A dormant customer is one who has bought before but has not purchased, or in some systems engaged, within a period the business considers normal. The term belongs to customer relationship management (CRM), the practice and software used to record a company's dealings with each buyer. It exists because most retail relationships end without notice. Nobody cancels a membership of a shoe shop; a buyer simply stops appearing. Dormancy is the assumption a marketer makes about that silence, and it decides whether a person receives a reactivation offer, is excluded from paid campaigns or eventually has their data deleted.

The pool is large. Between 20% and 60% of customers become inactive in any given year, according to a 2002 paper by Dominique Crié in the journal Interactive Marketing, many of them recoverable rather than lost.

How a customer becomes dormant

Dormancy is calculated, not observed. A CRM system, an email service provider (ESP) or a customer data platform (CDP) applies a rule to the gap since each customer's last transaction. The usual inputs are the three variables of RFM analysis: recency, the time since the last purchase; frequency, how often the customer buys; and monetary value, how much they spend. Recency triggers the label, but frequency sets the yardstick. A weekly buyer missing for two months has strayed further from habit than an annual buyer missing for ten. The gap between orders, the purchase cycle, governs lookback windows and ad timing.

Thresholds come from the business or from its software vendor. Klaviyo, an ESP, treats a purchase within four months as recent and groups buyers whose last order came four to 13 months ago into its lapsed personas, according to its help centre. Voyado, a Swedish CRM vendor, reserves its dormant stage for customers with the lowest recency score in its RFM grid. Email platforms often measure engagement instead. HubSpot marks a contact as unengaged after 11 consecutive ignored marketing emails if it has never engaged, or 16 if it engaged before, counting an open or a click as engagement, according to its knowledge base. HubSpot suppresses them by default.

A more formal method estimates the probability that a customer is still active. Probabilistic customer-base modelsassume each person buys at a personal rate until an unobserved moment of dropout; the beta-geometric/negative binomial distribution (BG/NBD) model, for instance, gives each customer a set probability of becoming inactive after every purchase. The output is a score rather than a flag, and a long silence lowers it faster for a frequent buyer than for an occasional one.

Once flagged, dormant records move. CRM teams send them to email tools for reactivation sequences, or upload them to advertising platforms as hashed lists for targeting or exclusion. In Google Ads the upload route is Customer Match, which has capped list membership at 540 days since April 7, 2025, set through the membership_life_span field in the Google Ads API and membershipDurationDays in Display & Video 360 (DV360). The clock runs from the upload or refresh date rather than the last purchase, so a re-uploaded member counts as current however long ago they bought. As of October 2026, a list needs at least 100 members added or updated within 540 days to stay eligible, according to Google's help centre.

Origin and evolution

Dormancy began as a mail-order cost problem: catalogue companies paid to post every book, so dropping names from the list saved money. RFM analysis came out of that trade and was probably the first predictive model used in database marketing, according to Robert Blattberg, Byung-Do Kim and Scott Neslin in their 2008 textbook Database Marketing.

Academics then gave inactivity a statistical form. Andrew Ehrenberg showed in 1959 that purchases follow the negative binomial distribution (NBD). In January 1987, David Schmittlein, Donald Morrison and Richard Colombo published the Pareto/NBD model in Management Science, estimating from each customer's frequency and recency whether that customer was still active. Because it proved hard to compute, Peter Fader, Bruce Hardie and Ka Lok Lee proposed BG/NBD in Marketing Science in 2005, testing it on 2,357 customers of the online music retailer CDNOW. Crié had argued in 2002 for a more rigorous test than arbitrary time cut-offs.

Platforms and regulators later imposed clocks of their own. Amazon introduced new-to-brand metrics in January 2019, counting a purchase as new if the buyer had not bought the brand on Amazon over the prior year. Apple's Mail Privacy Protection, released with iOS 15 on September 20, 2021, fetched email content automatically and severed the link between an open and a person reading. That month France's data protection authority, the CNIL, adopted a framework letting businesses keep customer data for prospecting until three years after the relationship ends. Google added Win-Back Mode and High Value Win-Back Mode, both in beta, to Performance Max in April 2025.

How advertising systems price dormancy

The label now changes what a customer is worth in an auction. Google's retention goal asks advertisers to upload a Customer Match list of lapsed customers and give them a value that should usually sit between those of new and existing customers, according to Google's help centre. Google does not define lapsed; the advertiser's list does, under an API audience category named DISENGAGED_CUSTOMERS.

Definitions also decide what counts as acquisition. Google's autodetection of existing customers draws on up to 540 days of campaign activity, according to its help centre, so older purchases fall outside it. Amazon's 12-month window means a buyer dormant for 13 months returns as new to the brand. Either way, a reactivated customer can be reported as newly acquired.

Data protection law adds another reason. The storage limitation principle in Article 5(1)(e) of the General Data Protection Regulation (GDPR) governs how long identifiable data may be kept. For prospects, the CNIL's three years run from collection or the prospect's last contact, and a click in a marketing email counts as contact, restarting the clock. Google Analytics wrote the same logic into its data retention controls in 2018: with "reset on new activity" enabled, only users who stop generating events see their data expire.

Deliverability adds a commercial penalty. Since February 1, 2024, Gmail has required bulk senders to keep spam rates below 0.3% as reported in Postmaster Tools. Google said in May 2023 that personal accounts unused for two years could be deleted from December that year, and deleted Gmail addresses hard bounce. Some abandoned mailboxes are later switched back on as recycled spam traps that identify senders still mailing them.

Limitations and disputes

There is no common standard. The same person can sit in a lapsed segment in Klaviyo after four months, stay in Customer Match for 540 days after upload, count as new to the brand at Amazon after a year and remain a lawful prospect in France for three years. In non-contractual retail, a lapsed buyer and a slow one look identical, so every dormancy flag is an inference.

