Published October 7, 2026
A customer never announces that they are leaving. There is no cancellation notice for a shopper who stops shopping, no exit interview for an investor who stops investing, no farewell from a traveller who books elsewhere next time. The brand learns about it later, when a report finally shows no purchase for a defined period and the customer is filed under a word that sounds like a diagnosis: dormant.
By then, much of the story has already happened. The customer may have opened fewer messages. They may have stopped answering questions, checking new arrivals, using the app or visiting the site between purchases. The brand may have slipped out of the small set of names that come to mind when the category does. Purchase inactivity is the visible end of a process that may have started months earlier.
That matters because marketing spends heavily at two moments. It spends when the customer is buying: promotions, journeys, loyalty rewards, cross-sell and upsell. And it spends again after the customer has gone: win-back campaigns, discounts, retargeting, paid media and, eventually, reacquisition. The interval between those two moments is strangely under-designed. It is the time when the customer still knows the brand, still has given permission to hear from it, and can still be reached at very low cost — but has less and less reason to pay attention.
This essay is about that interval. Its thesis is simple: dormancy is an outcome, not an event. For many categories, the slide begins in attention before it becomes visible in transactions. If that is true, customer retention should begin earlier too.
Retain before you Recover. Recover before you Reacquire.
1
Dormancy Is an Outcome, Not an Event
Dormancy is recorded at the end of a slide that began long before.
Look at how dormancy is defined. It is almost always a statement about something that did not happen: no purchase in twelve months, no transaction in two cycles, no renewal by a deadline. The window changes by category, but the measure is retrospective by construction. It tells the brand that something did not happen. That makes it useful for reporting and late for intervention.
The problem is sharpest in categories without contracts. Nobody cancels a supermarket, a fashion brand, an airline or a restaurant. They visit less often, spend less often, and eventually drop out of the active base without ever saying so. Research on customer defection has long relied on exactly this kind of behavioural trail. In a study of a grocery retailer’s loyal customers, Buckinx and Van den Poel found that past behaviour — how recently, how often and how much customers bought — was among the most useful information for predicting which of them would partially defect. More recent work, using a large financial-services panel, shows that behavioural measures tracked over time are valuable inputs to churn prediction in their own right. Practitioners who build churn models treat declining engagement with marketing — fewer opens, fewer clicks, fewer responses — as one of the signals to watch.
None of this proves that attention always falls before purchases do. It supports a more useful hypothesis: the relationship often changes before the accounting definition of dormancy catches up. The practical implication is to watch the path, not only the endpoint.
Think of a customer who once engaged with a brand every week. Over time the pattern becomes every fortnight, then once a month, then not at all. If that customer still buys occasionally, a transaction-only dashboard calls them active. An attention dashboard would say something else: the relationship is weakening. The difference is the difference between a fire alarm and a smoke detector. Dormancy tells the brand the building is already burning. Attention decay can tell it something is changing while there is still time to act.
The consequence is a timing error. When dormancy is the trigger, every intervention starts after the most important part of the relationship has already been lost.
2
The Five Stages
Between active and dormant there are three stages marketing rarely names.
Most customer reporting uses two states: active or not. That segmentation is too coarse for this problem. The slide has at least five stages, and the three in the middle are where it is decided.
Figure 1. The five stages of the slide. Attention fades in the middle three; purchases stop only at the end.
Table 1. The five stages, defined by rules rather than fixed numbers
| Stage | Rule |
| Active | Engaging, and buying within the normal purchase cycle |
| Less attentive | Engagement well below the customer’s own recent baseline |
| Quiet | No verified engagement across a window set by the brand’s own cadence |
| Inactive | No engagement in any owned channel across a longer window |
| Dormant | No purchase for well beyond the category’s normal purchase interval |
The words are universal; the numbers are not. A grocer, a fashion retailer, a fund house, an airline and a car company cannot share the same thirty-, ninety- or 365-day definitions. A customer who has not bought milk for a month is in a very different place from one who has not bought a car for a year. So the rule is to publish the logic and calibrate the thresholds from the brand’s own data. The Alpha Audit, or a pilot, is where “well below baseline”, “quiet” and “dormant” acquire numbers.
