it is an enormous amount of employment.

When the machine makes the instances, that is the first line a CFO questions. Not the strategy retainer. Not the brand work. The execution hours — because those are the ones with a visible unit cost and a visible substitute. To claim otherwise, on the grounds that marketers’ jobs move up rather than out, is true at the level of the individual senior marketer and evasive at the level of the industry.

The danger for an agency is not that creativity disappears. It is that deliverables become cheap. Ten banners stop being a defensible unit of value when a system can produce a hundred. A monthly campaign calendar is worth less when the client can specify a goal and have the plan generated. Reporting retainers shrink when the analysis arrives continuously and nobody has to build the deck.

None of this is free, and it is worth being accurate about that, because the loose version of the argument is easy to attack. Models cost money. Inference costs money. Data, orchestration, quality control and governance all cost money. What collapses is not cost — it is the marginal cost of human execution, and that is the specific thing a retainer priced per campaign built is selling.

So my expectation is that the agencies which survive are the ones that stop selling deliverables and start owning an outcome — the same transition being asked of software vendors, arriving at the same moment, at organisations with thinner balance sheets and far less appetite for measurement risk. The shops that matter more will be the ones willing to say: give us the number, the constraints and the right to be measured. That is a better business if it works and a less forgiving one if it does not.

It is not comfortable to write that from inside a company selling the thing causing it. It would be less honest not to.

5   What does not change.

The failure mode of every essay like this one is that it becomes an advertisement for the future. So here is the floor — six things that will look exactly the same in 2027, and will still decide whether any of the rest works.

None of these is affected by model capability. All of them constrain it.

Economics. A thin-margin transaction cannot fund an expensive intervention, however confident the prediction behind it. The break-even on a recovery is still one divided by your gross margin, and no amount of intelligence moves that number. A system that ignores it will lose money at extraordinary speed and with excellent attribution.

Consent. Knowing more about a customer confers no additional permission to act. The two things are unrelated, and the temptation to conflate them grows exactly as fast as the context layer does. This is the single most likely place for the industry to embarrass itself in 2027.

Deliverability. A message the mailbox rejects cannot be made intelligent enough to matter. It is the least glamorous item on this list and the largest multiplier on every number above it. An agentic marketing programme sitting on poor sender reputation is a very sophisticated way of not arriving.

Brand. Ten million perfectly personalised bad offers are still ten million bad offers. Personalisation improves the fit between an offer and a person; it does nothing whatever about whether the offer was worth making. As content becomes free, distinctiveness becomes the scarce thing — and distinctiveness is a human judgement, made by people like Maya in the meeting that got longer.

Causality. A prediction is not an incremental outcome. This is the error the whole apparatus is built to make, and it will make it constantly, in good faith, at scale. A model that says a customer will return is describing the world. A control group is the only thing that can tell you whether you changed it.

Human nature. People still want relevance, usefulness, status, novelty, trust and convenience — and sometimes simply to be left alone. None of that has moved in fifty years of marketing technology, and none of it moves in 2027.

Which leads to the sentence this whole essay exists to support.

The holdout does not become obsolete as the machines get smarter. It becomes more necessary.

The logic is unavoidable. The more decisions a system makes, the more results it produces, and the easier it becomes to mistake a prediction for a cause. Ten campaigns a quarter could be assessed by argument. Ten million decisions a day cannot be assessed by anything except a control group, because there is no other instrument that can separate what the system did from what would have happened without it. Scale does not weaken the case for randomised measurement. It is the case.

And the three commitments I have been arguing for since this began do not change either, because they are not predictions. They are the floor the predictions stand on. Never lose customers. Never pay twice. Never pay fixed.

    What fades, what matters more.

