Why stages matter
A customer three days into their relationship with an operator and one three years into it need entirely different things. The first does not know how the product works. The second knows it better than most staff.
Applying the same programme to both is a common structural error in CRM, and it produces a function that is mildly useful to established customers and actively unhelpful to new ones.
This lesson works through the stages and what each requires.
New: the first session and first week
Usually the steepest part of the churn curve, and the period where CRM has the most influence and typically applies the least thought.
What actually happens to a new customer at many operators: they complete registration, deposit, land on a homepage designed for returning users, face a catalogue of thousands of games with no basis for choosing, and receive a welcome email containing further offers.
What they need is different.
Orientation, not promotion. Understanding what is here and how to find something they might enjoy. A customer who does not find anything in their first session has no reason to return, and no offer compensates for that.
A narrowed choice. Presenting a curated selection appropriate for someone new is more useful than the full catalogue. Popularity is a reasonable proxy in the absence of any personal data.
Clarity about the bonus. If they took a welcome offer, they should be able to see the wagering requirement, their progress against it and which games contribute. Burying this invites complaints: in a Gambling Commission study published in 2021, misleading promotions and adverts were the most common reason gamblers gave for complaining, cited by 16% of those who had complained or wanted to, ahead of incorrect bet settlement and non-payment of winnings. In Britain the terms themselves are now constrained: since 19 January 2026 bonus wagering requirements have been capped at ten times and incentives may no longer mix gambling products, such as betting and slots.
Early tool introduction. Offering limit-setting during onboarding reaches customers before any pattern has formed. In Britain it is no longer optional: since 31 October 2025 the Commission's technical standards have required operators to prompt customers to set a financial limit during registration or at their first deposit, with setting a limit presented as the default choice. This is a genuine protective measure and it also signals something about the operator.
Learning about them. Early treatment should be designed to gather information: which products they try, what they respond to, when they play. Segmentation on a new customer is guesswork, and the correct response to low information is to reduce the uncertainty.
The commercial case for investing here is that early retention compounds. An operator can test this in its own cohort data: where customers retained through the first month are much more likely to persist for a year, an improvement in early retention is worth considerably more than the same improvement applied later.
Establishing: weeks two to twelve
The habit-forming period, where a customer either settles into a pattern or drifts away.
The programme here should be learning and adjusting. What products have they engaged with? Which communications do they respond to? What is their natural frequency?
Offers become more appropriate at this stage and should be calibrated to observed behaviour rather than applied uniformly. A customer who plays slots twice a week does not need a sportsbook free bet.
The first significant negative experience frequently occurs here: a failed deposit, a slow withdrawal, a verification request, a bonus term they did not expect. How that is handled substantially determines whether they persist, and CRM should know when it has happened rather than continuing a scheduled sequence regardless.
This is also the stage at which early risk indicators may appear. A customer whose deposits escalate rapidly in their first weeks is displaying something worth noticing, and early-tenure customers can slip through monitoring designed around established behaviour, which is a gap worth closing. In Britain it is also a licence requirement: operators must monitor customer activity to identify harm from the point when an account is opened, and the Commission's guidance asks for a consistent level of protection for new customers.
Established
The mature relationship, where most of an operator's revenue typically sits.
The requirements shift from establishing a habit to maintaining an experience.
Relevance. Communications and offers matched to observed preference rather than to a general calendar.
Variety without disruption. Introducing new content the customer might enjoy, which requires understanding their preferences well enough to extend rather than to interrupt.
Recognition. Acknowledging the relationship, which loyalty structures formalise and which does not require them.
Restraint. Established customers are the group most likely to be over-contacted, because they are the largest group with known preferences and the easiest to build campaigns for. Contact frequency should be governed rather than maximised.
Monitoring. The established base is where the majority of high-value customers sit, and therefore where protective monitoring matters most.
Changing: the stage most operators do not have
A customer whose behaviour shifts materially warrants attention, and this is not a stage most CRM programmes recognise.
Increased activity is treated as unambiguously positive at most operators, and it is not. Rapid escalation in deposit frequency or size, extended sessions, play at unusual hours and repeated deposits within a session are recognised markers of developing harm: the Gambling Commission's guidance for remote operators lists escalating deposits, increasing session length, binges and late-night play among the indicators operators must monitor. A CRM function that responds to escalation with increased promotion may be accelerating the pattern that should have triggered a review.
The correct response is that escalation triggers a look rather than a campaign. Where the pattern is consistent with enthusiasm, nothing further is needed. Where it matches the indicators described in the Customer Service course, it goes to safer gambling.
