The largest lever and the least measured
Bonus spend comes straight off an operator's gross revenue: DraftKings, for example, reports the incentives it gives through loyalty programmes, free plays, deposit bonuses and rebates as a reduction of the revenue it records. Much of that spend is allocated by CRM decisions about who receives what.
It is also, at most operators, measured badly. Awarded value is reported rather than net cost. Response is reported rather than incremental effect. And offers are allocated by segment and calendar rather than by evidence that they change behaviour.
That combination, a large budget with weak measurement, is where the improvement in most CRM functions actually sits.
The offer types
Deposit match adds bonus funds proportional to a deposit. The standard acquisition and reactivation offer, and the most expensive in awarded value terms.
Free spins grant a number of spins on specified games. Cheaper, easily targeted to casino customers, and useful for promoting specific content.
Free bets grant a stake that returns profit but not the stake itself if it wins. The sportsbook equivalent, and cheaper than its face value implies for that reason.
Cashback returns a proportion of net losses over a period. Popular with customers, straightforward to understand, and structurally worth examining because the reward scales directly with how much a customer lost.
Price boosts enhance odds above the normal price, sacrificing margin on a specific market to generate attention.
Tournaments and leaderboards create competitive structures with prizes, which generate engagement at controlled cost since the prize pool is fixed regardless of participation.
Loyalty points accrue on play and convert into rewards, providing a continuous low-intensity incentive.
Non-monetary rewards including early access, exclusive content and recognition, which cost little and are systematically underused because they are harder to measure.
What a bonus actually costs
The calculation that most operators get wrong.
A £50 deposit match with a thirty times wagering requirement obliges the customer to stake £1,500 before withdrawing anything derived from it. Each of those stakes is exposed to the house edge. At a blended edge of 4%, expected losses on that wagering are £60.
The bonus therefore has an expected net cost that is considerably lower than £50, and depending on the requirement level and the games played, may approach zero or be positive for the operator.
The thirty times figure is illustrative, and in Great Britain it is no longer lawful: since 19 January 2026 the Gambling Commission has capped wagering requirements at ten times. At ten times the customer stakes £500, expected losses at a 4% edge are £20, and the same £50 bonus has an expected net cost of around £30. A cap therefore changes the economics of an offer, not just its presentation.
That is the mechanism wagering requirements exist for, and it means the correct cost figure requires knowing the requirement, the contribution rates, the games actually played and the completion rate.
Several things follow.
Awarded value overstates cost, sometimes by a large factor, and an operator budgeting on awarded value is budgeting for a number that does not correspond to anything.
Different offers have different real costs despite similar face values. A £50 match with a low requirement costs substantially more than one with a high requirement.
Completion rate matters. Bonuses never completed have a different cost profile from those completed, and many bonuses are never completed.
Free bets cost less than face value because the stake is not returned, typically making the real cost a proportion of the nominal amount.
The practical requirement is that bonus reporting shows net cost, calculated from actual play, and that this figure is what appears in campaign assessments and in the bonus cost line the finance function reports.
Incrementality
The second measurement failure, and the more consequential one.
An offer sent to a segment produces revenue from that segment. Some of that revenue would have occurred anyway. The offer's value is the difference, and establishing it requires a comparison group.
Without a holdout, an operator sees offer recipients generating revenue and concludes the offer worked. The customers who would have played regardless are counted as successes, and the promotional value given to them is counted as cost incurred to achieve something that would have happened.
Operators running holdouts for the first time frequently find that specific offers produce little or no incremental effect, particularly offers given to already-engaged customers who were going to play. That is not an argument against bonusing; it is an argument for directing it at customers whose behaviour it actually changes.
The practical implementation is a permanent randomly selected holdout on every significant offer programme, reviewed periodically. The forgone revenue is small and the information is the only reliable basis for allocation decisions.
Constructing terms
Bonus terms exist for good reasons, and promotions sit behind one of the sector's most common complaint categories: in Gambling Commission research published in 2021, misleading promotions and adverts were the most cited reason for complaining, given by 16% of gamblers who had complained or wanted to, ahead of incorrect bet settlement and non-payment of winnings. That is largely a presentation problem rather than a terms problem.
The conditions that recur.
Wagering requirement, the multiple that must be staked.
Contribution rates, since low house edge games contribute partially or not at all. This is a frequent source of customer surprise.
Maximum stake while active, breach of which commonly voids the bonus entirely.
Maximum conversion, capping how much of a bonus-derived balance may be withdrawn.
Time limits on completion.
Eligibility conditions relating to deposit method, market or account history.
Game restrictions, excluding specific titles.
