What an audit is for, and when to run one
The programme management hub sets an operating rhythm of weekly, monthly, quarterly and yearly checks that keeps most programmes from drifting. An audit is different: it looks at everything at once, from outside the routine, and asks whether the routine itself is working. Its example of an inherited programme cleaned up over a quarter shows the kind of problems an audit finds: partners who had stopped promoting, conversions from a partner using the brand name in paid search, no routine for reviewing rates.
Run a full audit when you inherit a programme, before moving it to a new platform, when results change and the regular reviews cannot explain why, or when nobody can answer basic questions such as what the programme costs per new customer. For a programme not yet open, the affiliate programme launch checklist is the gate instead.
The audit is a review, not a rebuild. Each area below asks questions and records evidence; it does not re-teach commission design, attribution or fraud patterns. Where a finding needs a fix, the guide that owns that area shows how. Other systems have audits of their own built the same way, such as the Google Ads account audit and the martech stack audit, and an affiliate audit often finds questions for both.
Before you start
- Read access to the affiliate platform, the order system, analytics and the finance ledger line for commission.
- The programme terms, current and previous versions, and any partner-specific agreements.
- The commission schedule and any partner-specific, promotional or placement deals, with their reasons and dates.
- Twelve months of data where it exists: tracked, validated and paid conversions by partner; orders, returns and new-customer flags from the order system; commission and fees from finance.
- The business's margin figures: contribution margin by product category and the target cost of acquiring a customer.
- Past reviews and tests: quarterly reviews, any incrementality tests, fraud cases and partner disputes.
- An audit log: one row per question, with the evidence, the finding and the guide that owns the fix.
The steps
Gather the records and map the programme
List what exists: the platform, and whether it is a network or an in-house platform, which the comparison of networks and in-house programmes shows decides how much of the tracking and payment work the business does itself; the partners by type and status, the commission schedule, the terms and their versions, the reports and who reads them, the routines and who runs them. Note what is missing as a finding in itself. A programme with no version history of its terms or no record of why special rates exist has a governance gap before any number is checked.
Check the economics against margin
For each segment the programme pays differently, such as partner type, customer type and product category, set commission plus platform fees beside the Contribution MarginRevenue left after every cost that varies with each sale. the orders produce. The commission structures guide sets the ceiling from margin and explains the return-on-spend floor; the break-even calculator keeps the arithmetic consistent. Ask: is any segment paying more than its margin allows? Is the same Commission RateThe share of a sale's value paid to the affiliate or partner credited with it. paid on new and returning customers? Are partner-specific and placement deals still justified by their recorded reasons? Record each answer with the figures.
Test tracking and reconcile with the order system
Run a test journey from a partner link to an order, and check the conversion arrives with the order reference, value and customer flag. Then reconcile a recent period: the platform's tracked and approved conversions, the order system's qualifying orders and the analytics view of affiliate traffic. The tracking guide says the three should tell a consistent story; where they diverge, record the size and the likely cause. If partners report missing conversions, the tutorial on tracing missing affiliate conversions runs that investigation.
Compare the terms with the platform settings
Put the terms beside the platform and check each setting: the window, the validation period, the DeduplicationRemoving double counts when one conversion is reported by more than one source. rule, product exclusions, rates by group, and how codes are tied to partners. Check the validation period against the business's returns window. Check the terms cover the hard cases: brand bidding, coupons, browser extensionsA partner whose software sits in the shopper's browser and pops up at checkout with codes or rewards, often taking the last click on an order already decided., disclosure, data, enforcement and changes. The tutorial on writing affiliate programme terms drafts any clause that is missing, and the attribution and deduplication guide covers the credit rule.
Review the partner mix and inactive partners
Count partners by type and by activity: how many are approved, how many sent a click, how many sent a validated sale in the period. Compare the mix with the customers the programme is meant to reach, as the recruitment, compliance and fraud guide describes. Look at concentration: what share of new customers and of commission comes from the largest few partners, and whether those partners are managed as accounts, which the guide to managing key affiliate partners covers. List inactive partners for review.
Sample compliance and brand search
Check a sample of partner content for disclosure, claims, prices, codes and link labels, following the tutorial on checking affiliate content. Search the brand's terms with common modifiers in its main markets and record any partner ads, as the guide to setting brand bidding rules describes. The audit question is whether checks happen and are recorded, as well as what this sample finds.
Check validation and fraud controls
Ask how pending conversions are validated: against the order system, by whom, how often, on what grounds. Check the Return RateShare of orders or units sold that customers sent back. on approved conversions by partner, and whether commission was approved on orders later returned. Look at the last few fraud or dispute cases: were they investigated, decided and recorded? The tutorial on investigating suspected affiliate fraud sets the standard for a case record. Check that the month-end routine produces figures finance can use, as the tutorial on closing the month on affiliate commissions describes.