There is a second reason to avoid fixed day counts. Attention is personal. A customer who normally engages with every edition and suddenly goes silent is sending a stronger signal than one who has always engaged with one in ten. For a weekly reader, three silent weeks may matter; for a monthly reader, they may be normal. Good retention logic is therefore partly absolute and partly relative to each person’s own rhythm.
The five stages also create a language for intervention. Active customers need relevance. Less attentive customers need renewed usefulness. Quiet customers need a reason to re-engage. Inactive customers need deliberate recovery. Dormant customers may, eventually, need to be reacquired.
The most useful feature of the model is the transition, not the label. A dashboard should not merely count how many customers sit in each stage. It should show the flows — Active to Less attentive, Less attentive to Quiet, Quiet to Inactive — and the smaller flows back when customers re-engage. That turns the base from a static pyramid into a moving system. Two quarters can show the same number of active customers and describe very different businesses, if one of them has a much larger stream heading towards Quiet.
And the order is a hypothesis, not a law. Many customers pass through Less attentive and Quiet before they become Dormant. Whether that sequence holds strongly enough to act on is the first thing each brand should test. If it holds, the middle stages are not noise; they are the warning.
3
Attention Moves First
Before a customer goes dormant in purchases, they often go dormant in attention.
There are three kinds of retention, and marketing measures only two of them.
Customer retention asks whether the customer bought again. Revenue retention asks whether their value continued. Both are measured everywhere, reported every quarter and discussed at board level. Both are lagging measures. The third is attention retention: whether the customer kept giving the brand voluntary attention between purchases. It is a description, not a product or an acronym, and it is almost never measured. The working hypothesis is temporal: attention weakens first; customer activity and revenue weaken later.
Figure 2. Conceptual timing, not observed data: attention as the leading signal, customer and revenue retention as lagging outcomes.
The leading measures are simple, and most brands already have the data to compute them:
- Real Reach: the share of the base that has engaged in the recent window, as opposed to the size of the list.
- CRR (Click Retention Rate): how quickly an engaged base decays over time.
- Habit Rate: the share of customers who engage with most of the contacts they chose to receive, rather than producing isolated spikes.
- Engaged → Quiet migration: how many customers crossed from engaged to quiet in the period.
There is a useful asymmetry here. Transactions are sparse; attention can be observed far more often. In a low-frequency category a brand may wait months to learn whether a purchase will repeat, but it can see every week whether the relationship is still producing voluntary actions. That makes attention valuable not only as an outcome but as a diagnostic. It gives the brand more chances to see deterioration before the economic event is final.
This is not an argument for watching customers more closely. The useful signals are the ones customers create through ordinary voluntary behaviour: opening something useful, answering a question, revealing a Card, clicking for more, changing a preference, visiting an owned property. The aim is to recognise fading interest, not to maximise instrumentation.
Nor is it an argument for being seen every day. The Ehrenberg-Bass Institute describes the goal as mental availability: the brand’s propensity to be thought of or noticed in buying situations. Attention retention is not the same thing, but it can feed it. A brand that remains periodically useful has more chances to refresh memory than one that appears only with a promotion. The aim is modest and precise: keep the brand easy to retrieve when the buying moment returns.
4
Marketing Arrives Late
Spend concentrates where the customer is buying and where the customer has gone. Almost nothing is designed for the fade.
Marketing already has sophisticated machinery at both ends of the relationship. Around the transaction there are campaigns, triggered journeys, recommendations, loyalty points, cart recovery, cross-sell and upsell — machinery built to convert an existing intent into revenue. After the customer has gone there is a second machine: win-back. The brand offers a coupon, a sale or a “we miss you” message, and when owned channels stop working, paid media takes over. The customer is placed in an audience, and the brand bids for attention it once had for free.