Decreased activity is treated as a churn risk and is frequently something else. A customer reducing their play may have lost interest, may have found a competitor, or may be deliberately exercising control over something that was becoming a problem.
The third case is the one that matters, and the systems should be able to identify it. A customer who reduced activity after setting a limit, after a time-out, or following an intervention should not receive reactivation contact, and that suppression should be automatic rather than depending on someone checking.
Lapsing and dormancy
Defining when a customer has gone requires care, because a fixed threshold applied to everyone produces false positives.
A customer who plays every day and has not appeared for three weeks has probably gone. One who plays only during a particular sporting season and has not appeared for three weeks in June has not.
Dormancy definitions should be relative to the individual's established pattern, which requires the pattern to be recorded. A fixed threshold is the simple option: the Belgian operator described in a 2022 University of Liège paper on churn prediction classed a customer as churned after 35 days without activity, applying different retention rules in the first 60 days after registration. Operators applying a uniform threshold of that kind send winback campaigns to seasonal customers who are behaving entirely normally, which is wasteful and mildly insulting.
The lapsing stage, where activity is declining but has not stopped, is where intervention is most likely to work, and it is frequently missed because reporting focuses on the binary of active or lapsed rather than on the trend.
Returning
A stage designed for almost nowhere, and one with specific requirements.
A customer coming back after months arrives with partial memory of the product, possibly expired payment details, potentially outdated verification, and preferences that may have changed.
What helps: reorienting them to what they played before and what has changed; surfacing practical problems such as expired cards proactively rather than at the point of deposit failure; offering the tools, since a return after absence is a natural decision point at which to offer limit-setting; and not assuming continuity, since their previous pattern may not describe them now, which matters for both recommendation and for risk monitoring.
And the protective consideration: why were they away. This cannot always be known and can sometimes be inferred, and a customer whose absence followed a protective action should be handled accordingly rather than welcomed back with an offer.
Designing the programme
Practically, a lifecycle programme is a set of triggered sequences rather than a campaign calendar.
Triggers fire on events: first deposit, first week complete, first withdrawal, product first tried, inactivity threshold reached, behaviour change detected, return after absence.
Sequences deliver appropriate content for that moment, with logic for what happens if the customer responds and if they do not.
Suppression applies throughout, checking protective classifications before anything sends.
Frequency governance limits total contact regardless of how many sequences a customer qualifies for, which prevents the accumulation that produces over-communication.
Measurement compares customers who entered a sequence with comparable customers who did not, which requires holdout groups and is the subject of the final lesson in this course.
The general principle is that a lifecycle programme responds to what a customer is doing rather than to what the calendar says, and an operator making this shift should expect contact volume to fall, and should measure whether effectiveness rises with it.
Contact frequency and governance
A practical mechanism that sits across all stages and prevents the most common failure.
A customer may simultaneously qualify for a welcome sequence, a product introduction, a weekly offer, a sporting event campaign and a milestone recognition. Without governance, they receive all of them, and the accumulation is what produces the over-communication described in the first lesson.
Frequency caps limit total contact per customer per period, regardless of how many campaigns select them.
Priority rules determine which communication sends when several qualify, so that the most valuable one is not displaced by a routine campaign.
Channel governance applies separately, since a customer tolerant of weekly email may find daily push notifications intrusive.
Quiet periods prevent contact at hours that would be unwelcome, and in this sector there is a further consideration: promotional contact delivered late at night reaches a population that skews towards the customers an operator should be most careful with, since the Commission's guidance notes that a higher percentage of overnight gamblers were found to be problem gamblers than at other times of day.
Recovery periods after a customer has been contacted intensively, so that campaign clusters do not repeat.
The governance also needs a protective layer: a customer who has been the subject of an intervention, set a limit or displayed indicators should have promotional contact suppressed entirely, and that suppression must survive every campaign selection. In Britain the floor is a licence condition: operators must prevent marketing and the take-up of new bonus offers where strong indicators of harm have been identified.
Frequency governance reduces total send volume by design; if response rates then rise and opt-out rates fall, that is the signature of a base that was being over-contacted.
Measuring lifecycle programmes
A note before the dedicated measurement lesson, because lifecycle work is measured badly in a characteristic way.
The temptation is to measure whether customers who entered a sequence performed better than those who did not. That comparison is confounded, because sequence entry is triggered by behaviour and the behaviour predicts the outcome. Customers who entered a welcome sequence deposited, which is why they entered it, and comparing them to customers who did not deposit measures the trigger rather than the sequence.
The correct approach is a holdout: a randomly selected proportion of qualifying customers who receive nothing, compared with those who receive the programme. That is the only way to establish what the programme caused.
