Rajesh Jain

Relationship Media and the Shift to Attention Routing

What happens when a brand knows exactly which customer action should happen next, but no longer has that customer’s attention?

Every brand knows more about its own customers than any advertising platform does. It knows what they bought, what they nearly bought, what is about to expire and what should logically happen next. And every brand has a growing share of customers who no longer hear it. The emails go unopened, the app sits unused, the WhatsApp messages stay unread. At that point the brand does something strange: it pays a platform to find a person whose email address already sits in its own database.

This essay proposes a step between those two moments. Before a brand rents its customer back, it should be able to reach that customer through attention someone else has earned — with the customer’s permission, without either brand seeing the other’s data, and carrying a specific, useful action rather than an advertisement. I call this a CRM Ad. The network that makes it possible is NeoNet. The category it opens is Relationship Media.

Your customer. Someone else’s attention. Your next-best-action.

CRM Ads look like a new advertising format. They are the visible consequence of a deeper change: marketing is moving from channel orchestration to attention routing. Marketing software has always sent campaigns to channels. It should route actions to attention.

1

The Customer Who Stopped Listening

Ananya has not opened an email from her beauty retailer in three months. She opens a news digest every morning.

She joined the retailer’s loyalty programme two years ago and bought regularly for a while. Then its emails became indistinguishable from the forty others she receives each day, and she stopped opening them. She has not unsubscribed. She has not left. She simply no longer hears the brand.

The retailer’s CRM knows something useful about her: she has $30 in loyalty points that expire on Sunday. It has tried email twice and a push notification once. Nothing. On Monday the points lapse, and with them the most natural reason she had to come back.

Meanwhile, every morning at 7.40, Ananya opens a news publisher’s NeoLetter, its curated daily digest. She has done so for months. It is part of her routine. On Thursday, that digest carries a small block.

Figure 1. A CRM Ad inside a publisher’s digest: at open, and after the reader taps ‘Show me’.

She taps Show me. The block expands inside the email to display three products her points will cover. She picks one and lands in the retailer’s own app, where she completes the redemption under her own login. The digest never handled her account. The publisher never learnt she was the retailer’s customer. The retailer never learnt she reads that digest.

Her reaction is the test this entire idea must pass: “Of course I wanted to see that.” Not: “How did they know?”

Consider what the retailer would otherwise have done. Next quarter its paid-media team would have uploaded her email address, with millions of others, to an advertising platform and bid in an auction to show her a generic banner. It would have paid rent to reach a customer it already had.

2

The Empty Quadrant

Marketing has two mature ways to reach people. A third has been missing.

Draw a grid. One axis asks whether the person is a known customer or an unknown prospect. The other asks whether the message appears on the brand’s own surface or on someone else’s.

Figure 2. The empty quadrant: a known customer, reached on another party’s surface.

A known customer on the brand’s own surface is CRM: email, app, SMS, WhatsApp — owned media. An unknown prospect on someone else’s surface is adtech: search, social, programmatic — rented media. An unknown prospect on the brand’s own surface barely exists as a category. The fourth quadrant — a known customer reached on someone else’s surface — has been occupied only by the platforms, and only on their terms.

Today that quadrant has a name, and the name is revealing: retargeting. Instead of treating the person as a known relationship that has temporarily lost attention, the brand sends the identity into an advertising system and buys the customer back through an auction.

Relationship Media is a brand reaching its own known customer through attention earned by another party. It sits between the other two: owned first, relationship next, rented last. In the language a CMO’s budget already uses: CRM, then CRM Ads, then adtech.

The defining fact of Relationship Media is not that the host is friendly. It is that the advertising brand already has a relationship with the customer. Three things have to come together, and each supplies something different. The host supplies the attention. The existing relationship supplies the relevance. Permission supplies the right to use it. Remove any one and the CRM Ad should not run.

3

The Inventory Nobody Counts

Every brand holds millions of unfinished customer actions. Nobody measures them as inventory.

Look inside any large consumer brand’s CRM and you find the same thing: actions that should happen and have not. Points expiring. A replenishment due. A travel booking started and abandoned. A quote waiting for a reply. A subscription about to lapse. A favourite product back in stock. A service visit unbooked.

Each has a value the brand can estimate and a customer the brand can name. And each follows the same path. CRM tries — an email, a push, perhaps a WhatsApp message. If the customer does not respond, CRM stops. Weeks later, if the customer drifts far enough, the paid-media team starts again from the other end and buys the customer back through an auction.

The two halves do not even share a vocabulary. CRM speaks of customers, states and journeys. Media speaks of audiences, impressions, bids and click-through rates. Somewhere in the translation, the unfinished customer action — the economically important object — disappears from view.

Between CRM stopping and adtech starting there is a gap. In that gap sit the unfinished actions of customers who are not lost, only unreachable on the brand’s own channels. In the language of Before the Slide, this is where the slide begins.

CRM Ads do not sell targeting. They give unfinished customer actions a second surface before the brand rents the customer back.

The framing matters. “Advertisers can target individual customers in other brands’ emails” sounds invasive the moment it is said. “Your points expire Sunday — here is what they cover” sounds like service. The mechanism is the same. The proposition is not.

4  From Channel Orchestration to Attention Routing

For twenty years, CRM decided what to say and adtech decided where to buy attention. The organisation chart fixed the hand-off between them.

The CRM team’s channels are email, push, SMS and WhatsApp. The paid-media team’s channels are search, social and publishers. Each optimises within its own boundary. Nobody starts from the customer and asks: this known customer has an unfinished action worth $12 — what is the cheapest available path to their attention?

Traditional martech asks: which owned channel should I use? Traditional adtech asks: which paid audience should I buy? NeoMarketing asks a different question: for this customer, in this state, with this action value, what is the best attention path available now?

Once that question is asked, the channel boundary stops making sense, and the optimal routing is plain: owned attention first, relationship attention second, rented attention last.

Figure 3. Attention routing: the decision moves down the ladder only when the tier above fails, and every outcome is written back.

Until now, NeoNet has been described mainly as a recovery network: when a customer has been dormant for months, reach them through a brand they still engage with, before paying adtech. CRM Ads widen that role from recovery to response. The trigger is no longer months of silence. It is an owned message that went unanswered hours or days ago.

That turns NeoNet from a dormant-customer network into a distributed execution network for CRM decisions. The integration is a single new node in the journey builder: Escalate to NeoNet. An M-Agent reaches it with a clear piece of reasoning: email, push and WhatsApp have failed; this action is worth $12; find her somewhere else.

The full loop runs from customer state to next-best-action, through owned attempts, escalation, a relationship surface and an ActionAd, to an outcome — and then back. The last step matters most. The outcome returns to the brand’s CRM and changes what happens next, which no rented impression does.

Adtech distributes creatives. NeoNet distributes decisions.

This also explains why the idea is arriving now. A human marketer cannot decide, customer by customer across ten million records, whether today’s best route is email, WhatsApp, NeoNet or a platform. An agent can. Attention routing is a decision made at machine scale. It became practical when decisioning did.

5  Customer Match Without the Auction

The matching is not new. The economic arrangement is.

It would be wrong to claim that the platforms cannot reach a brand’s known customers. Customer Match lets a brand upload first-party data and reach matched customers across Google’s surfaces, Gmail included; Meta’s Custom Audiences do the same. The platforms already sell ‘your customer, our attention’. They sell it as rent. Brands spend heavily on this. The demand is proven.

What differs is the optimisation problem underneath. A platform asks: how much can this impression earn in the auction? NeoNet asks: what is the cheapest trustworthy route to completing this known customer’s unfinished action? Those are different objectives, and they produce different products.

NeoNet can match deterministically wherever the same permissioned identity exists on both sides, rather than relying on a partial or modelled overlap. The decision is made at the moment of open, not when a campaign was set up days earlier. The customer acts inside the email, using AMP for Email, rather than clicking out to a landing page. The brand pays for incremental outcomes measured against a holdout, not for impressions.

There is a quieter advantage. The advertiser does not need to hand its customer state and next-best-action to a general-purpose advertising platform. On a platform, the brand’s knowledge of what its customer should do next becomes an input to someone else’s optimisation engine. On NeoNet, the decision travels to the customer and the outcome travels back to the brand. The decision stays the brand’s.

CRM Ads are Customer Match without the auction — and without the rent.

6

The Anti-Cookie

Cookies guessed relationships. CRM Ads use declared ones.

The comparison with third-party cookies is tempting and dangerous. Cookies let a brand follow a browser around the web and show it ads after it left. In that narrow sense, a CRM Ad is a descendant. But cookies became objectionable because they answered a question nobody had agreed to: where else can this browser be followed? CRM Ads answer a different question: where else has this customer allowed brands they already know to reach them?

Cookies did something else that is easy to forget. By making cross-site reach easy and invisible, they removed any need for brands to build cooperative, permissioned ways to reach customers whose attention had moved elsewhere. As invisible reach becomes harder to defend, the need it masked becomes visible.

	Third-party cookie	CRM Ad
Identity	An anonymous browser	A customer’s declared relationship with the brand
Basis	Inferred interest	Known customer state
Permission	Invisible and assumed	Explicit, visible and revocable
Timing	Decided before delivery	Decided at the moment of open
Carries	A creative	A next-best-action
Once done	Keeps chasing	Cancels itself

 

The timing row is where CRM Ads do something no cookie could. Because the AMP block is composed when the email is opened, it can check the customer’s state at that instant. If the points are still unredeemed, it shows Show me. If Ananya redeemed them yesterday in the app, it shows a thank-you — or nothing at all.