Each is defensible. The problem is where they are disclosed. Terms presented as a linked document that customers accept without reading satisfy a legal requirement and communicate nothing, and the resulting disputes are the predictable consequence.
The practices that reduce this substantially: material conditions at the point of decision, in the same view, in plain language; progress visible during the bonus, showing wagering remaining and which games contribute; warnings before a breach, where a customer is about to do something that will void the bonus; and system prevention rather than retrospective penalty, since blocking a non-compliant stake serves everyone better than allowing it and voiding afterwards.
In Britain part of this is now a licence requirement: licensees must present the significant conditions of any promotion at the point of sale and on its advertising, and advertisers must state significant conditions such as deposit and wagering requirements clearly and prominently under CAP Code rule 8.17. Many bonus disputes were never really disagreements about fairness but the result of conditions the customer did not see.
Abuse
A proportion of promotional value is extracted by customers with no intention of playing genuinely.
Multi-accounting to claim offers repeatedly. Coordinated groups exploiting offers systematically. Hedged positions, where a promotional bet is offset elsewhere to remove risk and lock in the value. Bonus-only play, where activity stops the moment conditions are met.
Detection uses the signals described in the Affiliate Marketing and Payment Operations courses: shared device and payment characteristics, clustered registration timing, activity confined to qualification, and withdrawal immediately on completion.
The more durable remedy is design rather than enforcement. An offer that cannot be hedged profitably will not be hedged. Terms that prevent the exploit are better than terms that permit it and then void the winnings, which generates disputes with legitimate customers who did something similar innocently.
Proportionality matters here. Aggressive abuse detection catches legitimate customers whose behaviour resembles the pattern, and voiding a genuine customer's winnings on suspicion generates a complaint the operator will usually lose in reputational terms even where it is contractually correct.
The protective dimension
The constraints that govern offers more tightly than any other CRM activity.
Offers must not reach customers who are self-excluded, restricted, or displaying risk indicators. This is a suppression requirement and, as established throughout these courses, it is where failures most often occur. In Britain a self-excluded customer must be removed or flagged in marketing databases within two days of the self-exclusion notification.
Offers should not be timed to moments of loss. A response-optimising system will find that these moments produce good response rates, for reasons that make the practice indefensible.
Reactivation offers must exclude customers whose absence was protective, which the next lesson develops.
Escalating offer value to unresponsive customers is a pattern worth examining. A customer who has not responded to three increasingly generous offers is being pursued, and the pursuit reaches some people who stopped for reasons the operator should respect.
Cashback and loss-based rewards warrant particular thought. A reward that scales with how much a customer lost is the standard structure of gambling loyalty, and it is difficult to defend for a customer whose losses are not affordable. The protective override described in the segmentation lesson is doing real work here rather than catching edge cases.
Several jurisdictions restrict bonusing directly: prohibiting certain structures, as Britain does with offers that mix betting, casino, bingo or lottery products; limiting offers to a single welcome bonus, as Sweden's Gambling Act does by allowing a licensee to offer a bonus only on the first occasion a player plays with it; or banning promotional contact with customers showing risk indicators, as Britain does by requiring licensees to prevent marketing and the take-up of new bonus offers where strong indicators of harm have been identified. Market-specific rules apply per recipient rather than per campaign, which is an operational requirement that catches operators applying a single global offer set.
Allocating well
To close, the discipline that distinguishes an effective bonus programme.
Measure net cost, not awarded value.
Measure incrementality, using holdouts, and accept the findings when they are unflattering.
Target on responsiveness, since customers who do not respond to offers should not receive expensive ones.
Vary value by evidence rather than by segment convention, since the amount required to change behaviour differs and defaults to generous.
Test frequency and value, which is among the most valuable analysis available and among the least run.
Retire offers that do not work, which requires having measured them.
Present terms honestly, which reduces disputes and improves the trust that makes the next offer worth reading.
An operator applying these typically finds it can reduce promotional spend materially without losing revenue, because a significant proportion was reaching customers whose behaviour it did not change. That finding is available to anyone willing to run the analysis, and is unavailable to anyone measuring bonus performance by uptake.
A worked bonus calculation
To make the cost mechanism concrete, an illustrative example. It uses a thirty times requirement to show the mechanism clearly; in Great Britain the maximum is now ten times.
An operator offers a £50 bonus on a £50 deposit, with a thirty times wagering requirement on the bonus amount, slots contributing 100% and table games 10%.
The customer must stake £1,500 in qualifying play before withdrawing bonus-derived funds.
