Churn is a symptom
Most CRM functions treat churn as their problem to solve, and most churn has causes they do not control.
Customers leave because deposits failed, because a withdrawal took a week, because verification was frustrating, because a bonus term surprised them, because support was unhelpful, because the product was slow, because they lost interest, because they found a competitor, or because they decided to stop gambling.
Of that list, CRM influences perhaps two items directly. The rest sit in product, payments, verification, support and the customer's own life.
A CRM function responding to churn with retention campaigns is treating a symptom at recurring cost. A CRM function that diagnoses the causes, quantifies them, and puts the evidence in front of the teams that own them is doing something more valuable and considerably cheaper.
Diagnosing causes
The evidence is available and is rarely assembled.
Behavioural sequences before churn. What happened in the sessions preceding the departure? A failed deposit, a withdrawal request, a support contact, a large loss, a bonus dispute? Patterns emerge quickly when the data is examined, and they point at causes.
Support contact. Customers who contacted support before leaving told someone why they were unhappy. The transcripts are the most direct evidence available and are almost never read by CRM.
Complaint themes. Similar, at higher intensity.
Exit surveys, where they can be conducted, though response rates are low and the responses skew.
Cohort comparison. Customers acquired through different sources, in different markets or during different periods churn at different rates, and the differences are informative.
Product event correlation. Churn spikes following a release, a terms change or a process change indicate what caused them.
Failed action data. Customers whose last recorded action was a failed deposit, a rejected document or an abandoned withdrawal did not simply lose interest.
The output should be a quantified list of causes with owners. That is a different deliverable from a retention campaign and produces durable improvement rather than recurring spend.
Intervening at the lapsing stage
The most effective moment, and the one most operators miss because their reporting distinguishes active from lapsed rather than tracking the trend.
A customer whose activity is declining is still engaged, still reachable and has not yet formed the habit of not playing. Intervention here is more likely to work than reactivation afterwards, and it costs less because the customer has not yet gone.
Identifying them requires relative measurement. A customer who played daily and now plays twice a week has declined by more than one who played weekly and now plays fortnightly, in terms of what it signals. Absolute thresholds miss this entirely.
The intervention should be light and relevant rather than promotional. A customer drifting away from a product they used to enjoy may respond to content about what has changed in it. One whose decline followed a specific negative experience may respond to that being acknowledged and resolved.
And the protective check applies: a customer whose decline followed a limit being set, a time-out or an intervention should not receive anything.
Reactivation
Legitimate, commercially worthwhile, and subject to more constraints than most operators apply.
The commercial case is real. A lapsed customer already has an account, has been verified, has a known history and has demonstrated willingness to play. Reactivating them is considerably cheaper than acquiring a new customer, which makes it one of the higher-return activities available.
The constraints are equally real.
Exclusions must be applied before anything sends. Self-excluded customers, obviously: in Britain, licensees must remove a self-excluded customer from marketing databases, or flag them as someone who must not be marketed to, within two days of the self-exclusion notification, and the Commission's code says no marketing should follow when the exclusion expires unless the customer has asked for or agreed to it. Customers subject to restriction. Customers who opted out, or who never opted in to that product and channel: since 1 May 2025 British online operators have had to offer marketing opt-ins per product and per channel. And critically, customers whose absence followed a protective action.
The last category requires systematic identification. A customer who set a restrictive limit and then stopped, took a time-out and did not return, or was the subject of a safer gambling intervention before lapsing, should be permanently excluded from reactivation. Note that a British online time-out can run for up to six weeks, so a customer on a time-out can look exactly like a lapsed one to a fixed-threshold report. This cannot depend on someone checking each campaign; it must be a flag applied automatically and respected by every selection.
Escalation should be limited. A sequence that increases offer value each time a customer does not respond is pursuit. It reaches people who stopped deliberately, and it reaches them with progressively more attractive reasons to reconsider a decision that may have been protective.
Frequency should be bounded. A lapsed customer contacted repeatedly over months has been told the operator does not accept their departure.
Timing should be reasonable. Reactivation attempted immediately after a customer stops is premature; attempted a year later it is contacting someone who has moved on.
The uncomfortable case
An honest treatment requires stating this directly.