Figure 4. Decided at open: the same CRM Ad renders differently depending on the customer’s state.

Ordinary advertising acts on a decision made before delivery and keeps chasing a customer who has already converted. A CRM Ad cancels itself when it is no longer needed. That makes it less an advertisement than a remote CRM action slot: a small, permissioned window in which a brand’s live decision about its own customer can appear, update and disappear.

7

Permission, Not Plumbing

Hashing is security architecture. It is not permission architecture.

The obvious objection to CRM Ads is privacy, and the obvious answer — the exchange is blind, so neither brand sees the other’s data — is necessary but not sufficient. A hashed email address is still personal data: pseudonymised, not anonymous. Blind matching protects the data. It does not, by itself, create the right to use it.

India’s Digital Personal Data Protection Rules were notified in November 2025, with the substantive obligations on notice, consent and children’s data phasing in over eighteen months, to May 2027. The law offers no general legitimate-interest route for marketing of the kind European law allows. Consent carries the weight: it must be specific to a purpose, and withdrawing it must be as easy as giving it. Targeted advertising directed at children is prohibited outright.

Three flows need cover. The advertising brand’s notice must tell its customers it may reach them through partner surfaces. The host’s readers must have agreed to receive messages from brands they already use. And the operator’s role must be settled: an email platform that matches one client’s list against another’s is no longer simply processing on each client’s behalf. That changes who decides the purpose of the processing, and the design must face it squarely.

One trap deserves naming. Existing customer lists were not collected with this use in mind. CRM Ads cannot be built by quietly reinterpreting old consents. They have to be built on new, visible permission. The cleanest design places that permission with the reader:

Your Brands

Let brands you already have a relationship with reach you here when something useful is waiting. Your email address is never shared between brands.

The sequence follows from this: design the reader’s permission, define the data roles, take DPDP counsel, and only then decide which entity operates the network. A new legal entity does not create permission. A well-designed consent surface does.

No cross-brand reach without permission the customer can see, understand, revoke and control.

Privacy pressure does not make cooperation between brands necessary. It changes what acceptable cooperation looks like. Cooperation can no longer mean pooling customer files. It has to mean permissioned matching without exposing them.

8

Push, Pull and the Lines We Will Not Cross

The more specific the action, the more the customer should pull it rather than have it pushed.

CRM Ads come in four levels of specificity, and each level carries its own rule.

Figure 5. Four levels of CRM Ads, from push to pull.

A cohort message — points expiring this month — can reasonably appear in a host’s digest for any reader who has granted the Your Brands permission. A state message, such as a replenishment due, can be pushed where that permission exists. An individual next-best-action belongs only on surfaces the reader has opted into. The exact items left in a basket should appear only where the customer has asked to see their pending actions. That rule replaces a vague instinct about creepiness with a design principle.

Some lines hold at every level.

No payment, credential or identity-verification actions on third-party surfaces. Brands and banks have spent years teaching customers never to act on a payment request or a login prompt that arrives from somewhere else. A CRM Ad that asked for payment details, a password, a one-time password, an identity document or KYC information inside another brand’s email would undo that training and hand fraudsters a template. Relationship Media must not train the opposite behaviour. At most, an account-sensitive action becomes a pointer back to the brand’s own app or site: open your app, something needs your attention.

No banks as hosts. Banking communications carry their own regulatory weight, and third-party messages inside them are out of bounds. No prospecting, no lookalikes and no list resale in the first version. CRM Ads exist for customers a brand already has.

The test for every CRM Ad remains Ananya’s reaction. “Of course I wanted to see that” passes. “How did they know?” fails.

9

Alternative Reach: The Number That Sells the Network

Network size is vanity. The useful question is how much lost attention the network restores.

Email networks traditionally sell themselves by audience size: so many million addresses. That number tells a brand nothing about its own problem. The brand’s question is narrower and more useful: of the customers I cannot reach today, how many can you?

Call that Alternative Reach: the share of a brand’s attention-dark customers who are currently reachable, with permission, through another member of the network. It belongs on the same dashboard as Real Reach, the brand’s 90-day engaged base. One says how much attention the brand retains. The other says how much of the lost attention can be recovered cooperatively before renting it.

Figure 6. Alternative Reach for an illustrative brand.

An illustrative brand has 20 million customers in its database. Its Real Reach is 6 million, leaving 14 million attention-dark. If NeoNet can find 4 million of those on active, permissioned surfaces elsewhere, its Alternative Reach is 29%. The pitch becomes one sentence: your CRM reaches 30% of your base; NeoNet gives you a second path to 29% of the rest before you spend on adtech.

Universal reach is not required. What matters is overlap among commercially valuable customers, and overlap is uneven. A person who ignores one retailer still opens their airline’s updates, a news digest, a food-delivery app’s daily email or another retailer’s NeoMail. The overlap matrix — each brand’s dark customers against each host’s active readers — is one of the network’s core assets, and it improves with every member.

Reach alone is not value. Value is Alternative Reach multiplied by the worth of the actions waiting and the incremental response they draw. Five per cent reach into $500 actions is worth far more than thirty per cent reach into $5 actions. A network with modest reach can still be valuable if the actions it carries are dense with value.

Fill follows the same logic. If a reader is a known customer of five brands on the network, and each has a live action for them on one day in ten, a useful CRM Ad is available on roughly four days in ten: the chance that at least one of the five has something waiting is one minus 0.9 to the power of five, or about 41%. Real behaviour will not follow independence neatly, but the direction holds: the density of participating brands matters more than total audience.

10

The Economics: Price the Outcome, Not the Rectangle

An unfinished action can be worth far more than the impression that carries it.

Because the advertising brand already knows the customer and the value of the action, CRM Ads do not need to be priced by the thousand. They can be priced on incremental outcomes, measured against a holdout — the Alpha logic of NeoMarketing applied to reach. If Relationship Media is sold by the thousand, it will be compared with newsletter advertising and, in time, optimised like newsletter advertising. That throws away the most important thing the advertiser knows: the expected value of the unfinished action.

Take an illustrative pilot. A brand has 350,000 attention-dark customers reachable through NeoNet. It holds out 50,000 at random and sends CRM Ads to the other 300,000. The exposed group completes 8,000 more actions than the holdout rate predicts — an uplift of under three percentage points. At $40 of contribution per completed action, that is $320,000 of incremental contribution the brand would not otherwise have earned. If NeoNet takes 15–25% of that verified Alpha, between $48,000 and $80,000 is shared between the host, the network and, through Mu, the reader whose attention made it possible.

That changes bid logic too. The advertiser can value one customer differently from another by customer value, urgency, state and expected response. A $5 action should not displace a $100 action merely because both wanted the same slot.

This is where the Alpha Ledger earns its place. Every action needs a pre-agreed baseline, a holdout where feasible and a decision trace from state to action to surface to outcome. The aim is not to claim every conversion after exposure. It is to measure what changed because NeoNet existed.

The host does not sell impressions. It leases moments of trusted attention for verified outcomes. Every party gains something it cannot get elsewhere. The advertising brand completes actions more cheaply than buying the customer back. The host earns from attention it has already built, without selling its readers to strangers. The reader receives something useful instead of something random, and, where Mu is in play, a share of the value.

One number governs the host’s side: net attention yield — revenue earned minus the cost in unsubscribes, complaints and falling engagement. A CRM Ad that costs its host more attention than it earns should not run. The principle that protects NeoMails protects the network.

11

Why Nobody Built This

It was not one missing technology. It was six capabilities living in different businesses, none of which had a reason to combine them.

When an idea this obvious has not been built, either something blocked it or it is not worth building. Here it is mostly the first — with a warning attached.

Figure 7. Six capabilities, six holders, six reasons the pieces stayed apart.

CRM decisioning — what should happen next for this customer — lives in marketing clouds and email platforms. Deterministic identity and inbox supply were assembled by email advertising networks. External attention at scale belongs to the platforms. In-email action arrived with AMP. Privacy-safe matching and credible incrementality came from clean rooms and holdout methods. And routing at machine scale arrived only with agents. Each holder had a reason not to assemble the rest.

The platforms could have built much of it, and their incentives run the other way. Google has the inbox, created AMP for Email and runs Customer Match; Meta runs Custom Audiences. But making a brand’s known customers cheaply reachable would cannibalise the auction that funds them. Google also stopped scanning Gmail content for ad personalisation in 2017; it has little appetite to put brand-to-brand messages into the inbox it hosts.

The marketing clouds and email platforms were best placed technically and least permitted culturally. They held the lists of thousands of brands — Netcore among them. But they are built on tenant isolation. Their trust promise is that one client’s customer data never becomes another client’s asset. NeoNet requires controlled cooperation between tenants, which is close to the opposite design philosophy. When the marketing clouds did connect CRM to advertising, they piped customer lists into the platforms — routing towards rent, not away from it. And the emails they sent were not worth hosting anyone else: promotional email with decaying opens is not inventory.

The companies best placed technically were the companies least permitted culturally.

Email advertising networks came closest. LiveIntent, founded in 2009, built an identity graph of more than 235 million hashed email addresses a month across thousands of publisher newsletters, and was acquired by Zeta Global in 2024 for $250 million. That validates the foundation: deterministic email identity and distributed inbox inventory can become a substantial business. But its unit ran from identity to audience to creative to impression. It did not carry a brand’s live decision, could not let the customer act inside the email, and was bought and priced as display advertising. It shows the value realised by one architecture, not the ceiling of Relationship Media. LiveIntent monetises identity. NeoNet executes customer intent. Zeta’s later moves to buy an enterprise email business and a newsletter publisher show that at least one company is now assembling the pieces.