Fades	Matters more
Campaign calendars	Business goals and guardrails
Static segments	Customer context and decision traces
Journey spaghetti	Continuous decisioning
More screens for humans to operate	Capabilities for agents to invoke
Manual campaign operations	Human judgement, taste and restraint
Content scarcity	Brand distinctiveness
Channel silos	Lowest-total-tax routing
Last-click attribution	Incrementality against a control
Vanity engagement	Movement between customer states
Pure fixed SaaS	Outcome accountability, and the risk that comes with it
Marketing to humans	Marketing to humans and to their agents
Renting every interaction	Owned identity, attention and memory

Read that table in one direction and it looks like a technology story. It is not. Almost every item in the right-hand column is a governance, judgement or accountability item — things that cannot be bought, installed or prompted into existence.

Which is the surprising shape of the year ahead. Martech 2027 will contain more technology than it ever has, and marketers will spend less time using technology than they ever have. The systems get more complex underneath and simpler above. The marketer specifies the outcome. The agents run the instances. The customer, increasingly, delegates too. And the scarce things become the ones machines cannot cheaply manufacture: judgement, trust, permission, context, brand and accountability.

MarTech 2027 will be defined not by how much AI marketing uses, but by what humans no longer need to operate, what customers no longer need to endure, and what vendors are finally willing to be accountable for.

Three people, three clocks, and one card that turns out to connect them

The previous essay in this series described a design: a card that arrives sealed, asks before it shows, keeps teaching for weeks and then goes quiet; a set whose membership the world chose rather than an editor; an album that separates what you own from what you know; and a season built so that no one person can finish it alone.

All of that can be correct on paper and dead in somebody’s hands. Mechanics are easy to defend in the abstract and easy to misjudge in the specific, because the thing they have to survive is not an argument. It is a Friday.

So this essay follows three people on three different clocks — one day, one season, nine months. Two of them are in the same Circle. The third has never heard of either. By the end, one card will have passed through all three of their hands.

Nothing described here has been built. These are illustrations of a design rather than observations of a product — and setting them down this way exposed three real errors in it, which is the best argument I know for writing a product out as a story before building it.

Figure 1 — three people, three timescales, and the question each one is there to test

1

Meera, one Friday

Meera is twenty-four, designs products for a company in Baner, and has a commute of about forty minutes on a bus that is never quite full enough to sit down on. She joined on the first day of the season because a cousin sent a link into a family group and said something dismissive about it, which is how most good things reach most people.

Her first morning was not a success. Three cards arrived face down. Each one asked her something before it opened — which of these two is older, which came first — and she guessed at all three, got one right, and watched them turn over anyway. Seventeen slots were blank. Her reaction was neither delight nor rejection. It was the single most dangerous response in consumer products.

Fine.

Day one is not adoption. It is inventory. Whatever happens next has to come from the fact that yesterday changes today, and on day one there is no yesterday.

It is now the Friday of week four — the last Hunt of the season. She holds thirteen cards of India 20 and would still not describe herself as somebody who plays games on a phone.

Figure 2 — the twenty-fifth of September, three visits, eleven minutes

07:40 — something new

The Hunt is at the top of her album where it has been on every Monday, Wednesday and Friday of the season. Three questions.

The first is about cards she already holds: one of IH-03 or IH-11 fought a war he later regretted, and which was it? She knows this, because IH-11 spent a week telling her about Kalinga. She taps it.

The second asks her to connect two cards. That emperor’s grandfather founded the dynasty — which of her cards names the man who advised him? She has to think. She has IH-03, and IH-03 is Chanakya, and yes.

The third is a clue rather than a question, and it uses both answers. Their capital sat on the Son. Name the city. She types Pataliputra, and a card arrives face down with one line on it: is this one older than IH-11, or younger? She says older, which is right, and IH-14 turns over — not at the beginning of its journey but two rungs up, because she reached it by knowing rather than guessing at either step. Fourteen of twenty.

Four minutes, and the last thirty seconds of it were the part she will remember, because for those thirty seconds she did not know what she had.

Two things are worth noticing about that Hunt. Nobody wrote it for Meera this morning; it was assembled from facts already verified for the cards already in her album, which is why the person beside her on the bus would have received a different one. And at the bottom of the album, in small type, is a line that does more work than anything else on the screen: one of your older cards wakes after lunch.

13:15 — something she owns has moved

She opens her album at lunch because she was told this morning that there would be something there. Not a new message — the same thread she has been opening since week one.