Judge the reporting against incrementality
Read the programme's reports. Do they show quality signals, such as new-customer share, returns and order value, beside volume? Is cost of sale calculated after fees and returns? Has any test of IncrementalityConversions caused by marketing. ever been run on the partner types or practices that cost most? The reporting and incrementality guide sets out the report and the tests. A programme judged only on tracked sales has not been judged.
Find the causes behind the biggest findings
For each significant finding, ask why it happened, not only what it is. The five whys helps: a validation period shorter than the returns window may trace back to a platform default nobody changed, which traces back to no one owning the settings. Fixing the cause stops the finding returning.
Rank the fixes and set the plan
Place every fix on the impact-effort matrix: impact measured in money or risk where possible, effort in time and in the disruption to partners. Do quick, high-impact fixes first; plan the large ones; drop low-impact, high-effort ones unless they carry risk. Give each fix an owner and a date, tell partners in advance about any change that affects them as the terms require, and carry the plan into the business's own review using the quarterly marketing review template.
Audit areas, questions and evidence
| Area | Key questions | Evidence to record | Guide that owns the fix |
|---|---|---|---|
| Economics | Is each segment above its margin floor after fees? Are new and returning customers paid differently? | Commission and fees against contribution margin by segment | Commission structures |
| Tracking | Do conversions carry order references? Do platform, orders and analytics agree? | Test journey; reconciliation for a period | Tracking guide; missing conversions tutorial |
| Terms and settings | Do the platform's window, validation, credit rule and exclusions match the terms? Are hard cases covered? | Side-by-side comparison; list of gaps | Writing programme terms; attribution and deduplication |
| Partner mix | Does the mix match the target customers? How many partners are active? How concentrated is the programme? | Counts by type and activity; shares of new customers | Recruitment guide; key partners guide |
| Compliance | Are disclosure, claims and offers checked and recorded? | Sample findings; check log | Content check tutorial |
| Brand search | Are partners on brand terms? Is the rule written and monitored? | Search log; screenshots | Brand bidding guide |
| Validation and fraud | Is every conversion validated against orders? Are cases recorded? | Validation process; returns on approved conversions; case records | Fraud tutorial; month-end tutorial |
| Reporting | Are quality signals and incrementality in the report? Has any test been run? | Report review; test history | Reporting and incrementality guide |
A craft supplies retailer audits an inherited programme (illustrative)
Craflin Supplies audits the programme a departed manager left
Craflin Supplies, an invented craft supplies retailer, inherits its affiliate programme when its manager leaves. The new owner spends two weeks on an audit.
The economics show the programme pays 9% on every order. Returning customers make up 64% of affiliate orders, and on the lowest-margin categories, yarn and paper, 9% plus the platform fee comes to more than the 11% of order value Craflin has set as the most it will pay to keep a returning customer. The terms and platform show a validation period of 14 days against a 30-day returns policy; over the past year, commission was approved on 1,140 orders later returned. The terms say nothing about browser extensions, and two extension partners earn 18% of commission. Of 512 approved partners, 61 sent a sale in the past quarter. No incrementality test has ever been run.
Asking why, the owner finds the 14-day period was the platform's default and nobody owned the settings. Ranked by impact and effort, the fixes become a three-month plan. In the first month: extend validation to 35 days, give partners notice of a new-customer rate of 10% and a returning-customer rate of 4%, and remove partners inactive for a year. In the second: add an extensions clause to the terms, with notice. In the third: run a holdout on the extension partners to see whether they add orders, and name the five content partners that bring most new customers as key partners.
When the economics review finds one rate for every order, the affiliate programme version of the break-even calculator rebuilds the commission ceiling separately for new and returning customers.
Common mistakes
- Auditing from the platform's reports alone. Tracked sales are one view; reconcile with the order system and finance.
- Comparing with someone else's programme. The test is the business's own margin and returns, not another programme's rates or shares.
- Recording opinions instead of evidence. Each finding needs the figure, the setting or the page that shows it.
- Fixing symptoms. A setting corrected without an owner drifts back; find the cause.
- Changing everything at once. Partners need notice, and a programme that changes every rate in one week loses trust and makes the effect of each change impossible to read.
- Leaving out compliance and brand search. Economics and tracking are easier to measure, but terms nobody enforces are also findings.
- No plan at the end. An audit that ends with a list rather than ranked, owned and dated fixes changes nothing.