Cooperation itself has a poisoned history. In 2000, DoubleClick’s plan to combine the offline buyer data of the Abacus catalogue cooperative with online browsing cookies became one of the internet’s first great privacy controversies, and DoubleClick backed down. For a generation, brands pooling customer data has been toxic, rightly. Clean rooms now make matching possible without exchanging files, but plumbing is not a customer proposition. The cooperation CRM Ads need is the inverse of pooling: no shared files, only permissioned matching that the customer can see.

Others built adjacent pieces. Shopify pooled signals across merchants for advertising audiences, then handed those audiences to the platforms. Indian bill payment and card apps already show customers a list of dues across billers — a pull version of pending actions, confined to one vertical. Loyalty coalitions showed the appeal of a multi-brand membership, and its flaw: the coalition owned the member, not the brands.

What changed is not that these technologies suddenly appeared. They have become good enough, cheap enough and organisationally plausible enough to connect. Attention-earning brand email turns a cost centre into a surface worth hosting. AMP turns that surface into an interactive action surface. Agents make individual routing decisions at scale. Clean-room matching makes cooperation possible without pooling. Holdouts make outcome pricing credible.

The deepest reason is organisational.

The market was built around channels, not customer states.

CRM owned some channels; paid media owned others. Nobody owned the question of which route a particular unfinished action should take. Each industry optimised inside its own boundary, so each built one piece and stopped. The missing technology was not matching. The missing mental model was routing. Once an agent can value every unfinished action and choose the cheapest attention path, a layer between owned and rented media becomes possible. NeoNet is that layer.

The technology existed. The category boundary was the constraint.

12

What Would Disprove or Kill This

Relationship Media should be tested as hard as it is argued. Five findings would disprove the thesis. Three would kill the opportunity without disproving it.

What would disprove the thesis

Thin Alternative Reach in a dense cluster. The initial hypothesis is that a deliberately dense cluster of brands and hosts can reach at least 15% of a member’s attention-dark customers. That figure is a hypothesis, not evidence. If reach comes in far lower, the test becomes whether value density compensates, since a small reach into high-value actions may still pay. If neither holds, the network is not ready.

No incremental response. If customers exposed to CRM Ads complete actions at no higher rate than the holdout, the mechanism is reaching people without moving them. Outcome pricing exposes this immediately, which is the point of using it.

Negative net attention yield for hosts. If CRM Ads raise unsubscribes and complaints faster than they earn, hosts will withdraw, and they should. The network needs supply that grows healthier as it carries more.

Readers refuse the permission. If the Your Brands permission is declined at high rates, or ‘Why am I seeing this?’ produces more anxiety than reassurance, the consumer proposition is wrong whatever the economics.

Hosts will not carry other brands. Brands may not want other brands inside their messages, however well governed. If too few hosts open their surfaces, there is no network to route through.

What would kill the opportunity

The permission architecture cannot be made to work. If counsel concludes that cross-brand matching cannot be structured with visible, revocable permission and clear data roles, the product does not ship in that form. The need would remain; this route to it would close.

The inbox changes the rules. In-email action depends on AMP, which only a few mailbox providers support and one dominates in practice. If that support narrows, CRM Ads fall back to a link out of the email: still useful, but weaker.

A better-capitalised player assembles the pieces first. That would not disprove Relationship Media; it would confirm it. NeoNet would simply have lost the race. The defence is speed and density within one cluster, not breadth across many.

13

Three Phases

Prove the economics before expanding the permission. Expand the permission before expanding the specificity.

Figure 8. The rollout: economic proof, then permission, then specificity.

Phase 1: Prove Alternative Reach. Start with NeoNet Signals — Alternative Reach measurement, overlap analysis and suppression lists for paid media — alongside Reconnect Ads and Level 1–2 actions. Suppression is the easiest first sale: these customers are reachable through owned or relationship attention, so stop paying the platforms to reach them. It is Never Pay Twice made operational, and because it is not advertising it sits in its own product family. Reconnect Ads aim to win back the owned channel rather than the sale: a WhatsApp opt-in, an email re-permission, an app return. Publisher digests provide early density. The visible pilot uses an action that passes the test: points expiring, a subscription renewal, a booking left unfinished, a replenishment due. Two numbers decide everything: Alternative Reach and incremental Alpha.

Phase 2: Earn the permission. Launch My Next Actions as an opt-in Magnet inside NeoMails: a block where the reader sees useful pending actions from brands they already use. Opting in is what unlocks Level 3, individual next-best-actions.

Phase 3: Turn actions into objects. Transaction-level actions appear only on surfaces the reader has pulled. In NeoCards they become persistent objects the reader can open, save, snooze, dismiss or complete. An advertisement vanishes once scrolled past. A pending action waits until the customer is ready.

The sequence does three things. It proves value before the network needs broad density. It makes consent stronger as specificity rises. And it lets the customer’s experience move from another ad towards a useful place that organises what is waiting for them. Do not begin with the most specific version of the idea. Earn the right to become more specific.

14

Beyond Ads

The more the customer pulls, the less this looks like advertising and the more it looks like an agent working for them.

Push CRM Ads are a brand asking to reach me elsewhere. My Next Actions is me asking the network to organise what is waiting for me. The second may become the larger product.

Picture a block in a daily email: My Next Actions — 3. A retailer: points expire Sunday. An airline: a fare you tracked has dropped. A grocer: your usual order is due. Each with Act · Snooze · Dismiss. The reader no longer experiences brands interrupting independently. They see their own commercial relationships, organised around them.

That is where BrandTwins enter. CRM Ads can evolve into a reader-controlled My Next Actions surface, where pending actions become persistent objects to save, snooze or dismiss. In time, a BrandTwin — the customer’s own advocate — could organise these relationships on the customer’s behalf. NeoNet moves from an advertising network towards a customer-controlled relationship network. The direction of optimisation shifts too: from how each brand maximises its response to how the customer manages competing brand demands. That is far larger than newsletter monetisation. It is a new coordination layer between brands and the customers they already have.

The division of labour is now clean. The brand’s CRM and its M-Agents decide what should happen next. Progency takes accountability for executing the recovery programme and its outcomes. NeoNet finds additional reach when the brand’s own attention has failed. CRM Ads carry the decision there. ActionAds let the customer act. The Alpha Ledger measures what was incremental. NeoMails supply the attention, CRM Ads supply unusually valuable demand, and NeoNet clears the two.

The better starting point for every marketing decision is the customer state. Is this person already known? Is there a valuable unfinished action? Does the brand still have their attention? If not, can a permissioned relationship surface restore reach? Only when those questions are exhausted should the marketer enter an auction. Real Reach and Alternative Reach together turn an invisible problem into a measurable one.

For two decades a brand that lost its customer’s attention had two choices: keep shouting into a channel the customer had stopped hearing, or pay a platform to rent the customer back. Relationship Media adds a third: reach them through a brand they are still listening to, with their permission, carrying something they want. That is Never Pay Twice turned from a principle into a route.

CRM Ads are the first proof of a larger shift. Marketing software should route actions to attention, not campaigns to channels.

When owned attention fails, do not rent your customer back. Route to relationship attention first.

Your customer. Someone else’s attention. Your next-best-action.

**

Key points

• Relationship Media is a brand reaching its own known customer through attention earned by another party. It fills the empty quadrant between owned and rented media.
• CRM Ads carry unfinished customer actions — not generic advertisements — to surfaces where the customer is still paying attention, before the brand rents them back from adtech.
• Attention routing replaces channel orchestration: owned first, relationship second, rented last. The market was built around channels, not customer states; agents make routing practical at the scale of millions of customers.
• Permission, not plumbing: the host supplies attention, the relationship supplies relevance, and only permission supplies the right. Blind matching protects data but does not create that right. The reader’s Your Brands permission is the foundation; no payment, credential or KYC actions ever appear on third-party surfaces.
• Alternative Reach — the share of attention-dark customers reachable through the network — is the number that sells NeoNet, and it sits beside Real Reach.
• Price the outcome, not the rectangle: outcome pricing against holdouts lets the host lease trusted attention for verified results, governed by net attention yield.
• The rollout proves the economics first, then earns permission through My Next Actions, then allows specificity on reader-pulled surfaces — a path that leads from advertising towards the BrandTwin.

Search is taking back the readers it lent. The inbox holds the ones publishers own.

For twenty years, publishers ran on a simple bargain. They produced the journalism; the platforms sent the readers. Search sent the most, social sent a surge and then a trickle, and advertising monetised whatever arrived. Newsroom budgets, the number of ad slots on a page, even the length of a headline were all built on the assumption that the readers would keep coming.

That assumption has broken. Search engines are answering questions themselves. Social networks stopped sending readers years ago. The AI assistants that were supposed to become the new front door send almost no one. What remains is the audience that comes back on its own, and for most publishers it is far smaller than the cost base they have built.

A little over a year ago, in News Media’s Renaissance, I argued that the inbox offered news media a way back. The data since then have made the case more urgent, and the tools more concrete. This essay sets out the crisis, how publishers are responding, and why I think the answer lies in an asset most of them already own and underuse:

The email list is the largest first-party audience most publishers have, and the one they control most directly. Run as a product rather than a pipe, it can multiply the revenue each reader brings tenfold.