IH-06 has been quiet for eleven days. The album has decided it is close to slipping away, and before it goes for good it wants to know which river the capital sat on. She gets it. Two marks of three. One more retrieval, at a longer gap, and it is Mastered.

This is the property that has no equivalent on paper, and it is the strangest thing about the whole design. Nothing was sent. No notification arrived. The state changed on a server and the thread she already had is what showed it to her.

The email in her inbox from four weeks ago knows what happened this morning.

21:30 — somebody she knows has what she needs

The evening return is not the product’s doing at all. The family WhatsApp group has become briefly incomprehensible to anybody not playing, which is a good sign, and Fatima has been asking about IH-04 for twenty minutes. Meera opens the album to see where things stand.

Eighteen of twenty held between the six of them. She is on fifteen, which is the ceiling — three starter cards, then three hunts a week for four weeks, and no amount of diligence takes anybody past it.

She is missing IH-17. Nagesh has a spare.

Meera   Nagesh kaka, spare IH17?

Nagesh   What is IH17

Arjun   😂 Ashoka

Nagesh   Sent

Four lines, and everything the product needs is visible in them: the codes, the channel, the family Circle, a fifty-one-year-old who does not know what he is holding, and a gift.

Eleven minutes across the whole day, in three visits.

One message was sent to her — the morning Hunt. The other two returns happened inside a thread she already had, and each had a different trigger: an appointment in the morning, an announced arrival at lunch, and other people in the evening. A product that answers every design question by sending another email will not survive contact with an inbox.

2

Nagesh, one season

Nagesh is fifty-one, Meera’s father’s younger brother, and runs the hardware business his own father started in Kolhapur. He has never installed a game on a phone. This is not a technology problem — he uses WhatsApp constantly, pays for everything by scanning codes, and books his own train tickets. He simply has no interest in acquiring an application in order to do something he did not previously want to do.

When Meera sent the Circle invitation he tapped Join, mostly so that she would stop asking. Then he ignored it.

Week one: he does not play

Three cards arrive and he glances at them. The first Hunt arrives and he does not answer it. The second he opens, reads, does not know, and closes. This continues for six days.

It is worth sitting with that stretch, because a story where everybody is delighted from the first message is not a story about a real product. For six days Nagesh is a name in a Circle and nothing else, and if the design depended on his enthusiasm it would already have failed.

Week two: he answers one, and something is minted

On the Wednesday of the second week the Hunt asks about a Mauryan emperor, and he happens to know it, because he went to Sanchi as a boy and his father told him the story on the way back. He answers, and a sealed card asks him to guess whether it ruled before or after the one he already holds. He guesses right, which surprises him. IH-17 turns over two rungs up.

Over the following fortnight the album asks him about it twice more, at widening intervals, and both times he gets it right. On the second, it turns into three marks — and a spare.

That last detail is where the first of the three errors turned up, and it is worth stopping for.

Figure 3 — where a spare copy comes from, and the third route that scales

The season as first designed had Hunts on Monday, Wednesday and Friday, each minting one new card, and mastery days on Tuesday and Thursday. That produces fifteen unique cards per player and not a single duplicate anywhere in the system.

Which makes the whole social economy impossible. If nobody ever holds two of anything, nobody can give anything away. Meera can ask Nagesh for IH-17 all she likes; he has one, and he needs it.

The fix, and it improves the design rather than patching it.

Mastery and connection days mint spares — a second copy of a card the player has already mastered, which can be given away. A spare is therefore earned by knowing something, never by luck, and the reward for learning is something to give to somebody else. Nagesh had two Ashokas because he mastered one.

There is a third route to a spare, and it is the one that will do most of the work once a Circle is larger than a family. Two people can put a card each on the table and call a category — the length of a reign, the reach of an empire, the depth of the card itself — and settle it on the numbers printed on every card in the set. Only a mastered card can be entered, so the wager is knowledge rather than luck. The winner mints a spare. The loser keeps everything they arrived with.

