Why affiliate numbers are contested
Every channel's platform reports its own results generously, but affiliates have three features that make the problem sharper. First, the report is also an invoice. The conversions an affiliate platform tracks become commissions, so overstated credit is not just a vanity metric; it is money paid out. Second, the platform only sees affiliate touchpoints. It cannot know that the customer clicked a paid search ad after the partner's link, or opened an email the day before, so it credits the partner by default. Third, a large share of affiliate activity sits at the very end of the journey, where coupon, cashback and comparison partners meet customers who have often already decided to buy.
None of this means affiliates add little. Content partners, comparison sites and business referrals often bring customers the business would not otherwise have reached. It means the number on the platform dashboard is a starting point, not an answer. A return on spend (ROASRevenue generated per unit of ad spend.) calculated from platform-tracked revenue will almost always look better than the programme's true contribution, and decisions on commission, partners and budget should rest on deduplicated, quality-adjusted and tested figures.
The affiliate marketing guide introduces measurement as the third discipline of the model; this section treats it in depth.
Attribution, reporting, incrementality
Three activities make up affiliate measurement. They answer different questions and use different evidence.
| Activity | Question | Evidence | Typical output |
|---|---|---|---|
| Attribution and deduplication | Which channel, and which partner, gets credit for this order? | Click paths, platform records, analytics, order data | Deduplicated affiliate conversions; credit rules |
| Reconciliation | Why do the platform, analytics and the order system disagree? | Matched order references across systems | A reconciliation table explaining each gap |
| Programme reporting | What is each partner delivering, at what cost? | Validated conversions, customer type, order value, returns, commission and fees | Partner-level KPIA metric tied directly to an objective. report |
| Quality analysis | Are these the customers we want? | New-customer share, return rates, repeat purchase | Partner quality scores and tiers |
| Incrementality testing | What would have happened without the programme or a partner? | Holdouts, coupon tests, brand-term tests, time-based comparisons | An estimate of incremental orders and true cost per incremental customer |
Attribution decides how credit is split; IncrementalityConversions caused by marketing. asks whether there was anything to credit at all. They are complementary. A programme can be perfectly deduplicated and still be paying for orders that would have happened anyway.
The core guides
Affiliate attribution and deduplication explains why affiliate platforms default to last-click among affiliates, how to deduplicate credit against paid search, email and other affiliates, how windows and overrides interact, how returns and chargebacksA sale reversed by the customer's card issuer, which can arrive after other checks have closed and should not stay in payment or attribution figures. change the picture, and how to reconcile the platform report with analytics. It applies the general attribution models to the affiliate case.
Affiliate reporting and incrementality builds the programme report: the partner-level indicators that matter, quality signals such as new-customer share and return rates, cost of sale against margin, and the incrementality tests that show whether high-volume partners, coupons and brand-term activity add anything. It finishes with the review cadence and partner tiers that turn the report into decisions.
For the theory underneath, the analytics root's guide to choosing an attribution approach is the right starting point. The affiliate guides assume it and apply it.
Further guides
Further pages take the validated figures to finance each month and settle when a partner's own code earns credit:
- Closing the month on affiliate commissions — a tutorial for validating pending sales, approving, estimating what is owed, matching the invoice and reporting the cost to finance clearly.
- Crediting affiliates for orders made with their codes — a guide to when a partner's own code earns commission without a click, which partner is paid when a code and a click disagree, and what a leaked code does to that credit.
- Setting the affiliate lookback window from your own orders — a tutorial for choosing how long a click counts: the delay between click and order by partner type, what each candidate window would credit and cost, its place beside the credit rule, and the terms and platform settings that record it.
Affiliate data in the measurement plan
Affiliate measurement fails most often because nobody planned it. Tracking was set up by the platform, the report is whatever the platform shows, and analytics was never told how to recognise partner traffic. A measurement plan prevents that by deciding in advance what is measured, where, and how it connects.
For affiliates, the plan should cover a few specific points. Identification. Partner traffic carries consistent analytics parameters, so that analytics can see it as a channel and, ideally, by partner. Order references. Every affiliate conversion carries the order or lead reference, so that platform records can be joined to the order system and to analytics. Customer type. The order system supplies a reliable new or returning flag, which both commission and reporting depend on. Credit rules. The business has one written rule for deduplicationRemoving double counts when one conversion is reported by more than one source. across channels, applied in the platform and in reporting. Testing. The plan names the incrementality questions to answer each year and the method for each.
A fashion retailer adds affiliates to its measurement plan
Northvale, an invented fashion retailer, finds that its affiliate platform reports £1.9 million of sales a year while analytics attributes £0.8 million to the affiliate channel. Its measurement plan for the next year adds consistent analytics parameters for every partner link, passes the order reference in every postback, and writes a rule that paid search and email clicks within 24 hours of purchase override an affiliate click. After the first quarter, deduplicated affiliate sales are £410,000 against £520,000 tracked on the platform, the gap is explained order by order, and the team schedules a coupon holdout for the summer sale.
Common affiliate measurement problems
| Problem | Usual cause | Guide |
|---|---|---|
| Platform and analytics disagree sharply | Different attribution rules and windows; no shared order reference | Attribution and deduplication |
| Affiliates and paid search both claim the same orders | No cross-channel deduplication rule | Attribution and deduplication |
| One partner overwrites another at checkout | Last-click among affiliates with no protection for earlier partners | Attribution and deduplication |
| Partners whose readers buy with their code but never click go unpaid | No written rule for orders made with a partner's code and no tracked click | Crediting affiliates for orders made with their codes |
| Partners say orders arrive after the window closes | The window was never read against the delay in the programme's own orders | Setting the lookback window from your own orders |
| Commission paid on orders later refunded | Reporting on tracked rather than validated conversions | Attribution and deduplication, Closing the month on commissions |
| Programme looks profitable but margin falls | Reporting revenue, not cost of sale against margin | Reporting and incrementality |
| Voucher partners dominate the top of the report | Ranking partners on volume alone | Reporting and incrementality |
| Nobody knows what the programme adds | No incrementality test ever run | Reporting and incrementality |
| Reviews happen only when finance asks | No review cadence or partner tiers | Reporting and incrementality |
When the platform's total and the analytics figure are already being quoted against each other, the affiliate troubleshooting section splits the gap into its separable causes and matches both to the order system before either is treated as the record.