Underneath the products is a bigger shift, and it is the real argument of this essay: from monetising visits to monetising relationships.

The programme rests on three innovations: NeoLetters, NeoNet and NeoCards. The essay closes with the tests that would prove the argument wrong.

The argument in one paragraph. Search referrals to publishers are collapsing, and AI is accelerating the collapse. Advertising falls with pageviews, and subscriptions have stopped growing in most markets. The readers who stay are the loyal ones, and email is the one channel where a publisher knows them by name and reaches them every day. Yet the newsletter has barely changed in 25 years: it is still a list of links asking for a click. Rebuilt as a product, with editions that update when opened, actions that happen inside the email and collections readers come back to complete, it multiplies four numbers at once: subscribers, newsletters per reader, engagement and yield. None of them has to double: 1.5 × 2 × 1.5 × 2.25 comes to just over ten. The shift underneath is from monetising visits to monetising relationships.

1

The Floor Gave Way

Search, the largest free source of readers publishers ever had, is shrinking fast, and AI is the reason.

The numbers moved from worrying to alarming within a year. Data from Chartbeat for the Reuters Institute’s annual industry report showed Google organic search traffic to more than 2,500 news sites down by a third globally between November 2024 and November 2025, and by 38% in the United States. Chartbeat’s own 2026 playbook is worse: across its network, Google Search referrals fell 40% between July 2025 and July 2026, almost double the fall of the year before. Google Discover fell 34%. Search Engine Journal’s report of the playbook puts Google Search at about 5% of the network’s pageviews, down from 9% two years earlier.

The damage is worst at the bottom. Over two years, small publishers lost 60% of their search referrals, mid-sized publishers 47% and large ones 22%. And the leaders of the industry expect more of the same: in the Reuters Institute’s survey of 280 media executives in 51 countries, publishers expect search traffic to fall by a further 43% over three years, and about a fifth expect to lose more than three-quarters of it.

The mechanism is no mystery. When Google shows an AI summary at the top of the results, users click a traditional link in 8% of visits, against 15% when there is no summary, and they click a link inside the summary in 1% of visits. The summary is the answer. As The Atlantic’s chief executive, Nicholas Thompson, put it, “Google is shifting from being a search engine to an answer engine.”

Nor are the AI assistants filling the gap. In the Reuters Institute’s data, ChatGPT’s share of referrals barely registers: Google sends about 500 times as many referrals from search alone, and about 1,300 times as many once Discover is included. Social went first: over the past two and a half years, referrals to news sites from Facebook fell 43% and from X 46%, after Facebook spent years deliberately deprioritising publisher pages. There is no third platform to lean on now.

Figure 1. Referrals from the platforms have fallen sharply, and publishers expect worse. The figures cover different periods and show direction, not a single index. Source: Reuters Institute, Trends and Predictions 2026, using Chartbeat data.

The consequences are in the headcount. In February the Washington Post cut about a third of its staff, with leadership citing losses and a near halving of search-driven traffic over three years. The year before, Business Insider cut 21% of its workforce so that it could “endure extreme traffic drops outside of our control”, in its chief executive Barbara Peng’s words. For much of the industry, this is no longer a downturn. It is a question of survival.

2

The Appointment, and the Three Doors

The newspaper created a daily appointment. The platforms broke it into visits, and AI is now replacing the visit with an answer.

The newspaper’s great invention was not ink on paper. It was the appointment: a trusted bundle of news, judgement, utility and surprise that arrived every morning at the same time. The web replaced the appointment with a stream. Search replaced loyalty with intent. Social replaced intent with distraction. AI is now replacing the visit with an answer. The publisher’s job is to recreate the appointment, and email is unusually well suited to it, because it combines identity, cadence, personalisation and the ability to act.

Today a reader can reach a publisher through three doors. The owned door: the website or app opened directly, an email, a push alert. The rented door: search, Discover, social and video platforms and aggregators, where someone else’s ranking decides what the reader sees. And a new agent-mediated door: ChatGPT, Gemini, Perplexity and personal assistants that answer on the publisher’s behalf, often without a visit.

For two decades the rented door brought the most readers, and publishers built their businesses around it. That door is now narrowing, and readers are spending more of their time behind it. In this year’s Digital News Report, for the first time, social media and video networks overtook news websites and apps as the most widely used source of online news, 54% against 51%. Use of publishers’ own sites and apps has fallen 12 points since 2020. The agent-mediated door is still small but growing: weekly use of AI chatbots for news rose from 7% to 10% in a year.

Yet the same data contain the clue to what works. Across Chartbeat’s network, overall pageviews held up far better than search, because loyal readers, those who visit on at least eight days in sixteen, stayed steady while new readers fell 14% and returning readers 8%. Chartbeat says loyal readers “are a major reason why overall pageviews haven’t seen a major decline.” The Reuters Institute finds the same pattern in its survey: the most engaged news users are fewer but undiminished in their habits, what it calls “a smaller, but durable, target market.” And at The Atlantic, Thompson has noticed that the readers Google still sends are the valuable ones: “Google is sending us fewer people who aren’t going to subscribe.”

The audience that survives is the one that comes back on its own. Every strategy that matters now is about growing it.

3

Two Revenue Lines, Both Under Pressure

Advertising follows pageviews down, and subscriptions have stopped growing in most markets.

Display advertising is priced per impression, and impressions are a function of pageviews. When a third of the pageviews go, a third of the inventory goes with them, and what remains sells into a programmatic market where most publishers have little pricing power. Cutting costs keeps pace with that for a while. It does not reverse it.

Subscriptions were meant to be the answer, and for a handful of brands they are. The New York Times ended 2025 with 12.78 million subscribers, having added 1.4 million digital subscribers in the year. But the market is winner-takes-most. Across the 20 richer countries the Reuters Institute tracks, the share of people paying for online news is stuck at 17%, unchanged on the year, and it fell four points in the United States. The Institute’s explanation matters: as fewer people visit news sites and apps, fewer enter the subscription funnel at all. In price-sensitive markets, including most of Asia, the share paying is lower still.

Even the Times’s success carries a lesson. More than half its digital subscribers take a bundle or more than one product, and the Reuters Institute notes that the share paying for news alone fell from 17% to 12% last year. Most pay for a bundle that includes puzzles and cooking. People pay most readily for the things they return to every day, not only for the news.

Both revenue lines were built on the same funnel: a stranger arrives from a platform, reads one article, and is shown ads or asked to pay. As the platforms stop sending strangers, the top of that funnel empties. A business model that depends on strangers cannot survive the end of free distribution.

4

What Publishers Are Doing, and What Others Suggest

Publishers are cutting, licensing, diversifying and doubling down on distinctiveness. Most of these responses still rent attention rather than own it.

The Reuters Institute’s survey gives the clearest picture of the industry’s plans. Subscriptions and memberships remain the top revenue focus for 76% of commercial publishers, ahead of display advertising (68%), native advertising (64%) and events (54%). The biggest new opportunity they see is getting platforms to pay for content, cited by 37%, though only a fifth expect AI licensing to be a substantial source of income, and another fifth expect nothing at all.

On content, the consensus is to be more distinctive: more original reporting, analysis and human stories; less of the service and evergreen material that AI answers best. Aftonbladet’s head of AI, Martin Schori, frames the goal as “journalism that can’t easily be summarised in the three bullet points.” The Wall Street Journal’s head of digital, Taneth Evans, argues for doubling down on what makes a publication unique and on direct relationships with readers. AFP’s Latin America bureau chief, Maria Lorente Estrada, describes a shift away from volatile platforms towards owned communities, including premium newsletters. Nine’s Sophia Phan, in Australia, makes the point that matters most here: as search traffic falls, publishers must engage readers directly through newsletters, podcasts and puzzles.

Others are diversifying into businesses that earn per action. Condé Nast says its commerce revenue has grown 170% since 2020 and its US tentpole event revenue ninefold. Many publishers now run vertical platforms in personal finance, autos and product reviews, earning as a trusted middleman. Dotdash Meredith’s chief executive, Neil Vogel, has said Google’s share of his company’s traffic fell from about 60% to about a third, with growth now coming through newsletters and specialised platforms. In Finland, one local publisher is even designing a news product modelled on Duolingo, with stars and tokens that reward participation.

Two cautions run through these responses. First, verticals fed by search inherit search’s fragility. When Business Insider cut staff, it also said it was exiting most of its commerce business because that business depended on search traffic. Second, much of the new effort, from video on YouTube and TikTok to newsletters on Substack and answer-engine optimisation, still builds audiences on someone else’s platform. A strategy built on rented attention is only as durable as the landlord’s next algorithm.

5

The Asset Hiding in Plain Sight

The email list is the largest first-party audience most publishers have, and it is run as a pipe.

Every publisher of any size has one, built from registrations, sign-ups and subscriptions over years or decades. It has three properties no platform can offer. It is known: each address belongs to a person who chose to hear from the publisher. It is direct: no search or social ranking decides whether the reader is referred; mailbox providers still filter and sort, but the publisher does not have to rent the reader back. And it is daily: a publisher can reach the same reader every morning without paying anyone for the privilege.

Yet the newsletter has barely changed in 25 years. A typical edition is a list of headlines, each asking for a click back to the website. Its job is to generate pageviews, the very metric that is collapsing. Its success is measured in opens, a number inflated by Apple’s Mail Privacy Protection and by security scanners that open emails no person has seen. It carries a few banner slots sold at low rates. Most titles reach only a small share of the registered base, and millions of addresses sit dormant.