Nothing is confiscated and nobody is raided, which is what stops the trade desk becoming a market. What a duel does is manufacture supply out of the one input this system can produce at will, which is somebody having learned something.

That reversal matters more than the arithmetic it fixes. In most collectible systems, generosity costs you something. Here it is a by-product of having paid attention.

Week four: he is asked for something

This is the moment the design either works or does not.

Figure 4 — the message that arrives, and where it arrives

Look at where it sits. Between the electricity bill and the bank statement, in the same list, in the same place he goes every morning anyway to see what he owes and to whom. He did not travel to receive it. He did not have to remember that he was playing anything.

And look at what it says. Not your streak is at risk. Not come back and play. It says that Meera needs IH-17, and that he has a spare.

A streak is between a person and some software. A missing card is between a person and his niece.

He taps once — under ten seconds, no password, no app, no account he has to remember creating. Then he does something nobody asked for: he writes into the family group that he has sent it, and mentions that he got both of his right, which is not entirely true and which everybody lets stand.

That evening he opens the album for the first time and looks at what the Circle has. It is the first time he has gone to the product rather than the product coming to him. Nobody persuaded him. He was needed, and being needed is a stronger force than being entertained.

The weekend: what six people can do and one cannot

Figure 5 — the same six people, three times in one season

The middle panel is the design. Six people, fifteen cards each, ninety cards between them, and not one person able to finish. Every gap in every row is somebody else’s spare.

That configuration is arithmetic decided before the season opened: three starter cards, three Hunts a week for four weeks, a ceiling of fifteen, and twelve Hunt cards drawn differently for every player so that no two people are short of the same five.

Compare it with the panel that would exist if the design were wrong — twenty filled squares in every row by the Friday of week four. No gaps. Nothing to ask for and nobody to ask. Six people playing alone in the same room, and a social layer that exists in the marketing and nowhere else.

Over the Saturday, the asking happens. Arjun has a spare Fatima needs. Priya has been sitting on something Sunil has wanted for a fortnight. Nagesh, who a fortnight ago did not know what IH-17 was, turns out to have spares of three cards — two from mastering them and one from a challenge he did not expect to win.

Now look at the right-hand panel, and read it down the columns rather than across the rows.

The rows say 16, 17, 18, 19. Not one of the six has twenty. That is not a failure of the weekend; it is what the season is for. Ninety cards plus a handful of spares does not become a hundred and twenty just because people are generous, and a design that claimed otherwise would be lying about its own arithmetic.

The columns say something different. Every column has at least one filled square. Between the six of them, every card in India 20 is held by somebody.

Nobody completed the set. The Circle did.

Figure 6 — Sunday evening

What arrives is not points. It is a card carrying the Circle’s name, the six of them, and the date — untransferable, unrepeatable, and of no value whatsoever to anybody outside those six people. Nagesh, who six weeks ago would not have installed anything, has his name printed on it.

And then two words from him that are worth more than any retention chart: next one when?

Why the gift and not the swap.

Nagesh gave Meera a card and received nothing back. A settled exchange would have closed the matter on Thursday morning. Instead, eleven days later, in a different set, Meera found she was holding a spare of something he needed and sent it without being asked. There was nothing to settle. He had helped her; later she happened to be able to help him. That is how things move between people who know each other, and it is why the group chat stays alive between seasons.

3

Thara, nine months

Thara is nineteen and studying in Madurai. Neither Meera nor Nagesh has heard of her. She joined in the second month, on her own, because a friend forwarded something during a lecture.

Her story runs on a different clock because the questions that matter at nine months are not the ones that matter on a Friday. Anything can be interesting for a fortnight. The design claims something harder: that an album becomes more interesting as it fills, that knowledge compounds rather than evaporating, and that a collection acquires history. Nine months is the shortest honest test of any of that.

Figure 7 — three numbers measuring different things, on three different clocks

The two lines, and the third number under them

The red line is cards held. It climbs steadily and jumps whenever a new set opens — the fourteen peaks above eight thousand metres in month three, the Chola temples in month five, the Eighth Schedule languages in month seven. By month nine she holds a hundred and sixty cards across five sets.