Both machines have a clear budget owner and a familiar return-on-investment story. The middle has neither. It is the long period in which the customer may have no reason to transact: the first-time investor waiting for the next SIP debit, the fashion shopper waiting for the next season, the traveller who will not fly again for months, the homeowner who will not need paint for years. In that period, the customer who is less attentive typically receives the same promotional stream as everyone else — often part of why they became less attentive — and the customer who has gone quiet receives more of it, until list-hygiene rules suppress them.
Figure 3. Conceptual, not measured: heavy effort around the transaction and after the loss, thin exactly where the slide is decided.
The two established systems explain the gap. Loyalty rewards the transaction. Points are earned by spending and matter at the moment of purchase; between purchases they sit on a balance the customer rarely checks. CRM manages the customer. It records what the company knows, organises journeys and triggers, and sends accordingly. Both are valuable. Neither was designed primarily to make the relationship worth opening when there is nothing to buy.
More promotional frequency is not the answer. An offer asks for attention; it does not earn it. If a customer has no buying need, sending more things to buy can accelerate the very decay the brand is trying to prevent. The middle needs a different contract: be useful even when you are not selling.
Why has the gap persisted? Partly because organisations are built around what is easy to count. Campaign teams own sends. Loyalty teams own points. Performance teams own paid conversion. CRM teams own journeys. Nobody owns the slow erosion of voluntary attention across months. Budgeting makes it worse: a win-back campaign has a visible target list and a paid campaign has a visible acquisition cost, while a programme whose job is to keep a customer from becoming expensive later asks the organisation to spend before the loss is booked. That is why attention retention needs both a metric and an economic bridge. Without the first it looks soft; without the second it looks like another content initiative.
Loyalty rewards the transaction. CRM manages the customer. Nothing yet retains the attention.
5
What Retaining Attention Takes
Win-back begins after attention has gone. Attention retention begins before.
If attention retention is a real job, it needs four things.
- Usefulness on days the brand has nothing to sell. A contact has to earn its place in the inbox without depending on urgency, discount or a launch: something the customer is glad to know, from their world rather than the catalogue.
- A rhythm the customer chooses. Frequency should belong to the reader. The brand can produce every day; the customer decides whether they want one, two, three or seven editions a week. Volume is valuable only when it is requested.
- Memory across contacts. Most brand email starts from zero every time. A retention system should remember what the customer saw, did and kept, so that the next contact feels like a continuation rather than another campaign — and the twentieth is worth more than the first.
- A concurrent holdout. Attention moves with seasons, offers, news and category demand. A brand that wants to claim an intervention slowed the slide needs comparable customers receiving its current best effort at the same time.
One answer is the Brand Digest described in the previous two essays in this series, Points Reward Purchases. Cards Reward Attention. and Between Purchases: Four Brand Stories. It is a small publication from the customer’s world, sent under the brand’s name: a few things worth reading, one small thing to do, and something persistent to keep — Cards that accumulate into Sets, each opened after a quick prediction. The reader controls frequency, and each edition recalls the last. It creates two reasons to return: the publication earns today’s open, and the accumulating collection creates a reason for the next one.
But the Brand Digest is not the thesis. The broader claim is that the fade needs a product and a measurement system of its own. Another company could invent a different format and test the same idea. And the decisive question is not whether open rates rose. It is whether fewer customers moved from engaged to quiet than would have done under the brand’s current best effort.
The publication also has limits. Customers who have already gone quiet or dormant are not its audience; mailing a dormant list twice a week damages the deliverability everything depends on. They need a different intervention.
6
Retain, Recover, Reacquire
Brand Digest slows the flow into dormancy. Progency Recover reverses it. Adtech reacquires after both fail.
Once the slide is visible, the customer base can be defended in three lines, in rising order of cost.