A newsletter built to send readers somewhere else inherits the economics of somewhere else.

The fix starts with a different question. A link-list newsletter asks: did this email create a visit? A newsletter built as a product asks: did this edition make the next one more likely to be opened? I have called this the Next Open. Recurring attention is not produced by one excellent send; it is produced by a sequence of editions worth opening, and it is the one metric search cannot take away.

Search economics ran from content to a stranger, a visit, a pageview and an impression. Relationship economics run from identity to attention, habit, intent, action and revenue.

The opportunity is not to make newsletters better at producing pageviews. It is to change the unit of publisher economics from the pageview to the relationship.

6

The 10X Equation

Email revenue is the product of four numbers, and each of them can double.

Strip a newsletter programme to its economics and four numbers decide what it earns:

Email revenue = Subscribers × Newsletters per reader × Engagement × Yield

• Subscribers: how many readers are on the list and reachable. Capture on the site and app, a weekly digest for the dormant base, and reactivation all move it.
• Newsletters per reader: how many of the publisher’s newsletters each reader takes. Most publishers run many titles that only a small share of their base has ever seen.
• Engagement: how often readers come back and do something: engaged days per reader, not opens alone, which privacy features and scanners inflate. Live content and a reason to return tomorrow move it.
• Yield: what that engagement earns, from advertising, from actions completed in the email and from subscriptions sold there.

These numbers multiply rather than add. Growing subscribers grows the audience every other lever works on; raising yield raises what all that engagement is worth. None of them has to double. Grow subscribers by half, double the newsletters each reader takes, lift engagement by half and raise yield by 125%, and the product is just over ten. Yield carries the largest multiple because it has the most room: most newsletters today earn from a handful of banners. Newsletters per reader carries the next largest, for reasons section 8 sets out.

Figure 2. An illustrative path to 10X. The multipliers are targets to test, not forecasts.

A word on the claim itself. 10X here means the revenue a publisher earns from its email relationships, not the whole company’s revenue overnight. That stream starts small for most publishers, which is exactly why it can grow tenfold, and why it matters: it is the revenue that does not depend on a platform’s next decision.

Yield is where the arithmetic is most striking. Take an illustrative newsletter with four banner slots, sold at a $4 CPM and 60% filled. It earns about $10 per thousand opens. Now replace one slot with an action the reader completes inside the email. If an action is worth $2, the publisher needs one action in every 200 opens to match all four banners. If it is worth $20, a qualified finance lead or a test-drive request, it needs one in 2,000. A subscription sold in the email at $100 needs one in 10,000.

Figure 3. The breakeven is a threshold to beat, not a forecast. The higher the value of the action, the fewer readers need to act.

7

NeoLetters: Attention You Own

A NeoLetter is composed when it is opened, not when it is sent, and it lets the reader act without leaving the inbox.

NeoLetters are the publishing form of NeoMails, messages worth opening, built for media companies and for brands that publish. They rest on AMP for Email, which lets supported inboxes run live, interactive content inside a message. I described the underlying shift in What Changes When Emails Are Composed at Open, Not Send. For publishers it changes five things.

• Living Emails. Markets, scores, prices and the latest headlines refresh every time the edition is opened, so a morning edition is still current at lunch.
• Stories read in the email. Readers can expand a story inside the edition. That creates more surface for advertising, without asking for a click the reader may not make.
• Polls, quizzes and one-tap registrations work inside the email. Every answer is declared interest, the most valuable data a publisher can hold. I wrote about this years ago in Quizzing in Email; AMP has made it practical at scale.
• Subscriptions started in the email. A reader who has hit the meter can choose a plan and confirm intent inside the message, and where the inbox and payment rail support it, complete payment there too. Elsewhere the email collapses most of the journey and hands off only the final authenticated step. Renewals work the same way, from the reminder itself.
• Reach across titles. A weekly digest carries one item from every newsletter the publisher runs, with a subscribe button beside each. Reactivation gives dormant readers a one-tap choice to stay, go weekly or pause.

Two design rules keep this honest. Every interactive block should end in a link back to the publisher’s site, so engagement adds clicks rather than replacing them. And AMP renders only in supported clients, including Gmail, Yahoo Mail, AOL Mail, Mail.ru and FairEmail, and senders must register and authenticate, so every edition must travel with a strong static version carrying the same content and links. For a large share of readers, the static version is the edition.

None of this is exotic. Brands have spent fifteen years learning lifecycle marketing: segmentation, preference capture, triggered journeys and win-back flows. Publishers can borrow the playbook and apply it to readers rather than shoppers.

8

Publishers as Aggregators Again

AI makes a new publication cheap to start. That turns newsletters per reader into one of the biggest multipliers.

In the early days of the Indian internet I built Samachar to bring the day’s news from many sources into one place, and later experimented with the idea again in Vartam. Techmeme has done it brilliantly for technology for two decades. What limited the model was the cost of curation: every beat needed editors watching it all day.

AI changes that cost. A publisher with trust and distribution can now launch an AI Daily, a Markets Daily, a Health Daily, a City Daily, an edition for each major sports team, a jobs digest or a personal morning edition, each assembled from its own reporting and the best of the web, without building a newsroom for every vertical. Explainers on allied themes such as health and personal finance, and reader-written formats in the spirit of the Times’s Tiny Love Stories, extend the range further. Editors set the rules, check what matters and credit and link the sources the edition draws on.

This is not newsletter personalisation; it is portfolio expansion at a near-zero marginal cost of assembly. Each new edition is a subscription a reader can add with one tap, a new context for ActionAds and a new set of declared interests. It is the most believable way to double the second lever of the equation.

9

NeoNet: Action Without Friction

To raise yield, publishers must sell actions, not impressions.

Ad servers and programmatic exchanges fill slots with banners. A banner asks the reader to click, leave the email, land on a page and act there; most never do. An ActionAd lets the reader act where they already are: check card eligibility, get a quote, request a callback, book a test drive. Because the action happens in the email, there is no click to lose, and advertising becomes lead generation and performance rather than display. I set out the larger idea in Inbox Media Network: How the Next Ad Category Has Been Hiding in Plain Sight.

The first advertiser should be the publisher itself. A publisher with a personal-finance platform, an autos site, a product-review brand, an events business or a subscription to sell already has offers that pay per action. Those can fill the ActionAd slot from the first day, with no sales cycle and no revenue share. This also solves the vertical trap from section 4: email gives the verticals their own distribution, so they stop depending on search.

Then comes outside demand. NeoNet is the network layer: it brings lenders, insurers, carmakers and other performance advertisers to the slots the publisher’s own verticals leave empty, alongside the existing ad stack. The order of priority is simple: direct deals first, the publisher’s own verticals next, then network demand, and programmatic banners last.

Two cautions belong in the design. Regulated offers such as loans and cards should start in the email and finish in the publisher’s app or site, where consent, credit checks and lender disclosures live. And some affiliate programmes restrict their links in email, so affiliate offers should point to the publisher’s own page, which carries the retailer links.

10

NeoCards: Driving the Next Open

News is read once and forgotten. A collection is kept.

Every edition is gone by evening. Nothing pulls a reader back tomorrow except the next headline, and the next headline is now one an AI can summarise without a click. Publishers need magnets: reasons to return that the news alone cannot provide. Games are the proven example. The Times has built its bundle around them, the Wall Street Journal and the Washington Post have followed with daily puzzles, and the Finnish experiment above borrows from Duolingo. NeoCards add a different kind of magnet, built from an asset every established publisher already holds: its archive.

A NeoCard is a collectible card issued inside the newsletters, app and site. A publisher might issue a Set of a hundred front pages from its history, one card a day. Or a Set on the companies that built its country, the great matches of a sport, or the cars that changed a nation’s roads. Some cards can be Living Cards that keep updating after they are collected: a player whose statistics refresh after every match, a company whose share price moves on the card. One version of the game runs like this:

• A few cards a day, inside the emails readers already get. One stays sealed until the reader answers the day’s quiz or poll.
• Duplicates are swapped with friends. Large Sets are hard to finish alone, so trading brings in new readers.
• Each Set runs for a season, and some cards are rarer than others.
• The first finishers get prizes only the publisher can give: a printed edition of the hundred front pages, an ebook, an invitation to an event. Sponsors can add their own.

These mechanics are choices to test, not laws. The principle is what matters: NeoCards turn ephemeral media into persistent objects.

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

Cards work on all four levers. They lift engagement, because every missing card is a reason to come back. They grow subscribers, because trading pulls in new readers. They raise yield, because a brand can present a Set as a new kind of sponsorship. And they route readers into the verticals: a card in a personal-finance Set can open a credit-score tool, a card in an autos Set a test-drive offer. For readers drifting away, Atomic Rewards can add a small incentive to open, used sparingly and only where attention is fading.

One design question comes up early: should the cards live in the publisher’s own album? My strong preference is no. A private album starts with no collectors and no one to trade with, so Sets stall before they are finished. Cards belong in one shared album that readers of many issuers already use. The publisher keeps what matters: its name on every card, its own shelf in its app and site, its distribution and the data on its own Sets. Other issuers never see its readers, and it never sees theirs.

11

The Second Reader

Publishers used to publish for millions of humans. Soon they will publish for millions of humans and millions of agents.

