Judging partners on player value rather than volume, detecting fraud, and understanding what the channel actually costs.
In this lesson:
- Build a partner assessment framework based on downstream player value rather than registration volume
- Apply cohort analysis to affiliate traffic and interpret what it reveals about partner quality
- Identify the main categories of affiliate fraud and the detection signals for each
- Calculate the true cost of the affiliate channel and compare it against alternative acquisition routes
Volume is the wrong measure
An affiliate programme reporting the number of players each partner delivered is measuring the thing that is easiest to count rather than the thing that matters.
Consider two partners over a quarter. The first delivers 500 first time depositors. The second delivers 200. On volume, the first is clearly superior. Now add that the first partner's players average a single small deposit and are largely inactive within a month, while the second partner's players deposit repeatedly and a meaningful proportion remain active a year later. The second partner is worth considerably more, and a programme managed on volume will have spent the quarter cultivating the wrong relationship.
This lesson is about measuring what actually matters, and about the analytical work that separates a programme from a payment mechanism.
The quality framework
Assessing a partner properly means following its players downstream through several stages.
Conversion quality covers the funnel from click to depositing player: click-to-registration rate, registration-to-deposit rate, and the time between them. Genuine high-intent traffic converts consistently at each stage, and unusual ratios are informative in both directions.
Deposit behaviour covers first deposit size, second deposit rate and deposit frequency in the early period. Second deposit rate is one of the most predictive early indicators available, because a player who deposits twice has demonstrated something a player who deposits once has not.
Retention covers what proportion of a partner's players remain active at thirty, ninety and three hundred and sixty-five days. Retention curves differ substantially by source and are stable enough per partner to be predictive.
Revenue and contribution covers what those players actually generate, net of bonus cost, commission, payment costs and servicing. Contribution per player, rather than revenue per player, is the figure that determines whether the relationship works.
Vertical and product mix covers what the players actually do, which matters because a partner delivering players who cross into casino from a sportsbook acquisition is delivering materially more value than one delivering single-vertical players.
Risk profile covers the proportion of a partner's players who trigger fraud flags, fail verification, are found to be duplicates, self-exclude shortly after registering, or generate chargebacks. A partner with an elevated rate on any of these is costing more than its commission suggests.
Cohorts, applied per partner
The technique introduced in the iGaming Basics course becomes considerably more powerful when applied at partner level.
Group each partner's players by the month they were acquired, then track each group's contribution over subsequent months. What emerges is a picture of whether the partner's traffic is holding quality, improving or deteriorating.
Deterioration is the pattern to watch for, and it has several common causes. The partner may have expanded into lower-quality traffic sources to grow volume. Its content may have shifted towards offer-seeking audiences that convert well and retain poorly. Its rankings may have moved from high-intent comparison terms towards broader informational terms. Or it may have begun buying traffic to supplement organic volume, which typically produces measurably weaker players.
None of these is necessarily grounds for termination, and the productive response is usually a conversation rather than a decision. A partner shown clear evidence that its recent traffic underperforms will often know exactly why, and the problem is frequently fixable. Partners are rarely told, because most programmes do not measure at this granularity.
The reverse pattern matters too. A partner whose cohorts are improving is worth investing in, and identifying that early allows the operator to secure a stronger position before competitors notice.
Effective CPA and comparability
Programmes routinely run revenue share, CPA and hybrid arrangements simultaneously, which makes partners difficult to compare directly. Effective CPA solves this.
Take the total paid to a partner over a period and divide it by the number of qualifying players delivered in that period. The result expresses what the operator actually paid per player regardless of the commercial model.
This is straightforward for CPA deals and requires care for revenue share, since payments in a given month relate to players acquired across many previous months. The more informative version calculates, for a specific cohort, the total commission that cohort will generate across its expected life, divided by the number of players in it. That figure is comparable across models and across channels, and it is the number that should inform budget allocation.
Set alongside contribution per player, effective CPA produces the ratio that actually matters: what the operator paid to acquire a player against what that player is worth net of everything. A partner with a high effective CPA and correspondingly high contribution is a good partner. A partner with a low effective CPA and negligible contribution is not a bargain.
Fraud and its signals
Affiliate fraud is varied, and almost all of it is detectable through analysis of ratios rather than through investigation of individual cases.
Incentivised registration produces registrations from people rewarded for signing up rather than motivated to play. The signal is a registration rate far above the norm paired with a deposit rate far below it, and minimal activity from those who do deposit.