Holdouts are resisted because they mean deliberately not sending to some customers, which feels like forgone revenue. The counter is that without them the operator does not know whether the programme generates revenue at all, and a first holdout test can show that a specific programme generates little or nothing while costing a great deal in bonus value.
The practical recommendation is a permanent small holdout on every significant lifecycle programme, reviewed periodically. The cost is the forgone response from a small share of qualifying customers, and the information is the most reliable evidence the function produces.
A worked lifecycle programme
To make this concrete, a programme for a casino-led operator, described in outline.
Trigger: first deposit. Immediate confirmation, orientation content explaining how to find games, clear presentation of any bonus and its requirements, and an introduction to limit-setting. No further offers.
Trigger: forty-eight hours, no play. A single prompt with a curated selection of popular titles. If no response, no further contact for a week.
Trigger: first week complete, active. Content based on what they actually played. Introduction to a second product area only if their behaviour suggests interest.
Trigger: first withdrawal requested. Nothing promotional. Status communication only, since this is the moment trust is established or lost and an offer at this point reads as an attempt to prevent the withdrawal.
Trigger: bonus wagering complete. Confirmation and a clear statement of what is now withdrawable, which prevents a substantial complaint category.
Trigger: thirty days active. Recognition, preference-based content, and a limit-setting prompt if none is set.
Trigger: deposit escalation beyond established pattern. No campaign. A flag to safer gambling for review.
Trigger: activity decline relative to their own pattern. A single relevance-based contact. If the decline followed a limit being set, a time-out, or an intervention, suppression instead.
Trigger: dormancy relative to their pattern. A winback sequence subject to the exclusions above, limited in length, with no escalation of offer value if unanswered.
Trigger: return after absence. Reorientation, practical checks on payment and verification status, tool offer, and no assumption that previous preferences still hold.
Throughout: frequency governance capping total contact, protective suppression checked before every send, and a holdout on each sequence.
The notable features are how much of it is not promotional, how many triggers produce no contact at all, and how many produce a review rather than a campaign. That is the shape of a lifecycle programme designed around what customers need rather than around opportunities to communicate.
Where lifecycle work goes wrong
A short catalogue of the recurring failures, drawn from the stages above.
No distinct new-customer experience. A common gap, and a costly one, since it sits at the steepest part of the churn curve.
Offers instead of orientation for people who do not yet know what the product is.
Bonus state invisible, feeding one of the most common causes of complaint in the sector.
Escalation treated as success, which is the failure with the most serious consequences.
Decline treated uniformly as churn risk, which means reactivation contact reaches people who reduced their play deliberately.
Fixed dormancy thresholds, producing winback campaigns aimed at seasonal customers behaving normally.
No return experience, so customers coming back after months are treated as though they never left.
Sequences that continue regardless, so a customer having a bad experience with a failed withdrawal receives their scheduled promotional email.
No frequency governance, so customers qualifying for several programmes receive all of them.
No holdouts, so the operator cannot establish whether any of it works.
Each of these is addressable and none requires sophisticated capability. Most require someone to look at the programme from the customer's perspective and ask whether what they receive makes sense given what is happening to them, which is an afternoon's work and is rarely done.
Adapting stages to the vertical
A final refinement, since lifecycle patterns differ meaningfully between products.
Sportsbook customers are frequently seasonal and event-driven. Their natural rhythm follows competitions, and a dormancy definition that ignores this will misclassify a large proportion of them. Their onboarding is also different: someone who joined to bet on a specific event may have no intention of becoming a regular customer, and treating them as a failed retention case misreads what happened.
Casino customers have more continuous patterns and shorter natural cycles, which makes activity decline a more meaningful signal and dormancy thresholds shorter.
Poker customers build habits around specific games, stakes and times, and their lifecycle is heavily influenced by the liquidity available at their preferred tables. A poker customer lapsing may be responding to the room rather than to anything CRM controls.
Bingo customers frequently form attachments to the community aspect, which means their lifecycle involves social factors that other verticals lack, and interventions that treat bingo as a slot lobby with a different game type reliably damage what made it work.
Multi-product customers are often among the most valuable and are the most complex to stage, since their activity in one vertical may be steady while another declines. They also need care: the Gambling Commission cites evidence that consumers are more at risk of harm when they gamble on multiple products, and its guidance lists gambling on multiple products among the indicators of harm.
The practical implication is that lifecycle definitions should be built per vertical rather than applied uniformly, and that operators serving several products with a single lifecycle model are misclassifying a substantial share of their base. That misclassification then propagates into every campaign selection built on top of it.