Personal AI agents are arriving. The Reuters Institute points to tools such as Huxe and OpenAI’s Pulse that compile personalised news briefings, and three-quarters of the executives it surveyed expect such agents to have a large impact on the industry within a few years.

The two readers want different things. The agent wants structure: facts, entities, prices, provenance, permissions and actions it is allowed to take. The human wants what no summary delivers: narrative, judgement, humour, surprise, play and a sense of belonging. An agent can summarise a story. It cannot be surprised by one, laugh at a cartoon, argue in a poll or feel the satisfaction of finishing a Set. The future news product may need two editions inside one object: one that people enjoy, and one that agents can understand and act on. A Living Email, composed at open, is a natural place for both to meet.

Habit has to be designed, too, because it can no longer be inherited. Among 18–24-year-olds who did not read a newspaper last week, the Reuters Institute finds that 56% have never read one regularly, and most people of every age say their news routines vary from day to day. Interest in news has fallen 13 points since 2021. The daily newspaper habit is fading; a daily inbox habit has to be earned, edition by edition.

12

How the Pieces Compound

The order matters more than any single tool: identify, earn attention, build habit, learn intent, and only then monetise.

There is an obvious failure mode. A publisher sees a new email surface and fills it with advertising, repeating the web’s old mistake in a new place. An address is not an audience, an open is not attention, and an empty slot is not inventory. The sequence has to be disciplined, as Figure 4 shows. A network of weak relationships merely networks weakness; NeoNet becomes valuable only after the editions have earned their opens.

Figure 4. Each step earns permission for the next.

Run in that order, the loop compounds. The digest and reactivation bring more of the base into play. Cards and Living Emails give those readers a reason to open every day. Each open carries live content, a question and an action, so engagement produces both declared interest and revenue. Declared interest makes the next ActionAd more relevant, which raises yield. Yield, much of it from the publisher’s own verticals, funds better editions and richer Sets.

The website does not disappear in this model; its role changes. It remains the home for deep reading, archives, video and tools. What changes is the hierarchy: the website no longer carries the whole economic burden of the relationship. Search rewarded publishers for attracting strangers once. The inbox rewards them for keeping readers every day.

13

What Would Prove This Wrong

Every claim in this essay comes with a test that could kill it.

Each test can be read within months, and each uses a concurrent, randomised holdout: part of the list keeps today’s best newsletter while the rest gets the new one. Last year’s numbers are not a baseline.

• Inbox AI eats the open. If AI assistants in email apps summarise newsletters the way AI Overviews summarise search, the open itself erodes. Test: human opens per engaged reader over twelve months. If they fall as search clicks did, the inbox is no refuge.
• Actions stay below breakeven. If, after six weeks, actions per thousand opens sit below the level that matches display, ActionAds do not pay. Stop and redesign the offer.
• Interactivity is not the driver. AMP reaches only some inboxes. If the static version captures most of the gain, the interactive layer is not earning its cost.
• Cards are a novelty. If engaged days per reader fall back to the holdout level once the first Set ends, cards are a gimmick, not a habit.
• Reach costs reputation. If sending the digest to dormant readers pushes spam complaints towards the limits Gmail sets for bulk senders, reach is being bought with deliverability. Narrow the audience before widening it.

14

Where to Start

Start with ninety days and four pilots, each with an owner, a clock and a holdout.

• Yield: ActionAds for the publisher’s own verticals, and subscription offers started in the email. Measured as revenue per thousand opens against display.
• Subscribers and newsletters per reader: a weekly digest with one-tap subscribe, one AI-assembled vertical edition, and a reactivation flow for a tested slice of the dormant base. Measured as new subscriptions and declared interests per thousand sent, with complaint rate as the guard.
• Engagement, Living Emails: a Living Email version of the flagship daily edition. Measured as engaged days per reader and the Next Open rate.
• Engagement, NeoCards: one Set, drawn from the archive. Measured as engaged days per reader against the holdout.

None of this needs a new audience, a new platform or a bigger marketing budget. It needs publishers to stop treating newsletters as a way to send readers somewhere else, and to start treating them as the place where the relationship lives.

Search was a gift, and it is being taken back. The inbox was never a gift. Publishers earned it one subscriber at a time, and no answer engine can summarise it away. The publishers who turn their newsletters into messages worth opening and objects worth keeping will not need the traffic back.

For twenty-five years, publishers taught themselves to maximise the value of the pageview. The next twenty-five will be about maximising the value of the reader.

Messages Worth Opening, Objects Worth Keeping

NeoMarketing is a way of running marketing so that brands stop paying to win back customers they already have. It rests on two pillars and one operating model.

1

The problem

Brands pay to win back the attention of customers they already have.

The pattern is familiar. A customer buys, and the relationship goes quiet. The brand’s emails are opened less often, then not at all. Months later the same customer is reached again through paid ads on someone else’s platform, and the spend is booked as acquisition.

The money spent buying back attention the brand once had is AdWaste. It grows because attention decays between purchases, and because almost everything a brand sends is designed to be consumed and forgotten. Nothing from the last message remains in the customer’s hands, so every reminder starts from zero.

2

The Three NEVERs

NeoMarketing is a promise in three parts.

• Never Lose Customers. Keep attention before customers drift. Marketing should notice attention fading and reverse it while the customer is still cheap to reach, rather than discovering the loss once they have gone.
• Never Pay Twice. If the brand acquired a customer once, it should exhaust its own relationship with that customer before renting access to the same person from an ad platform. Paying a second time for a customer the brand already had is the clearest form of AdWaste.
• Never Pay Fixed. Move what the brand pays towards measured outcomes. Agree the baseline, measure the uplift, and pay increasingly for what changed rather than for licences, seats or sends.

3

Two pillars

Messages worth opening. Objects worth keeping.

NeoMarketing needs two different primitives because attention works on two different time scales. A message does a job at a moment: it can win attention now. An object has properties that last: it keeps its state after the moment is over. Marketing has spent thirty years improving the first and has barely designed the second.

So NeoMarketing rests on two pillars, NeoMails and NeoCards. One creates the moment; the other gives the moment somewhere to accumulate.

 

Figure 1. One doctrine, two pillars, one operating model.

4

NeoMails

NeoMails do four jobs: Sell, Notify, Deepen, Recover.

Job	What it does	Example
Sell	Creates and converts demand	A launch, an offer, a reminder
Notify	Services the customer; often the brand’s highest-attention emails	Order shipped, statement ready, flight confirmed
Deepen	Strengthens the relationship between purchases	A digest worth reading
Recover	Brings drifting and lapsed customers back before the brand pays to reacquire them	A reason to return, sent as attention starts to fade

Table 1. The four jobs of NeoMails.

Two more names sit inside the pillar. NeoLetters is the publishing form of NeoMails: a recurring, curated edition that gives people a reason to open between purchases. Brands use it to deepen customer relationships, often as a weekly digest; publishers use it to grow and monetise their audiences. Same machinery, different economics. Living Emails are interactive emails that can be updated after they are sent, so the inbox holds something current rather than something stale.

Earlier essays also described NeoNet, a network that connects attention across brands and turns it into outcomes, including through ActionAds. It is a later layer that sits across both pillars, not a pillar itself. Networks need many participants before they work, and NeoMarketing starts with what a single brand can do with its own customers.

A note on the name. Earlier essays used NeoMails for one specific thing: a daily relationship email. The word now covers the whole of a brand’s next-generation email, across all four jobs. The daily relationship email remains, as one pattern within Deepen and Recover.

5

NeoCards

NeoCards are built to KEEP: Collect, Evolve, Exchange, Play.

A NeoCard is an object worth keeping: a fixed front that carries identity, and a reverse that keeps growing after the card is acquired. Collecting creates progress, evolution creates return, exchange creates reach, and play creates declaration, meaning things the customer chooses to tell the brand. Underneath sits a trust layer of identity, provenance, permissions and history. The essay NeoCards sets out the idea in full.

6

How the pillars connect

NeoMails earn attention. NeoCards compound it.

Figure 2. Each pillar gives the other something to work with.

The connection runs both ways. A message can deliver a card: the order confirmation a customer was always going to open can carry one, so the brand attaches an object to attention it already has rather than launching a new programme. A card, in turn, gives the next message a reason to be opened: one of your cards has changed, your set is nearly complete, a friend has sent you one.

NeoMails improve the economics of the next interaction. NeoCards improve the economics of the time between interactions. Over time the brand builds a relationship it does not have to rent back from a platform, because its customers hold something that brings them back. That is the practical meaning of Never Pay Twice.

7

The operating model

Agents operate. Accountability proves. Alpha pays.

Both pillars run on the same operating model.

• AI agents do the work no marketing team could staff: writing the digest, choosing which card goes to whom, deciding when to try to recover a lapsed customer, and updating cards as the world changes.
• Every claim is measured against a holdout and a baseline agreed in advance. Real Reach, the share of the base that engaged in the last ninety days, shows whether owned attention is growing. Issuer Recall shows whether customers remember who gave them their cards.
• The economics follow the results. Alpha pricing starts from an agreed baseline and ties fees for the upside to incremental outcomes above it. The aim is not to remove every fixed charge overnight; it is to make impact visible and payable. That is what Never Pay Fixed means in a contract.

8

What would prove this wrong

The architecture makes three claims that can each be tested and each fail.