Bonus abuse networks coordinate accounts to extract promotional value systematically, often with hedged positions that eliminate risk. Signals include clustered registration timing, shared device or payment characteristics, activity confined to promotional qualification, and withdrawal immediately upon meeting wagering conditions.
Cookie stuffing claims attribution for players the affiliate never introduced. Signals include click volumes implausible relative to the affiliate's visible traffic, and attribution appearing for players whose journey shows no plausible contact with the partner.
Brand bidding where prohibited intercepts players already searching for the operator. Detection requires automated monitoring of paid search results across markets, since the activity is often geographically targeted and scheduled to avoid observation.
Traffic laundering disguises the true source of traffic, typically to conceal that it originates from prohibited markets, prohibited methods or purchased sources the contract excludes. Signals include geographic patterns inconsistent with the partner's stated properties and referral data that does not match the sites it claims to operate.
Identity fraud delivers players using stolen or fabricated identities to trigger CPA payments. Signals include elevated verification failure rates, chargebacks and duplicate detection hits.
The common analytical approach is baselining. Establish what normal funnel ratios, retention curves and risk rates look like across the programme, then investigate partners whose figures depart materially without an explanation. Fraud almost always produces metrics that are implausible rather than merely poor, and the implausibility is what identifies it.
A note on proportionality is warranted. Unusual metrics have innocent explanations, and a partner whose audience differs genuinely from the norm will produce different figures. The appropriate response to an anomaly is enquiry, not immediate termination, and programmes that terminate on statistical grounds alone lose good partners and acquire a reputation for doing so.
What the channel actually costs
Assessing the affiliate channel against alternatives requires honesty about its full cost, and the comparison is frequently done badly.
The apparent attraction is that nothing is paid until a customer arrives. The full picture includes the commission itself, which under lifetime revenue share continues indefinitely; the cost of running the programme, including staff, software and compliance monitoring; the cost of disputes and reconciliation; and the strategic cost of building no owned asset, since money spent here produces customers rather than brand.
Set against paid media, the comparison should be made on fully loaded lifetime cost per player rather than on upfront outlay. A player acquired through paid media at a known cost may be cheaper over three years than the same player acquired through a lifetime revenue share arrangement, and operators that have modelled this carefully often find the gap larger than expected.
Set against owned channels, the comparison is starker. Investment in an operator's own search presence, content, brand and retention capability builds assets that continue producing without further payment. Affiliate spend does not.
None of this argues for abandoning the channel, which would be neither possible nor sensible given the traffic it controls. It argues for assessing it on the same basis as every other acquisition route, rather than treating it as costless because the payment arrives later.
Building the reporting that supports this
A closing practical point. Everything in this lesson depends on being able to link a player back to the partner that introduced them and then follow that player's contribution over time.
That requires attribution data to persist beyond acquisition into the player's ongoing record, which many systems handle poorly. It requires bonus cost, payment cost and commission to be attributable at player level rather than only in aggregate. It requires cohort reporting by partner as a standard view rather than an occasional analysis. And it requires consistent definitions, so that the quality figures discussed with a partner reconcile with the figures used internally.
Programmes that build this capability manage partners on evidence. Programmes that do not manage them on volume, on relationships and on assertion, and they systematically reward the partners best at generating registrations rather than the partners best at generating customers.
Setting benchmarks and acting on them
Measurement is only useful if it produces decisions, and that requires benchmarks against which partners can be assessed.
The practical approach is to establish, across the programme, what normal looks like for each metric: typical click-to-registration and registration-to-deposit rates, typical second deposit rates, typical retention at thirty and ninety days, typical contribution per player, and typical rates of verification failure and early self-exclusion. These benchmarks should be segmented, because a sportsbook-led partner in one market and a casino-led partner in another are not comparable, and holding them to a single standard produces meaningless conclusions.
Against those benchmarks, partners sort into recognisable groups.
High volume, high quality partners are the ones to protect and grow. The commercial question is how to secure a larger share of their traffic, and the answer is usually improved terms, better placement or product advantages rather than anything clever.
High volume, low quality partners are the ones that flatter a programme's headline numbers while damaging its economics. The productive response is a conversation supported by evidence, followed by restructuring terms towards CPA with stricter qualifying conditions if quality does not improve.
Low volume, high quality partners are frequently overlooked because they do not appear prominently in volume reporting. They are often the most valuable growth opportunity in a programme, since the traffic is proven and the constraint is scale rather than quality.
Low volume, low quality partners consume administrative attention disproportionate to their contribution and should be moved to self-service handling or terminated.
