Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

Affiliate Marketing › Affiliate Measurement

Affiliate Reporting and Incrementality

How to report an affiliate programme and test what it adds: partner-level KPIs, quality signals, cost of sale, incrementality tests, and the review cadence.
Guideadvanced8 min readUpdated

What the programme report must answer

A programme report exists to support decisions, so start from the decisions it has to inform. Which partners deserve more commission, more support or a higher tier? Which should be restricted or removed? Is the programme's total cost justified against its contribution? Are the rules on coupons, brand bidding and deduplication working?

Those decisions need answers to five questions. Volume: how many validated conversions, and how much validated revenue, did each partner deliver after deduplication? Quality: were those new customers, at good order values, who kept what they bought? Cost: what did each partner cost in commission, bonuses and the platform's share or fee? Efficiency: what is the cost per action (CPA) and return on spend (ROAS) on validated figures, and how do they compare with margin? Contribution: how much of this was incremental?

The platform answers the first question partially and the others hardly at all. Most of the report has to be built by joining platform data to the order system, which is why every conversion should carry an order reference. How credit is assigned in the first place, and how the platform's figures are reconciled with analytics, is covered in affiliate attribution and deduplication; this guide assumes validated, deduplicated conversions.

Partner-level KPIs

A small set of indicators, calculated for each partner and for the programme, covers most decisions. A KPI tree helps show how they connect to the business's goals, and the analytics root's guide to KPI trees explains how to build one.

IndicatorWhat it showsCalculated fromWatch for
Validated conversionsOutcomes that survived validation and deduplicationPlatform data matched to ordersGap between tracked and validated
Validated revenueValue of those outcomesOrder systemRevenue concentrated in a few partners
New-customer shareProportion of conversions from first-time customersOrder system flagPartners with very low shares
Average order valueBasket size from the partner's customersOrder systemPartners sending small, discount-led baskets
Return or cancellation rateShare of orders reversedOrder systemRates well above the business average
Total costCommission, bonuses, placement fees and platform costsPlatform and financeCosts outside commission left out
Cost per new customerTotal cost divided by new customersDerivedThe single most telling efficiency measure
Cost of saleTotal cost as a share of validated revenueDerivedCompare with margin, not with other channels' platform figures
Return on spendValidated revenue divided by total costDerivedAbove break-even, after margin

Choose a few headline KPI figures for the programme, usually validated revenue, cost per new customer and cost of sale, and use the rest for partner-level diagnosis. The ROAS calculator and CPA calculator help turn the raw inputs into rates.

Quality signals: new customers, returns, order value

Volume tells you which partners are busy. Quality tells you which are worth paying.

New customers. The proportion of a partner's conversions that come from first-time customers is the best single quality signal. Partners at the end of the journey, voucher, cashback and browser-extension partners in particular, tend to show low shares because they reach people already buying. Content, comparison and referral partners tend to show higher ones. A partner with a high share is probably bringing demand the business would not otherwise have; one with a low share may be collecting commission on orders already in progress.

Order value and basket composition. Partners whose customers buy larger baskets (AOV) or higher-margin products are worth more per order. Partners whose customers buy only discounted items, or only with a code, may cost more than they appear to.

Returns and cancellations. Return rates well above the business average can signal misleading promotion, incentivised orders or fraud. Check them before approving tiers.

Retention. Where the business can follow customers over time, compare repeat purchase and lifetime value by acquiring partner. The LTV and CAC guide explains how to compare that value with acquisition cost. A partner whose customers return often can justify a higher commission than its first-order figures suggest.

Cost of sale and margin

Affiliate efficiency is usually expressed as cost of sale: total programme cost as a share of revenue. It is a useful figure only when compared with margin. A cost of sale that looks low beside other channels can still be unprofitable on a low-margin category, and one that looks high can be excellent on new customers with good lifetime value.

Build the figure carefully. Include every cost: commission, bonuses, placement fees, the network's share or the in-house platform's licence, and a fair share of the people who run the programme. Use validated revenue, not tracked. Calculate it by partner type, by product category and by customer type, because the averages hide where money is lost.

Then compare with contribution margin. A segment whose cost of sale exceeds the margin left after product, fulfilment and payment costs is losing money on each order. That is acceptable only if the customers are new and their later purchases repay the loss within the business's payback period. The commission guide covers how to reshape the structure when a segment fails this test; the report's job is to show it clearly.

Where the report already shows the affiliate line climbing faster than validated revenue, the troubleshooting page on affiliate commission costs rising works from that rebuilt cost of sale to the cause behind each segment.

Incrementality: coupons, brand terms, holdouts

Attribution decides how to split credit. Incrementality asks whether there was anything to credit. For affiliates, the doubt concentrates in a few places, and tests can be designed for each. The analytics root's guide to incrementality testing covers test design in depth: control groups, test length, power and reading the results. The affiliate-specific versions are below.