• NeoMails do not earn durable attention. Against a holdout, they fail to raise opens, actions or recovery, and the brand keeps paying to win customers back.
• Cards do not compound it. NeoCards raise engagement without raising return, so attention is earned but not kept.
• Outcome pricing proves unworkable. Brands and vendors cannot agree baselines, and Never Pay Fixed stays a slogan.

If those tests fail, the architecture fails where it matters: in behaviour and economics, not in storytelling.

9

Where to start reading

Four essays are enough to start.

• Email’s Next Act. How email changes when messages are built to be opened, acted on and returned to.
• Kept Media. Why persistence is the missing dimension of media, and why owned and kept are different things.
• When a Card Becomes Software. What kind of object a card is: persistent identity, evolving capability.
• NeoCards. The name, the four jobs, the trust layer and the ethos.

**

Something kept

NeoMarketing starts from a simple observation. Brands lose customers they already have because nothing they send is worth keeping, and then pay to win those customers back. The answer is to earn attention with messages worth opening, compound it through objects worth keeping and returning to, and be paid for the difference that makes.

An audience receives what you send. A relationship keeps something after you leave.

**

Key points

• Brands pay to win back the attention of customers they already have. That spend is AdWaste.
• The Three NEVERs: Never Lose Customers, Never Pay Twice, Never Pay Fixed.
• Two pillars: NeoMails, messages worth opening, and NeoCards, objects worth keeping.
• NeoMails do four jobs: Sell, Notify, Deepen, Recover. NeoCards are built to KEEP: Collect, Evolve, Exchange, Play.
• A message can deliver a card; a card gives the next message a reason to be opened. NeoMails earn attention; NeoCards compound it.
• Agents operate. Accountability proves. Alpha pays.

Previous essays have described the same object without naming it. This one gives it a name: NeoCards, objects worth keeping.

1

Why a name now

Over the last two months I have written about one object from several directions. It began as a card arriving in an email, with a front that never changes and a flip that can. It became a card that accumulates history, then a card that rewards attention where points reward purchases. Most recently it became something a customer keeps rather than consumes, and a small piece of software the person holds.

Multiple essays have described the same object without naming it.

Ideas without a name get discussed as features of something else. Named, they can be discussed as a category, tested, and built on. So the object now has a name: NeoCards. The name is not a new feature. It names an object whose behaviour, economics and architecture differ from a message, a web page or an app.

The name places it beside NeoMails within NeoMarketing. NeoMails are messages worth opening; NeoCards are objects worth keeping. A separate primer sets out how the two fit together. This essay is about the object.

2

What a NeoCard is

A NeoCard is an object worth keeping: a fixed front that carries identity, and a reverse that keeps growing.

The front holds everything that makes the card recognisable: its artwork, name, number, set, edition and issuer. It never changes, because ownership, trade and completion only mean something if everyone is holding the same thing.

The reverse holds everything the card can do. On the day it is collected, that is a story, much as the back of a cigarette card carried one a century ago. Over the months that follow, the issuer can add a guide in the owner’s language, a short film, a question, fresh data or a relevant offer. The card gains capability without becoming a different card.

A NeoCard can arrive anywhere: in an email, attached to an order confirmation, through a code on a pack, or from a friend. Once acquired, it lives in the owner’s album, whichever route it took.

A message is something you were sent. An app is somewhere you go. A card is something you have.

3

Built to KEEP

NeoCards are built to be kept, and four jobs earn that.

Figure 1. Four jobs earn the keep, and each creates something the brand can measure. Trust sits underneath all four.

Collect. Every card belongs to a finite set, so what the owner has and what they are missing are both visible. An incomplete set is a reason to come back that no reminder email can match. Collecting creates progress.

Evolve. The reverse gets deeper after the card is acquired. A card that held only a story on day one may hold a spoken guide, a film and a challenge by the end of the year. Evolution creates return.

Exchange. Cards move between people, as gifts or trades. A duplicate becomes a reason to contact a friend, and the brand earns reach to that friend rather than buying it through media. Exchange creates reach.

Play. The owner does things with the card: listens, watches, answers, declares a plan, acts on an offer. What the owner declares (going in December, with family) is written by them, on an object they keep. Play creates declaration.

Collect creates progress. Evolve creates return. Exchange creates reach. Play creates declaration.

Play means interaction, not gamification. A bank’s card on investing ideas is played by learning and answering, not by spinning a wheel.

4

Trust by design

A card that can change must prove what it was, who changed it and why.

Trust is not a fifth job. It is the layer underneath the four, and it has four parts: identity, provenance, permissions and history. Every card has a permanent identity. Every version the issuer publishes is kept. Every update is dated and attributed. The owner’s own notes and answers are stored apart from the issuer’s content, and belong to the owner.

The issuer can add to a card. It cannot silently take it back or rewrite what was there. Without that guarantee, Evolve would be a liability: an object that changes without a record is an object nobody can rely on.

5

The ethos

Five rules keep NeoCards honest.

• The brand issues the card but is not its subject. A paint company issues a set on the colours of India, not twenty of its own products. The set must be worth collecting even if the logo disappeared.
• Rewards amplify play; they do not define the card. A prize can start a habit. If a card needs a prize to be kept, the card is weak.
• The card may grow deeper. It must not grow wider. Capability arrives one layer at a time, and every layer is about the same subject. A card that fills with unrelated functions becomes an app nobody opens.
• Identity is fixed. Rendering can adapt. One card can speak Kannada to one owner and English to another, and still be the same card.
• Kept is not enough. Kept and returned to is the asset. A card that is kept but never revisited is an archive, not a relationship.

6

How the idea developed

The idea was not designed in one piece; it was assembled.

Got, Got, Need introduced the object: a card that arrives by email, with a permanent front and a flip that can change, and an argument that the inbox was the right room for it.

IndiaVotes added time: a physical card freezes a moment; a digital card accumulates history.

The Brand Digest showed what cards do for a channel: cards, sets and albums give a stateless medium memory.

Points Reward Purchases. Cards Reward Attention. made the economic case. Points cover the moments of purchase; cards cover the far larger stretch of time between them, with the brand as issuer rather than subject.

Kept Media found the missing dimension. Paid, owned and earned describe how media arrives; kept describes what remains. It also corrected the first essay: I had confused the best delivery rail with the thing being delivered.

When a Card Becomes Software described the object’s physics. The front is identity; the reverse is software; the card can change without becoming a different thing.

NeoCards gathers all six under one name.

Figure 2. Six essays, each adding one piece, before the object had a name.

7

What would prove this wrong

A name is only worth having if the thing it names works. Three failures would show it does not.

• No return. Owners collect cards but never come back to them, and the reverse grows for nobody.
• No credit. Engagement rises while Issuer Recall stays flat. People love the cards and forget who issued them.
• Wider, not deeper. Issuers keep adding unrelated functions until the card becomes an app in miniature, and owners stop turning it over.

The test is not whether people like the idea of cards. It is whether a card received today raises the chance of attention tomorrow.

**

Objects worth keeping

NeoMails are how a brand earns attention, one message at a time. NeoCards are how it keeps that attention once it has been earned. The primer that follows explains how the two pillars work together.

NeoMails earn attention. NeoCards compound it.

**

Key points

• NeoCards names an object six essays had described: an object worth keeping.
• A NeoCard has a fixed front that carries identity and a reverse that keeps growing.
• NeoCards are built to KEEP: Collect, Evolve, Exchange, Play, creating progress, return, reach and declaration.
• Trust sits underneath: identity, provenance, permissions and history.
• Five rules: issuer not subject; rewards amplify play; deeper not wider; identity fixed, rendering adapts; kept and returned to is the asset.
• A message is something you were sent. An app is somewhere you go. A card is something you have.

A digital card can change without becoming a different thing. That one property makes it something new: a small piece of software that the person, not the vendor, holds.

1

Between the message and the app

In Kept Media I argued that marketing has a missing dimension: the things a customer chooses to keep. This essay asks a narrower and stranger question. If a digital card is worth keeping, what kind of object is it?

Start with the two objects most digital relationships are built from. The first is the message: an email, a notification, a chat, a post. Messages are weightless. They arrive anywhere, cost almost nothing, and disappear almost as soon as they are read. Nobody holds a message; they receive it.

The second is the app. Apps solved persistence. They remember who you are, keep state between visits, and can do almost anything. But they solved it the expensive way. An app has to be found, installed, logged into, learned and updated, and once inside it you are in the vendor’s room, on the vendor’s terms. Most people use a handful of apps. Most brands will never earn a place among them.

There are things in between, and section 8 deals with them. But none has become the everyday way a person holds a small piece of a brand’s world.

What is missing is not another format between the message and the app. It is a different ownership model.

2

The front is identity; the reverse is software

A printed card has two faces, and both are fixed the day it leaves the press. A web page can change at will, and its address may last for years, but nobody holds a copy of it. An app has persistent identity and evolving capability, but it belongs to the vendor and carries all the weight described above.

A digital card can split the two jobs between its two faces. The front holds identity: the artwork, the name, the number, the set, the edition, the issuer. The reverse holds capability: text, audio, video, live data, questions, forms, actions and history.

The front is identity. The reverse is software.

That split produces an unusual property, one that neither paper nor web pages have:

It can change without becoming a different thing.

Figure 1. The card combines the message’s ease of arrival with the app’s persistence, and adds the one thing neither offers: the person holds it.

3

Why only the reverse changes

Building a working card makes this rule concrete. There is a constant temptation to make the front live: a sky that matches today’s weather, a badge that updates, a picture that changes with the season. It should be resisted.