Assume the customer plays slots at a 96% return to player, giving a house edge of 4%. Expected losses across £1,500 of staking are £60.
The operator has awarded £50 and expects to recover £60 through the wagering the bonus requires. On expectation, the bonus costs nothing.
That calculation is theoretical and several factors move it.
Completion rates. Many bonuses are never fully wagered, with the customer either exhausting the balance first or abandoning it. Where the balance is exhausted, the operator retained the deposit and the bonus cost nothing. Where it is abandoned with funds remaining, the treatment depends on the terms, although in Britain the Competition and Markets Authority's 2018 guidance tells firms to let customers exit a promotion at any time with their remaining deposit and any winnings from play with it.
Game choice. A customer playing lower edge games generates less expected loss per unit staked, raising the real cost.
Variance. Individual outcomes differ enormously from expectation, and a customer who wins substantially during wagering represents a real cost even though the average holds.
Contribution rates determine how much play counts, and customers who play non-contributing games are staking without progressing.
The practical output is that net bonus cost should be calculated from actual observed play rather than assumed, and that the resulting figure is frequently far below awarded value. Operators reporting awarded value may be budgeting against a number several times the real one, which distorts every allocation decision built on it.
The reverse error also occurs. A low wagering requirement, a high contribution rate on low edge games, or a generous conversion cap can produce a bonus whose real cost approaches its face value, and operators assuming that requirements always neutralise cost are equally wrong.
Loyalty structures
A distinct category worth separate treatment, since it operates continuously rather than campaign by campaign.
A loyalty programme accrues points or status on play and converts them into rewards. Its attractions are that it operates without individual campaign decisions, that it creates a reason to concentrate play with one operator, and that status carries value beyond its monetary worth.
The design considerations.
Accrual rate determines the cost, and is usually expressed as a proportion of stake or of revenue.
Tiering creates status, and the thresholds determine how many customers reach each level and what it costs.
Reward mix, which should include non-monetary elements, since recognition and access cost little and are valued.
Expiry, which limits liability and irritates customers.
Visibility, since a programme customers do not understand generates cost without engagement.
The structural point, made in the segmentation lesson and worth repeating, is that loyalty rewards scale with loss. That is the ordinary logic of loyalty in most industries, where spend correlates with satisfaction, and in gambling spend correlates with loss. Rewarding customers in proportion to what they have lost requires the affordability foundation described in the Customer Service course, and a tier structure that elevates customers on spend without any assessment of whether that spend is sustainable is difficult to defend. In Britain a scheme designated "high value", "VIP" or equivalent must be offered in a manner consistent with the licensing objectives, taking into account the Commission's guidance on high value customer incentives.
Market variation
Bonusing is among the most divergent areas of gambling regulation, and an offer structure lawful in one market may be prohibited in another.
The dimensions that differ.
Whether bonuses are permitted at all beyond a limited welcome offer, since some markets, Sweden among them, restrict ongoing promotional offers to existing customers.
Wagering requirement caps, where some jurisdictions limit how high a requirement may be set: ten times in Great Britain since January 2026.
Presentation requirements, specifying how and where terms must appear.
Prohibited structures: in Britain, for example, an incentive may not mix gambling product types, or increase in value because the qualifying activity is completed in a shorter time.
Restrictions on who may receive offers, particularly customers displaying risk indicators, which several markets have made explicit.
Advertising rules governing how offers may be promoted, separate from how they may be constructed.
Free bet and free spin treatment, which differs in whether the stake is returned and how the value must be described.
The operational requirement is that offer eligibility, construction and presentation are applied per recipient according to their market rather than per campaign. An operator running a single global offer set with translated copy will be non-compliant somewhere, and the failure is entirely foreseeable.
The related requirement is that changes are tracked, since this area moves. An offer structure that was permitted when it was designed may not be a year later, and operators without a process for reviewing live offers against current rules will discover this through a complaint or an enforcement query.
Assessing an offer before it launches
A short checklist that catches most of the problems described in this lesson.
What is the net cost, calculated from expected play rather than from awarded value?
Who is it going to, and would they have played anyway?
Is there a holdout, and what will be compared?
Are the material terms in the message, at the point of decision, in plain language?
Can the system enforce the terms rather than penalising breach retrospectively?
Can it be hedged or otherwise extracted without genuine play?
Is it permitted in every market it will reach, in this construction and this presentation?
Are the suppressions applied, verified rather than assumed?
Does the value escalate to unresponsive recipients, and should it?
Does the reward scale with loss, and if so what affordability foundation supports it?
Ten questions, answerable in a short meeting, and they prevent the majority of the failures that make bonusing expensive and contentious.