Coupon and voucher tests. Remove a widely published code, or stop voucher partners publishing codes for a period in some regions, and compare total orders and full-price orders against a control. If total orders hold steady and fewer use codes, the partners were mostly capturing existing demand and the discount was being given away.

Brand-term tests. Where partners promote on the brand's own name, restrict that activity for a period and measure total brand-search traffic and orders. If the brand's own listings recover the same customers, the partner activity was not adding them. The brand search holdout test applies the same logic to the brand's own paid search ads, with competitor and affiliate presence on the brand terms read as guardrails.

Partner-type holdouts. Pause a partner type, such as cashback or extension partners, in some regions or for a randomly chosen share of traffic, and compare total sales with regions or groups where it continues. This is the strongest test, and the most disruptive to partners, so agree it with them and keep it short.

Before-and-after comparisons. When a partner leaves or a programme pauses for operational reasons, compare total sales with the trend and with similar periods. This is weaker evidence than a planned holdout but costs nothing. The hierarchy of evidence for marketing claims ranks these designs and matches the strength needed to the size of the decision.

Illustrative example — not a real result

A beauty retailer tests its voucher partners

Petal & Pine, an invented beauty retailer, pays voucher partners £180,000 a year in commission. It splits its regions into two matched groups and, for six weeks, asks voucher partners to stop publishing its codes in one group. In the test regions, orders using voucher codes fall by 71%, but total orders fall by only 4% compared with the control regions, and full-price orders rise. Petal & Pine estimates that roughly one order in eight credited to voucher partners was incremental. It cuts the voucher commission rate on returning customers to zero, keeps a reduced rate for new customers, and moves £60,000 a year into higher tiers for review partners, whose test-region sales were unaffected.

Review cadence and partner tiers

The report is only useful if it changes decisions, and that needs a fixed rhythm.

Monthly, alongside validation: check the headline indicators, look for anomalies by partner, and confirm that deduplication and validation rules are being applied. Small adjustments, such as a bonus for a strong partner or a question to one whose figures look wrong, happen here.

Quarterly: review every active partner against the full set of indicators and place each in a tier. A simple scheme works: partners to grow, with higher commission and more support; partners to maintain; partners to fix, with a conversation about quality or compliance; and partners to remove. Tiering on quality signals as well as volume keeps the programme from rewarding only the partners closest to the checkout. The few partners at the top of the grow tier usually need more than a rate, and the guide to managing key affiliate partners runs them as accounts with a joint plan and regular reviews. The tutorial on running the quarterly affiliate partner review walks the quarterly tiering from the partner table to the recorded decisions.

Once or twice a year: run an incrementality test on the partner type or practice that costs most and is least certain, and feed the result into the commission structure. Present the programme's tracked, paid and incremental figures side by side to leadership, so that the programme is judged on all three. The measurement hub shows how this report fits with attribution and the wider measurement plan.

The affiliate channel page for online sales turns these reports into a plan, with incremental orders by partner type and effective commission as the measures that decide rates.

Regional partner holdouts are a form of geo test, and the comparison of geo holdout tests and platform conversion lift studies explains why a regional design measures total sales independently of any platform's tracking.

Common mistakes

  • Reporting the platform's dashboard as the programme report. It shows tracked activity, not validated value or cost against margin.
  • Ranking partners on revenue alone. Volume rewards end-of-journey partners. Add new-customer share and cost per new customer.
  • Leaving costs out of cost of sale. Platform fees, bonuses, placements and staff time all belong in it.
  • Comparing affiliate return on spend with other channels' platform figures. Compare on validated, deduplicated figures, against margin.
  • Never testing incrementality. Without a test, nobody knows how much of the programme's credited revenue would have arrived anyway.
  • Testing everything at once. Pausing the whole programme answers one broad question and damages partner relationships. Test the doubtful parts.
  • Reviewing without acting. A quarterly review that never moves a partner between tiers or changes a rate is a report, not management.

Frequently asked questions

What is the single most useful affiliate metric?

There is no single one, but cost per new customer comes closest. It combines what the programme pays, including fees, with the outcome most businesses value most, and it exposes partners whose volume comes mainly from existing customers. Pair it with a return-on-spend figure on validated revenue and a periodic incrementality estimate, and the report covers efficiency, quality and causality.

How do I test incrementality for affiliates without switching the programme off?

Test parts of it. Pause or restrict a single partner type, such as voucher or cashback sites, in some regions or for some periods, and compare total sales against similar regions or periods where it continues. Withdraw a coupon from publication and see whether orders using it simply move to full price. Restrict brand-term activity and measure total brand traffic. Each test answers a narrower question than switching everything off, and together they show where the programme adds most.

How often should partner performance be reviewed?

Look at the main indicators monthly, when commissions are validated, so that anomalies are caught early. Make decisions on tiers, commission and partner removal quarterly, when there is enough data to separate trend from noise and partners have had time to respond. Run incrementality tests once or twice a year on the partner types or practices that cost most.

Terms used on this page

Tools to try

Frameworks

Where this page is used

Also related

Sources

  1. Performance Marketing Association — PMA

Related