Identity must hold still for ownership, trade and completion to mean anything.

If the front’s identity changes, every collecting behaviour breaks. The owner has to ask which version they hold. Two people trading must be holding the same object, or the trade means nothing. A completed set needs a stable grid of faces, or completion is never finished. The front works like a passport. It is not frozen because the technology cannot change it; it is frozen because meaning requires it to stay put. A front that changes every week turns the card into a web page wearing a border.

But identity is not the same as appearance, and the rule needs to be stated precisely. The artwork, card number, set, edition and issuer never change. How they are rendered can adapt to the person reading them: the language of the words, the size of the text, the contrast, a narration for someone who cannot see well. A Kannada reader and an English reader are holding the same card.

Identity is fixed. Rendering can adapt.

Figure 2. Three layers move at three speeds: identity never, rendering per reader, capability over time.

The front may also carry one thing that comes and goes: a small signal that something new is waiting on the back. Without it, nobody turns the card over. But the signal is a notification, not content.

4

What the reverse can become

Take a single card: Hampi, number 12 of 45 in a set of India’s World Heritage sites. On the day it is collected, its reverse does what the back of a cigarette card did a century ago. It tells a story. Hampi was the capital of the Vijayanagara Empire until 1565, when it was sacked and abandoned. Its famous Stone Chariot is in fact a shrine to Garuda, the eagle who carries Vishnu, and it appears on the back of India’s fifty-rupee note.

Then the reverse gains capability, one step at a time.

• The story the card arrived with.
• A short guide, spoken in the owner’s language, with captions that follow the voice.
• A film of sunrise over the boulders.
• A challenge with one right answer. The chariot’s sheltered undercarriage still carries traces of what? Mineral paint: the sculptures were once painted.
• The owner records an intention on the card: going in December, with family.
• The issuer offers a relevant next step, such as a trip.
• In future, the card could host a bounded expert that answers questions only from the sources behind that card. A general chatbot on every object would be a novelty; an expert that knows the object’s sources and history could be useful. This is a direction, not something the card does today.

One capability on the Hampi card shows what the reverse can do that the world cannot. The pillars of the Vittala Temple ring with musical notes when tapped. Tapping them is now banned, to protect them. On the card, the owner can play them.

There is a danger in all this capability, and it needs a design law rather than a warning. If the reverse fills with story, data, film, forms, offers and history all at once, the card becomes a small app, and small apps are exactly what people do not open.

The card may grow deeper. It must not grow wider.

The Hampi card stays about Hampi. Its capability arrives through progressive disclosure: one new layer at a time, each one about the same place, never a menu of unrelated functions.

Figure 3. Each step adds depth to the same object. Ask, drawn dashed, is a future direction.

5

Objects instead of applications

Software today is organised around destinations. To check a booking, you open the airline’s app. To learn about a fund, you visit the fund house’s website. Functionality lives in a place, and the customer travels to it.

The card reverses that. There is no Hampi app to open. The Hampi card is already in the owner’s album, and new capability arrives on it. The thing comes to the customer, and stays. Kept Media makes the commercial case that the delivery rail does not define the object: distribution is transient, ownership is persistent.

That does not make the app obsolete. It changes what requires an app. Complex, frequent tasks still belong in one. The card is for the smaller unit: one thing, one identity, one growing relationship. A brand whose forty-five cards sit in a customer’s album has forty-five small pieces of software installed, without anyone installing anything.

6

One object, every language

A printed card has its language fixed on it. Digital content mostly copies that habit: the English site and the Hindi site are separate things, built and maintained separately, and they drift apart.

Applying the rule from section 3 removes the problem at the root. There is no English Hampi card, Hindi Hampi card and Tamil Hampi card. There is one Hampi card. It speaks Kannada to one owner, Tamil to another and English to a third, with the same number, the same edition, the same history and the same place in the set.

Language becomes rendering, not identity.

India’s Constitution lists twenty-two languages in its Eighth Schedule, and most brand communication still reaches customers in English, which is not among them. The reason has always been cost. Writing, translating, narrating and maintaining every object in many languages was absurdly expensive. AI changes that arithmetic. A card can now be written once and rendered, spoken and kept up to date in every language its owners read. The idea is not specific to India: any issuer can maintain one object and render it by language, reading level or accessibility need.

Cheap translation is not good translation. Every rendering still needs a native reader before it reaches a customer, and regulated content needs more than that. But the architecture no longer makes the attempt prohibitive. That is the change.

7

If it changes, provenance is mandatory

Any object that keeps changing invites an obvious challenge from its owner: if the issuer can change it, what exactly do I own?

An object that can change must be able to prove what it was, who changed it, and why.

The answer cannot be “trust us”. A wrong fact on a static card is embarrassing. A changing object that quietly rewrites itself is a breach of trust. So a serious card is built with six properties.

Property	What it means
Permanent card ID	The card’s identity never changes, whatever is added to its reverse
Versioned issuer content	Every version of what the issuer published is kept, not overwritten
Append-only update log	Each change is dated and added to the history; nothing is deleted from it
Signed authorship	Every update carries proof of which issuer published it
Separate owner state	The owner’s notes, answers and marks are stored apart from issuer content, and the issuer cannot rewrite them
Exportable record	The owner can take the identity, history and their own notes away in an open format

Table 1. Six properties that let a changing object prove its own history.

In Kept Media I set out the rule: the issuer owns the publishing right, the customer owns the card, and the album’s operator runs the system. These six properties are what make that rule enforceable rather than a promise.

No blockchain is required. Verifiable authorship and append-only history are. The stakes rise as cards carry more: financial education, certificates, records of achievement. A card that cannot prove its own history will not be trusted with any of them.

8

Hasn’t this been built?

A technically literate reader will point out that pieces of this already exist. They are right, and the precedents are worth reading carefully, because each proves part of the combination.

Wallet passes are the closest. A boarding pass keeps its identity while its gate number updates; a loyalty pass can persist for years and carry images and live fields. What passes were designed as is utility credentials: tickets, memberships, coupons. They were not designed to tell a story, sit in a set or collect a personal history.

Mini-programs inside super-apps such as WeChat proved that useful software can run without a conventional install. But they generally belong to the platform they run inside, not to the person using them.

NFTs proved that digital ownership and provenance could be built, and some carried changing content. Their difficulty was not a missing capability. It was that speculation and price became the centre of the experience, and when interest in price faded, there was little else to hold on to.

Phone widgets showed that a small, live surface can be useful. But an operating-system widget extends an app; it has no identity of its own that the person keeps.

Each precedent proves part of the combination. None was designed as a person-held object with a fixed identity, a reverse that grows, and a place in a collection.

The precedents also name the main risk. The companies that run phone wallets could add sets, media and collecting to passes before an independent version reaches scale. If that happens, the idea will have been right, and someone else will own it.

9

The limits of owning software

There is an honest limit to all of this, and it should be stated plainly.

The front can outlive its issuer. The reverse cannot.

A prototype of the Hampi card exists as a single file. Opened on a laptop with no connection, its front, story, history and musical pillars all work, because the pillar notes are generated by the browser itself. The spoken guide depends on the voices installed on the device, and the typefaces fall back to the device’s own. Nothing new can arrive on its back. New layers depend on an issuer still publishing them and an operator still running the system.

The principle matters more than the prototype. A portable card should stay readable offline; its live layers simply stop arriving. Owning software has always been partial: you hold the object, and someone else holds the ability to keep it alive. For cards, portability has to be designed in from the start. The identity, the history and the owner’s own notes should be exportable in an open format. A card should remain viewable if its issuer stops publishing. And the operator should be bound to keep both promises.

10

What would prove this wrong

The conditions under which the idea fails should be written down before anyone builds on it.

• Change goes unnoticed. Owners do not come back to cards when new capability arrives. The layers are built, and nobody uses them.
• Trust breaks. Even with a visible history, owners are uneasy about an object that changes, and fixed content turns out to be the better design.
• The platforms absorb it. Phone wallets add sets and media to passes, and the independent version never reaches scale.
• Languages do not pay. Rendering in many languages costs more to create and review than the attention it earns.
• The card grows wider. Issuers ignore the design law, cards fill with unrelated functions, and owners stop turning them over.
• The app remains the natural unit. People prefer visiting destinations to holding objects. Then the card should stay content, not become a new kind of software.

All six are plausible. That is why the idea should be tested as an object rather than defended as a metaphor.

**

Something you have

Digital life has given us two familiar kinds of thing. A message is something you were sent. An app is somewhere you go.

The card is a third kind.

A message is something you were sent. An app is somewhere you go. A card is something you have.

**

Key points

• Messages are weightless and vanish; apps persist but are heavy and vendor-owned. What is missing is a different ownership model.
• The front is identity; the reverse is software. A card can change without becoming a different thing.
• Identity is fixed; rendering can adapt. The same card can speak Kannada to one owner and English to another.
• The reverse can grow from Read to Listen, Watch, Answer, Declare, Act and eventually Ask. The card may grow deeper; it must not grow wider.
• Software can be organised around objects people hold rather than destinations they visit. That does not make the app obsolete; it changes what requires one.
• No blockchain is required. Verifiable authorship and append-only history are.
• Wallet passes, mini-programs, NFTs and widgets each prove part of the combination. None was designed as a person-held, growing, collected object.
• A portable card should stay readable offline; its live layers simply stop arriving.
• A message is something you were sent. An app is somewhere you go. A card is something you have.