Some proportion of lapsed customers stopped because their gambling was becoming a problem. They may have recognised it themselves, may have been told by someone close to them, or may simply have decided to stop. Many of them will not have used any formal tool, which means the operator has no explicit signal.
Reactivation campaigns reach these people, and they reach them with an offer designed to restart the behaviour they decided to end.
This is not a hypothetical concern and it is not addressed by excluding customers with formal flags, because the group in question mostly has none.
What can be done is partial and worth doing.
Exclude on any protective signal, including limits set, time-outs taken, interventions conducted, and support contacts where concern was raised, even where no formal restriction followed. The Gambling Commission already requires British online operators to use the customer's use of gambling management tools as an indicator of harm, and its formal guidance gives changing deposit limits, previous self-exclusions and repeated use of the time-out facility as examples.
Exclude on behavioural indicators present before departure, since a customer who was displaying markers of harm and then stopped is precisely the case in question. For British online operators this is a rule, not a choice: licensees must prevent marketing and the take-up of new bonus offers where strong indicators of harm have been identified.
Limit escalation and frequency, so that a customer who does not respond is left alone rather than pursued.
Do not target customers whose spend before departure was inconsistent with their circumstances, where affordability assessment identified concerns. British online operators have had to run financial vulnerability checks since 30 August 2024, triggered since 28 February 2025 when net deposits exceed £150 in a rolling 30 days, and their results belong in the suppression list.
Make it easy to stop, with a clear opt-out that is honoured immediately. The law points the same way: under the UK GDPR, once a person objects to direct marketing their data may no longer be processed for that purpose.
None of this fully resolves the problem, and pretending otherwise would be false. The honest position is that reactivation carries a risk of reaching people who should be left alone, that the risk can be substantially reduced by exclusions most operators do not apply, and that the residual risk is a reason for restraint in how aggressively the activity is pursued.
Measuring reactivation
The measurement requirements are the same as elsewhere in this course, with one addition.
Holdouts are essential, because a proportion of lapsed customers return without being contacted. A reactivation programme measured on returners without a holdout is claiming credit for people who came back anyway, and the size of that group is unknown until a holdout measures it.
Contribution rather than return. A customer who reactivates, deposits once and lapses again has generated activity rather than value, and reactivation programmes optimised on return rates produce exactly this pattern.
Durability. How long reactivated customers remain active, compared with the base. Reactivation that produces brief reappearances is worth considerably less than its headline figures suggest.
Offer cost against contribution, since reactivation offers are frequently the most generous an operator sends and the customers receiving them are, by definition, the ones who left.
Applying these measures shows whether reactivation is worthwhile at moderate offer values, where it stops being worthwhile, and what share of the customers an operator was claiming to have reactivated would have returned regardless.
What good looks like
To close, the shape of a churn and reactivation programme that works.
The function knows why customers leave, in quantified terms, with causes routed to owners.
It identifies decline early, relative to individual patterns, and intervenes while the customer is still there.
It distinguishes types of absence, and treats seasonal, disengaged and protective departures differently.
It excludes systematically, with protective suppression applied automatically rather than by campaign-level judgement.
It limits pursuit, accepting that customers who do not respond have answered.
It measures with holdouts, on contribution and durability rather than on return rates.
And it spends more effort on the causes than on the campaigns, which is the reallocation that produces durable improvement rather than recurring cost.
Defining churn usefully
The definition determines everything downstream, and most operators use one that misclassifies a large share of their base.
A fixed threshold, such as thirty days of inactivity, is simple and wrong for anyone whose natural pattern is longer than that. A monthly bettor is not lapsed at day thirty-one.
A relative definition, based on a multiple of the customer's own typical gap between sessions, is considerably more accurate and requires the pattern to be recorded.
A probabilistic definition, estimating the likelihood that a customer has gone given their history and current gap, is better still and requires modelling capability.
The practical recommendation for operators without modelling capability is the relative approach, since it requires only transaction dates and eliminates the largest source of misclassification.
Several further distinctions are worth making.
Seasonal customers should have their pattern assessed within their season rather than across the year.
Multi-product customers may be active in one vertical and lapsed in another, which means a single account-level definition loses information.
New customers who never established a pattern require different handling, since there is no baseline against which to assess a gap. A customer who deposited once and disappeared has churned in a different sense from an established customer who stopped.