Retain is for customers who are still engaged, or beginning to fade. The job is to preserve voluntary attention before silence sets in. The Brand Digest is one mechanism; a brand can buy it as a product, or have it run as part of a Progency mandate.
Recover is for customers who have gone quiet or dormant. The job is to restore the relationship through the brand’s own identity and channels — email, messaging, the app — before paying an outside platform to find them. Progency Recover does this as a done-for-you service, priced on the customers and revenue it brings back.
Reacquire is what happens when the first two fail, or were never tried. The relationship is effectively lost, and the brand enters an auction — search, social, marketplaces — to reach a customer it once had.
Figure 4. Three lines of defence. Cost per customer, and loss of control, tend to rise as the relationship weakens. Bar lengths are illustrative.
Every customer Retain saves never enters Recover. Every customer Recover saves never enters Reacquire. The value compounds as the flow downstream shrinks.
The three lines also need different operating rhythms. Retain is continuous and light-touch; it should feel like a useful habit, not a campaign. Recover is episodic and targeted; it activates when behaviour crosses a threshold, and it can use stronger interventions because the relationship is already weak. Reacquire is selective and external; it should be reserved for cases where paying the auction is worth it. Treating all three as one undifferentiated “retention” budget hides the economics. Treating them as sequential defences makes the trade-off visible.
An honesty clause belongs here. This is not an argument that paid acquisition is bad; brands need new customers, and paid media can work. Nor is it an argument that every dormant customer could have been saved. People move city, change life stage, switch brands or stop needing the category. Some loss is structural. The target is avoidable reacquisition: customers the brand once had, lost for want of attention, and is now paying to find again.
7
The Budget Case
This is not new revenue. It is spend the brand stops paying downstream.
This is where the argument moves from the CMO to the CFO. The usual objection to anything upstream is that it adds cost without adding revenue. The reply is that revenue is the wrong comparison. Attention retention is justified by what it removes from further down the slide.
The wrong way to evaluate it is against the cost of sending an ordinary email, which treats it as one more messaging expense. The right comparison is with what happens when attention is lost. Every brand can size that from its own numbers, and should, rather than from anybody’s industry average:
- The dormant share of the base: how many customers the brand has acquired and no longer hears from.
- Reactivation spend: what it currently spends trying to bring lapsed customers back.
- REACQ%: the share of paid acquisition that is buying back former customers rather than finding new ones.
- The cost of recovering or reacquiring one of those customers.
The strongest business case will be the brand’s own waterfall. Start with the historical base. Trace how many customers moved from Active to Quiet, from Quiet to Inactive and from Inactive to Dormant. Find which of them later appeared in reactivation programmes or in paid-acquisition audiences. Attach the brand’s own recovery and reacquisition costs. Then ask what would happen if the flow into each downstream stage fell modestly.
The exercise has a second benefit: it forces the company to separate acquisition from reacquisition. A paid-media dashboard may report a “new conversion” when the identity belongs to a customer the brand acquired years earlier. REACQ% exposes that circularity.
Framed this way, retention needs to keep only a small share of the base out of the Reacquire line to pay for itself, before counting any revenue those customers go on to produce. This is the practical meaning of Never Pay Twice: do not pay an auction for a customer you could have kept, or recovered through a relationship you already own. The budget decision follows. Move some money from after the loss to before the loss — not all of it, and not on faith, but on evidence tested against a holdout.
8
How You Will Know
Attention can be tested in a quarter. Its economic effect takes a year.
The framework makes two claims, and they run on two clocks.
Figure 5. Two clocks. Attention is read at ninety days; the economics at six to twelve months.
The first claim is about attention: does retaining it slow the fade? Within ninety days, against a concurrent randomised holdout receiving the brand’s current best effort, measure Real Reach, CRR, Habit Rate and Engaged → Quiet migration. The question is not whether the treatment produced impressive engagement in isolation. It is whether attention decayed more slowly than it did for comparable customers at the same time.
