Why affiliate credit is contested
A typical online purchase involves several visits. A customer reads a review on a partner site, clicks through, leaves, sees a social ad, searches the brand name later, clicks a paid search ad, gets an abandoned-basket email, and buys after searching for a discount code on a voucher siteAn affiliate partner that publishes discount codes and offers, often reached by shoppers who have already chosen and search for a code at checkout.. Every one of those channels has a platform or a report that can claim the sale.
The affiliate platform sees two of those touches: the review partner and the voucher site. Under its default rule, the voucher site gets the commission, because it was the last affiliate click. The paid search platform claims the sale because its click was within its window. Email claims it because the customer opened the message. Analytics, using its own model, credits one or several of them. The order system records one order.
AttributionAssigning conversion credit to touchpoints. is the general problem of assigning credit across touchpoints. The analytics root's comparison of attribution models covers the models themselves. What makes the affiliate case distinctive is money: affiliate credit becomes a payment to a third party, and every order wrongly credited is commission paid for nothing. That is why affiliate attribution needs written rules rather than a dashboard setting. The problem is sharpest where buyers research across review and deal sites before paying a retailer, as the consumer electronics marketing guide describes.
How the click and conversion are recorded in the first place is covered in how affiliate tracking works; this guide starts once conversions are being tracked.
Last-click and the affiliate default
Affiliate platforms default to last-click among affiliates for practical reasons: it is simple to explain, easy to put in a contract and easy for partners to verify against their own records. The default has predictable effects.
| Rule | How it works | Who it favours | Main risk |
|---|---|---|---|
| Last affiliate click | The most recent affiliate click in the window wins | Partners at the end of the journey: vouchers, cashback, extensions | Content partners who introduced the customer go unpaid |
| First affiliate click | The earliest affiliate click in the window wins | Content and review partners | Rewards early touches that may have had little effect |
| Last click, with protection | Last click wins unless an earlier partner is in a protected group | Content partners, while keeping simplicity | Defining the protected group fairly |
| Split commission | Credit shared between introducing and closing partners | Both, in proportions the terms set | Complexity; small payments to many partners |
| Cross-channel last click | The last click across all channels, including paid and email, wins | The business's own channels | Partners lose orders they influenced; must be in the terms |
| Data-driven or modelled | Credit shared by a statistical model | Whichever channels the model favours | Hard to explain to partners; not usually used for payment |
Most programmes pay on one rule and analyse on another. Payment needs a rule partners can check; analysis needs a view of contribution. Keep the two distinct, and do not let the payment rule be mistaken for a measure of value. The attribution models framework sets out the general trade-offs.
Deduplication rules across channels
Deduplication is the decision about which conversions the affiliate programme should pay for at all, given everything else the customer did. It is usually done in one of two ways. The advertiser's site or server decides at the moment of conversion whether to fire the affiliate postbackA server-to-server message reporting a conversion, install or in-app event back to the ad network or affiliate platform that should receive credit., based on the last marketing touch it recorded. Or the advertiser reviews pending conversions during validation and rejects those another channel should win. The first is cleaner; the second is easier to implement.
The rule needs to specify three things. Which channels override an affiliate click. Common choices are paid search, paid social and email clicks, sometimes only on brand terms. Direct and organic visits usually do not override, because they cannot be distinguished reliably from the customer returning on the partner's recommendation. Within what period. An override that applies to any click in the cookie window is much harsher than one that applies only to clicks shortly before purchase. Between affiliates. Whether the last affiliate always wins, or whether protected partners keep their credit when a voucher or extension partner clicks last. Orders that carry a partner's own discount code, with another partner's click or with no click at all, need a decision too, and the guide to crediting affiliates for orders made with their codes settles it beside this rule.
To implement it, the site needs to know the channel of each visit. That means consistent analytics parameters (UTMURL tags identifying traffic source. tags) on paid and email links, a first-party record of recent touchpoints, and a checkout process that reads that record. The paid search team should not be adding affiliate parameters to their own links, and partners should not be passing paid search traffic through affiliate links without permission.
Write the rule in the programme terms before it is applied. Partners accept deduplication when they know it in advance; they leave programmes that apply it in secret. In a lead programme the same rule matches each partner lead against the CRM and recent direct enquiries, as the guide to affiliate lead generation describes.