The signals have also degraded. Apple accounted for about 62% of tracked opens in July 2026, and because HubSpot's default counts an open as engagement, automated prefetching can keep a dormant contact looking active. Sources disagree on how long abandoned addresses wait before becoming traps. Spamhaus describes them hard-bouncing for 12 months or more, Cordial gives a range of 90 days to over a year, and Adobe has said an address can be recycled within 30 days.

The economics of reactivation are disputed. A Boston Consulting Group estimate that winning a customer costs five times as much as keeping one featured in coverage of an Amazon Marketing Cloud audience tool in May 2024. Subscription data is less encouraging. RevenueCat's 2026 report found that 95% of annual app subscribers who cancel never return, with annual reactivation of 4.4% to 5.9%, while about 20% of monthly subscribers come back within a year. Those subscribers cancelled rather than drifted, and RevenueCat's sample covers only apps using its software.

Not the same as

Churn is the rate at which customers stop paying over a defined period, observed directly in contractual businesses when a subscription ends. Dormancy is inferred where no contract exists. Vendors apply "lapsed" and "at risk" inconsistently along the same continuum.

Unengaged subscriber describes email behaviour, not purchasing; a contact can ignore 16 newsletters and still buy. YouTube applies similar logic: since April 2026 it stops lock-screen push notifications to subscribers who have not watched a channel for about a month and ignored recent notifications.

Low activity in Google Ads describes campaign objects rather than people. The platform began pausing ad groups older than 13 months with no impressions in that period from March 11, 2024 and applied the same rule to search keywords from June 2024.

Dormant account in banking is a legal status. Under the UK's Dormant Bank and Building Society Accounts Act 2008, balances with no customer-initiated transactions for 15 years can pass to a reclaim fund, and the owner keeps a right to repayment.

Recent developments

Google has rebuilt the pipes that carry dormant segments. Since April 1, 2026, Customer Match uploads have had to use the Data Manager API, and integrations idle for 180 days fail if reactivated without migrating. Google Ads API v25, released on July 22, 2026, removed the CustomerLifecycleGoal and CampaignLifecycleGoal resources with no transition window, shifting retention and win-back settings to a unified goals schema. Google now applies a single 100-user minimum across Search, Display and YouTube, which keeps smaller reactivation lists usable. As of October 2026, the definition itself still belongs to the advertiser, and Google, Amazon, Klaviyo and HubSpot each run a different clock.

Timeline

  • 1959: Andrew Ehrenberg shows that repeat purchases follow the negative binomial distribution
  • January 1987: David Schmittlein, Donald Morrison and Richard Colombo publish the Pareto/NBD model in Management Science, estimating whether each customer is still active
  • 2002: Dominique Crié publishes research on when a customer should be defined as lapsed, estimating that 20% to 60% of customers become inactive each year
  • 2005: Peter Fader, Bruce Hardie and Ka Lok Lee publish the BG/NBD model in Marketing Science
  • November 26, 2008: The UK's Dormant Bank and Building Society Accounts Act receives Royal Assent, defining a dormant account as one with no customer-initiated transactions for 15 years
  • May 25, 2018: Google Analytics data retention controls, including "reset on new activity", take effect alongside the GDPR
  • January 2019: Amazon introduces new-to-brand metrics with a 12-month lookback
  • September 20, 2021: Apple ships Mail Privacy Protection with iOS 15
  • September 23, 2021: The CNIL adopts its reference framework for commercial activities, setting a three-year retention period for prospecting data
  • May 2023: Google announces that personal accounts inactive for two years may be deleted from December 2023
  • February 1, 2024: Gmail's bulk sender requirements, including a spam rate below 0.3%, take effect
  • February 12, 2025: Google announces a 540-day maximum membership duration for Customer Match lists
  • April 7, 2025: The Customer Match 540-day limit takes effect, applied retroactively to existing lists
  • April 2025: Google adds Win-Back Mode and High Value Win-Back Mode in beta for Performance Max
  • April 1, 2026: Customer Match uploads must use the Data Manager API
  • April 21, 2026: YouTube announces it will stop push notifications to inactive subscribers
  • May 2026: RevenueCat reports that 95% of annual app subscribers who cancel never return
  • July 22, 2026: Google Ads API v25 removes the CustomerLifecycleGoal and CampaignLifecycleGoal resources

Summary

Who. CRM, retention and email teams define dormancy, usually through software from vendors such as Klaviyo, HubSpot and Voyado. Advertising platforms including Google and Amazon receive dormant segments as uploaded lists and apply their own lookback windows, while data protection authorities such as France's CNIL limit how long dormant records may be kept.

What. A status assigned to a past customer or contact who has not purchased or engaged within a defined period. It is inferred from recency, frequency and value data or estimated with probabilistic models, then used to trigger reactivation, exclusion from campaigns or deletion.

When. The practice dates to mail-order catalogue marketing. Statistical models followed in 1987 and 2005, and platform rules arrived with Amazon's new-to-brand metrics in 2019, Apple's Mail Privacy Protection in 2021, Google's 540-day Customer Match cap in April 2025 and its win-back bidding modes the same month.

Where. Inside CRM systems, email service providers and customer data platforms, and downstream in Google Ads, DV360, Amazon Ads and other platforms that accept customer lists for targeting, exclusion and lifecycle bidding.

Why. Most customers leave without saying so, and treating a lost buyer as active, or an active one as lost, wastes budget and distorts acquisition metrics. Retention law, deliverability rules and platform expiry dates now force businesses to decide when silence means departure, with no common standard for where that line sits.