The second claim is about economics: does a slower fade reduce loss? Over six to twelve months, read the same cohorts for purchase dormancy, reactivation spend and REACQ%. Did the customers whose attention was retained stay economically active at a higher rate? Did fewer need a win-back programme? Did fewer reappear in paid audiences? The two-horizon design prevents a common mistake: claiming revenue from a short attention experiment. A ninety-day pilot can establish an attention effect. In a low-frequency category, it cannot prove a twelve-month purchase outcome.
Two further readings matter. Look for differences between groups: the effect may be strongest among customers who were active but beginning to fade, weaker among those already quiet, and unnecessary among the most engaged. That would be useful, not disappointing; it would show where the intervention belongs. And distinguish a novelty spike from a habit. A strong first edition followed by rapid decay is not success. The curve across weeks matters more than the launch peak.
Throughout, the baseline is concurrent: never last quarter or last year, which differ in too many ways to isolate anything. If a result cannot survive a holdout, it is not a result.
9
What Would Prove This Wrong
The thesis names its own failure conditions.
Four results would change my mind.
- Attention does not lead. In a brand’s own data, declining attention does not reliably precede purchase dormancy: customers reach Dormant without a detectable weakening in owned-channel behaviour first.
- The intervention does not slow the fade. Customers receiving it migrate from Engaged to Quiet no more slowly than the holdout.
- A slower fade does not reduce loss. Attention is retained, but at twelve months purchase dormancy, reactivation spend and REACQ% look the same as the holdout’s.
- The arithmetic does not close. The cost of retaining attention exceeds the recovery or reacquisition cost it avoids.
There may also be categories where the idea matters less. If purchases are very frequent and habitual, the transaction may already supply enough contact. If purchases are very rare and email identity is weak, another channel may serve better. And if the customer has no ongoing reason to care about the brand’s world, no publication can manufacture one. The framework is not a law. It is a hypothesis with a measurement plan.
10
From Consuming Attention to Retaining It
For years, customer retention has been discussed as if the decisive moment arrives when the customer does not buy again. That is too late. The more useful moment comes earlier: when attention begins to thin, when the next email becomes easier to ignore, when a once-familiar brand starts to recede from memory. At that moment the relationship is still owned, permission still exists, and intervention is cheapest.
Marketing has mostly treated attention as something to consume: bought, spent on an offer, and bought again. The shift this essay argues for is to treat it as something to retain. The goal is not to eliminate churn, which is impossible. It is to identify the avoidable part sooner, and to compare the cost of keeping attention with what the brand already spends, further down the slide, to buy back what it let slip.
The cheapest dormant customer to reactivate is the one who never became dormant.
Retain before you Recover. Recover before you Reacquire. Never Lose Customers starts with never losing their attention.
Don’t begin customer retention when the customer stops buying. Begin when their attention starts to fade.
Key points
- Dormancy is an outcome, not an event: it is recorded at the end of a slide that often begins months earlier. Dormancy is a fire alarm; attention decay is a smoke detector.
- Five stages — active, less attentive, quiet, inactive, dormant — defined by rules each brand calibrates to its own cadence, purchase cycle and each customer’s own rhythm. Watch the flows, not just the counts.
- Customer and revenue retention are lagging measures. Attention retention is rarely measured, observable far more often, and — as a hypothesis each brand should test — moves first.
- Marketing spends heavily around the transaction and after the loss. The middle has no owner, no budget line and no metric. An offer asks for attention; it does not earn it.
- Retain, Recover, Reacquire: Brand Digest slows the flow into dormancy, Progency Recover reverses it, adtech reacquires after both fail — each with its own operating rhythm.
- The case is avoided spend, built from the brand’s own waterfall, dormant share, reactivation spend and REACQ%. Move some money from after the loss to before it.
- Attention is tested in ninety days and the economics over six to twelve months, both against a concurrent holdout, watching for differences between groups and for novelty spikes.