Windows, overrides and chargebacks
Several time-based settings interact, and inconsistencies between them generate most disputes.
Cookie windows. Each channel and each platform has its own. An affiliate window may be weeks long while a paid search platform credits clicks over a different period and analytics uses yet another lookback. A long affiliate window and a short deduplication period means an affiliate click from weeks ago can win against nothing, while a recent one loses to a paid search click the same day. That may be the intended result, but it should be chosen deliberately.
Overrides. Some programmes let particular partners override the normal rule, for example a strategic partner whose clicks are never deduplicated, or an extension partner whose clicks never overwrite a content partner's. Keep a list, review it, and show it in the terms.
Returns, cancellations and chargebacks. A tracked sale that is returned, cancelled or reversed by the card issuer should not stay in either the payment or the attribution figures. The validation period handles returns before payment; 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. can arrive later and need a clawback clause. For reporting, use validated rather than tracked conversions, or the programme's return on spend (ROASRevenue generated per unit of ad spend.) and cost per action (CPACost of each conversion.) will look better than they are.
Reconciling with analytics
Even with good deduplication, three systems will report different numbers: the affiliate platform, analytics and the order system. Reconciliation is not about choosing the right one; it is about explaining each gap so the differences stop being a source of suspicion. The analytics root's guide to platform versus analytics discrepancies covers the general causes.
Start from the order system, the only complete record. Match orders to affiliate conversions by order reference, which is why every postback should carry one. Then classify each affiliate conversion: credited by both, credited by the platform but another channel in analytics, credited by the platform with no analytics session (consent declined, tag blocked), rejected at validation, or returned later.
When the platform's total is far above what analytics credits to the channel, the diagnosis of affiliate sales not matching analytics classifies every tracked conversion by the reason analytics does not show it.
Analytics tools document their own attribution settings and lookback windows; Google Analytics Help, for example, describes how its reports assign credit and how settings can be changed. Know the settings in your own property before comparing. When the reconciliation is one part of a wider review of an inherited programme, the tutorial on auditing an affiliate programme sets it beside the economics, terms, partner mix and reporting checks.
An electronics retailer reconciles a quarter
Voltara, an invented electronics retailer, reconciles a quarter in which its affiliate platform tracked 4,200 conversions worth £1.26 million. The order system matches 4,140 by reference; 60 are test or duplicate records. Of the matched orders, 2,350 are credited to affiliates in analytics, 910 to paid search, 380 to email and 290 to direct, and 210 have no analytics session because consent was declined. Validation rejected 340 orders for returns and cancellations. Applying its written rule, under which paid search and email clicks within 24 hours of purchase override affiliate clicks, Voltara deduplicates 640 orders. Its reconciliation table explains every gap between the platform's figure and the 3,160 orders it pays commission on.
Explaining differences to stakeholders
Finance wants one number, the affiliate team wants credit for its partners, and the paid search team wants credit for its clicks. A reconciliation table that runs from tracked conversions to paid conversions, line by line, ends most of those arguments, because each step has a reason.
Present three figures and say what each answers. Tracked conversions: what partners touched, from the platform. Paid conversions: what the business paid commission on after validation and deduplication. Incremental conversions: what the programme added beyond what would have happened anyway, from tests, which the reporting and incrementality guide covers. Each is correct for its question; none is the whole truth.
Show the trend in each, not just the level. A widening gap between tracked and paid conversions can mean a tracking problem, a growing share of voucher or extension partners closing orders other channels started, or a fraud pattern. And explain changes in rules before they appear in the numbers. A new deduplication rule that lowers paid conversions looks like a collapse if nobody was warned. The measurement hub places this work among the other measurement tasks.
Common mistakes
- Treating the platform's figure as revenue the programme created. It is what partners touched under the platform's rule, nothing more.
- Deduplicating without telling partners. Rules applied in secret destroy trust and lose the best partners first. Put them in the terms.
- Letting the payment rule pass for a measure of value. Last-click is a contract, not an analysis.
- Overriding on direct and organic visits. They often represent the customer returning on a partner's recommendation and cannot be separated reliably.
- Reporting tracked rather than validated conversions. Returns, cancellations and chargebacks belong out of the figures.
- Comparing systems with different windows without saying so. Know each system's lookback before explaining a gap.
- No order reference in postbacks. Without it, reconciliation becomes guesswork.