Never-active registrations, who completed registration and never deposited, are a distinct population and are frequently included in churn figures where they inflate the numbers without belonging.
The value of getting this right is that every downstream decision, from intervention timing to reactivation targeting to retention reporting, depends on knowing who has actually gone.
Retention reporting
A related discipline, since how retention is reported determines what the organisation attends to.
Retention curves by cohort are the fundamental view, showing what proportion of each acquisition cohort remains active at successive intervals. They expose whether retention is improving or deteriorating over time, which aggregate active-player counts conceal completely.
Retention by acquisition source identifies which channels and partners deliver customers who stay, which connects directly to the affiliate quality analysis in the Affiliate Marketing course.
Retention by first-week experience links early events to long-term outcomes, and frequently reveals that customers who encountered a specific problem in their first days retain substantially worse.
Retention by product shows which parts of the offering hold customers.
Revenue retention alongside customer retention, since a base that retains customers while their value declines is deteriorating in a way customer counts do not show.
The reporting failure that recurs is measuring active players as a single number. It rises when acquisition is strong and falls when it is weak, and it says nothing about whether the operator is keeping the customers it acquires. Cohort curves say exactly that, and they are the view worth putting in front of leadership.
The economics of reactivation
A closing commercial assessment, since the activity is frequently justified on intuition.
The case for reactivation rests on the customer already existing. No acquisition cost, no verification cost, a known history and demonstrated willingness. Compared with acquiring a new customer at market rates, reactivating an existing one looks inexpensive.
The case requires several corrections.
A proportion return unprompted. Without a holdout, the programme claims credit for these, and the proportion can be substantial.
Reactivated customers often churn faster. They left once, and many leave again quickly. The durability of reactivated activity is often lower than that of continuously active customers, which reduces the value of each return, and the durability measure above will show whether that holds for a given base.
Reactivation offers are expensive. They are typically the most generous an operator sends, and they go to a population that has demonstrated it was not sufficiently engaged to stay. In Britain their design is also constrained: since 19 January 2026 wagering requirements have been capped at ten times, and an incentive may not include more than one type of gambling product among betting, casino, bingo and lottery.
The cost of the excluded population. Applying the protective exclusions described above removes a meaningful proportion of the target list, and it should.
Netting these, reactivation can remain worthwhile, possibly at lower offer values than are typically used, and it stops being worthwhile above a threshold that is discoverable through testing.
The practical recommendation is to test offer value systematically, starting low, and to establish where the incremental return stops justifying the cost. Where the threshold turns out to be well below what was being offered, the programme was profitable at a fraction of its spend.
Bringing it together
The lesson's argument in summary.
Churn is mostly caused by things CRM does not control, and diagnosing those causes produces durable improvement while campaigning against the symptom produces recurring cost.
Intervention while a customer is declining is more likely to work than reactivation once they have gone, and requires relative rather than absolute measurement to identify.
Reactivation is a legitimate and worthwhile activity, subject to exclusions that most operators apply inadequately, and it must exclude the population whose absence was protective.
That population cannot be fully identified, which is a reason for restraint in how aggressively the activity is pursued rather than a reason to abandon it.
And all of it requires holdouts, without which an operator cannot distinguish the customers it brought back from the ones who were returning anyway.
A note on what customers are entitled to
A closing point that sits slightly outside the commercial frame.
A customer who has stopped playing has made a decision. They may have made it deliberately or drifted into it, and either way the operator's position is that it would prefer they had not.
Pursuing that customer is ordinary commercial practice and is not, in itself, objectionable. Businesses attempt to win back customers in every sector.
What differs here is the possibility that the decision was protective, that the operator cannot reliably tell, and that the consequence of getting it wrong is not a wasted email but a person restarting something they had stopped.
The proportionate response is restraint rather than abandonment. A single well-judged contact to a lapsed customer is reasonable. A sequence of six with escalating offers is pursuit, and pursuit reaches the people who most needed to be left alone.
Operators that have thought about this tend to arrive at similar positions: exclusions applied broadly rather than narrowly, contact limited in frequency and duration, offer escalation prohibited, opt-out honoured immediately and permanently, and a general disposition that a customer who does not respond has answered.
None of that costs much. It removes the tail of the reactivation programme, which as the economics section above suggests was generating the least value anyway, and it removes the cases that would be hardest to defend.