The green line is cards she can still answer for. It climbs more slowly, and — this is the part the design originally had wrong — it does not only climb. In her exam weeks it falls. She stops answering the recall prompts, three or four cards slip below the threshold, and the number goes down.

That decline is not a flaw to be designed away. A mastery number that can only rise is tracking badges rather than memory, and everybody can tell the difference. What the album does is notice, and pull the slipping cards back into rotation until they are secure again — which is why the line recovers rather than staying down.

The gap between the two lines is sixty-six cards that she holds and cannot yet account for. That gap is the product being honest. Somebody who has been given a great many cards and retained little of what is in them should have an album that says so, or the second number means nothing and the first one is just a score.

A third number runs underneath both and it is the one that keeps a large album bearable. Finish counts sets closed, and it moves weekly rather than daily or slowly. Hold answers what arrived. Know answers what stuck. Finish answers the only question that produces a feeling of arriving somewhere, which is whether anything is complete. Thara has four. It is the smallest of her three numbers and the one she would quote if somebody asked.

Anybody can be given a card. Nobody can be given the answer to a question about it six weeks later.

The machinery behind the green line is the least glamorous thing in the entire design. Every day the album picks perhaps three cards out of the hundred and sixty by asking which piece of knowledge is closest to slipping away. Spaced repetition, understood for decades, doing something rather elegant here: a hundred and sixty cards produce three prompts rather than a hundred and sixty obligations.

The same card, three different relationships

By month nine, Thara, Meera and Nagesh all hold IH-17. They hold the identical knowledge object — the same code, the same seven reveals, the same verified facts, rendered for Thara in Tamil and for the other two in English.

Nagesh has three marks and has had them since week four. Meera has two, and lost one during a fortnight she spent moving flat before earning it back. Thara has none at all, because she received the card eleven days ago.

Same object. Three entirely different relationships with it. That distinction is the reason mastery is worth displaying at all.

What her album says about her

Figure 8 — four cards from one set, four different origins, and nothing that expired

Look at what separates these four, because it is not the knowledge. Every fact inside every one of them is identical to the version in anybody else’s album, anywhere in India, in any language.

TM-08 Brihadeeswara carries a Thanjavur mint mark, because that is where she was standing when she earned it. A collector in Delhi can never mint that specimen — but can perfectly well be given one by somebody who did, which is the entire point. TM-11 came from a person in Pune whose name is on it. TM-14 she found herself on day forty-one. And MC-04 is not a temple at all: it appeared because she held five related cards and put them into the right causal order, which means it cannot be found in any Hunt and cannot be given to her by anybody.

Geography makes the object scarce. It never makes the knowledge scarce.

Everything about Brihadeeswara is reachable by anybody, from anywhere. The Thanjavur edition is not — which gives a collector in Delhi a reason to know somebody in Thanjavur. That is the only mechanic in the whole design that pushes the social graph outward rather than deeper into groups that already exist.

Month nine: a set that other people finished without her

In the ninth month Thara starts India 20 — the set Meera and Nagesh finished before she had heard of any of this. Nothing about it has been withdrawn. Every one of the twenty cards is still earnable, at the same difficulty, by anybody who turns up; the season is a pacing device for the people playing it together, not a shutter that comes down afterwards.

This is a correction rather than a description, and it is the third thing writing the story found. The design originally had a first-month printing that closed for ever, and the sentence defending it read well: what she cannot have is the edition, which is exactly the right thing to lose. It is not. A rarity that expires is a machine for manufacturing regret, and regret is the one feeling this product cannot afford in a country where almost everybody who ever plays it will hear about it in its second year. Thara is the reason the mechanic is gone.

Scarcity should survive. Missing out should not.

So she has two routes and neither is a consolation prize. She can work the set at her own pace, one card at a time, exactly as Meera did. Or she can find somebody who was there — which is faster, and which is the only thing the earlier players have that she does not. She posts five characters into a wider group — no explanation, none needed — and somebody two or three degrees away from her answers.