Anomalous partners, whose metrics do not fit any normal pattern, warrant investigation before conclusion. Some turn out to be fraud. Others turn out to have genuinely distinctive audiences that the benchmarks were never designed to describe.
Renegotiating on evidence
The most direct application of all this analysis is in commercial discussions, and evidence changes those discussions substantially.
A programme that can demonstrate, with cohort data, that a partner's players generate a specific contribution over a specific horizon is in a position to propose terms grounded in something. A programme that can only say the traffic feels weak is not, and the negotiation reduces to relative stubbornness.
The same works in the partner's favour, which is worth stating because this course serves both sides. An affiliate that can evidence the retention and value of the players it delivers, ideally with data from several operators, has a genuine case for improved terms, and operators respond to it. The affiliates achieving the best commercial outcomes in this sector are generally not the ones with the most traffic but the ones best able to demonstrate what their traffic is worth.
Where the analysis usually breaks down
Three practical obstacles recur, and anticipating them saves considerable effort.
Attribution does not persist. Many systems record which affiliate introduced a player at registration and then lose that link in downstream reporting, making it impossible to attribute revenue, bonus cost or retention back to source. This is the single most common blocker, and it is a data architecture problem rather than an analytical one.
Costs are not attributable at player level. Bonus cost, payment cost and servicing cost are frequently held in aggregate, which permits revenue analysis but not contribution analysis. Since contribution is the figure that matters, this limits conclusions materially.
Definitions drift. The qualifying player definition used for commission, the active player definition used in reporting, and the cohort definition used in analysis are often three different things, and reconciling them consumes more time than the analysis itself.
The remedy in each case is the same: agree the definitions once, build the attribution to persist, and treat the reporting layer as infrastructure worth investing in rather than as a byproduct of the payment system. Programmes that do this manage on evidence. Programmes that do not spend their time arguing about numbers nobody can verify.
A worked partner comparison
To draw the threads together, consider three partners assessed over a twelve-month horizon. The figures are illustrative.
Partner A delivered 600 players. Effective CPA worked out at £140. Second deposit rate was 34%, ninety-day retention 11%, and contribution per player £185. Total contribution £111,000 against £84,000 paid, a net of £27,000.
Partner B delivered 180 players. Effective CPA was £260, which on first inspection looks poor. Second deposit rate was 61%, ninety-day retention 29%, and contribution per player £640. Total contribution £115,200 against £46,800 paid, a net of £68,400.
Partner C delivered 950 players. Effective CPA was £95, the lowest of the three. Second deposit rate was 12%, ninety-day retention 3%, and contribution per player £88. Total contribution £83,600 against £90,250 paid, a net loss of £6,650.
The ranking on volume is C, A, B. The ranking on cost per player is C, A, B. The ranking on actual value delivered is B, A, C, with C actively destroying value.
A programme managed on the first two measures would prioritise Partner C, invest in growing it, and be pleased with the resulting volume growth while its economics deteriorated. This is not a hypothetical failure mode; it is the ordinary consequence of measuring what is easy to measure.
Notice also what the numbers suggest about each relationship. Partner B is the obvious growth opportunity, and the correct action is to find out what constrains its volume and whether improved terms or better placement would lift it. Partner A is solid and should be maintained. Partner C requires a conversation about traffic sources and probably a restructuring towards CPA with stricter qualifying conditions, since the current arrangement pays more than the players are worth.
Closing the loop
The material in this course connects at this point. The commercial models determine what is paid. The traffic sources determine what arrives. The tracking determines whether attribution is correct. The compliance framework determines which partners can be worked with at all. And the measurement framework determines whether any of it is producing value.
An affiliate programme that gets four of these right and measurement wrong will systematically reward the wrong partners, because it cannot see which ones are which. That is the argument for treating the analytical work as foundational rather than as reporting, and it is the single most useful conclusion to take from this course.
Key takeaways
- Registration volume tells you almost nothing. Two affiliates delivering identical numbers of players can differ by an order of magnitude in the value those players generate.
- Cohort analysis applied per partner is the single most informative tool available to an affiliate manager, and it exposes quality deterioration long before aggregate reporting does.
- Effective CPA allows revenue share and fixed-fee arrangements to be compared on a common basis, which is essential when deciding where to allocate budget.
- Fraud in this channel is varied and mostly detectable through ratio analysis, since fraudulent traffic almost always produces funnel metrics that differ implausibly from genuine traffic.
- The affiliate channel should be assessed against its alternatives on fully loaded lifetime cost, not on the apparent attraction of paying nothing upfront.