The card that arrives is an IH-17 that has been somewhere. Under the artwork is a short line of custody: minted in Kolhapur, passed to Pune, passed to Madurai. She could have earned an IH-17 herself in a fortnight and it would have taught her the identical thing. What she could not have earned is this one.

Figure 9 — one card, nine months, three cities

Meera has no idea who Thara is. She had a spare — minted by mastering a card that Nagesh had given her nine months earlier — somebody needed it, and it cost her nothing.

Sets do not close. Specimens do not repeat.

An old set does not become a graveyard and it does not become a locked door either. The knowledge stays open at the same price it always cost. What cannot be reproduced is a particular copy with a particular history — where it was minted, at what level of recall it was earned, and whose hands it passed through. A latecomer is never made to miss something. She is given a reason to reach past the people she already knows.

4

What the stories are really testing

Stories are persuasive in a way that ought to make a reader suspicious, and the person writing them more so. Nothing above has happened. It is a design rendered as narrative — a useful way to find out whether mechanics fit together, and a terrible way to find out whether anybody wants them.

What it did produce is three corrections that no amount of arguing in the abstract had surfaced. The duplicate economy did not exist, so the entire social layer was resting on nothing. The season ended with six people reaching twenty out of twenty when the copies in circulation made that impossible. And the rarity that closed after a month read as a nice piece of collector authenticity until a nineteen-year-old in Madurai walked into it, at which point it read as a punishment for having heard about something late. All three are fixed. All three would have been found far more expensively in a pilot.

Five things remain, and they are the questions the stories assume rather than answer.

— Do the second and third daily returns happen? The morning Hunt is well-understood shape. The lunchtime return depends on an old thread reliably coming back to life, and on people learning to expect it. The evening return happened because other people were talking. If only the morning survives, this is a quiz with a card attached.

— Does being needed activate a reluctant member? Nagesh does nothing for six days, and what eventually moves him is a specific person needing a specific thing only he has. That mechanism exists only because the arithmetic leaves a hole — and it works only because the message is about Meera rather than about him. Anything mentioning his streak, or his six days of silence, and he would have muted it.

— Does accumulation stay motivating after months? Spaced repetition works; that is not in doubt. What is in doubt is whether people tolerate it inside something they came to for fun, and whether a visible gap of sixty-six cards reads as an honest account or as an accusation. Which of the two numbers is shown larger is not a cosmetic decision.

— Do people care about the sets? Everything here rests on somebody caring about twenty cards on the Mauryan empire or the Chola temples. There is now a filter that rejects the obviously flat candidates before anything is drawn — whether the reader can name three and then stall, whether card three and card nineteen teach the same shape, whether the items relate to one another at all. A filter is a way of failing less often. It is not evidence that anybody wants the ones that pass, and the first set will find an audience because the idea is novel rather than because the subject is.

— Does anybody type five characters unprompted? This is the one I would watch first. Meera writing Nagesh kaka, spare IH17? into a family group is the smallest possible unit of demand, and no growth mechanism in the design produces anything if that behaviour does not appear on its own.

Nine months earlier, a hardware dealer in Kolhapur who had ignored the whole business for six days answered one question about a Mauryan emperor because his father had told him the story on a bus. Getting it right eventually gave him a spare.

He gave the spare away because his niece asked. She mastered the card herself, which gave her a spare in turn, and when a stranger in Madurai posted five characters into a group she had never paid attention to, she sent it without thinking about it.

The knowledge inside that card never changed. Ashoka fought at Kalinga in 261 BCE in all three albums, in two languages, on three phones. What changed was everything around it: three names on the back, three cities, and three people who each had to start their own understanding of it from nothing.

The reluctant sixth member is not the person you tolerate in order to reach five. He is the one who has the card.

That is why all of this has to live where he already is. Not because email renders well — it does not, particularly — but because the person who solves your problem may be somebody who would never have installed anything, never joined anything, and never wanted to play.

The uncle had two. Nine months later, one of them was in Madurai.

A card, a Circle, an album that remembers, and a graph that grew outward one gift at a time. At which point this stops being a collection and starts being somewhere people go.