Missing sales, rejected sales and sales another channel won
The tracking guide describes the chain that has to hold for a partner to be paid: the link carries a partner identifier and a click identifier as URL parametersA key and value appended to a URL after a question mark, such as a filter, sort order, session or tracking tag, that may or may not change the page., the click is logged, the identifier is stored on the advertiser's side or in a cookie, and the conversion reports back with it, usually through a postback with a browser pixel as backup. Every link in that chain can break. But when a partner says sales are missing, a broken chain is only one of three explanations, and the partner cannot tell them apart from their side.
Never recorded. The customer clicked, bought, and the platform never heard about it. This is a tracking fault, and it is usually on the advertiser's side: a lost parameter, a payment path that skips the page or event the postback depends on, a consent state that blocks storage, a checkout change nobody tested.
Recorded, then rejected. The platform saw the conversion, and validation rejected it: a return, a cancellation, a duplicate, a coupon used outside the terms. The partner's report may show it as declined, or the partner may be looking at a figure that excludes it. This is a decision, not a fault, and the question is only whether it was made under the terms.
Recorded for someone else. The order was tracked, but another partner or channel took the credit under the programme's deduplicationRemoving double counts when one conversion is reported by more than one source. rule, or the click fell outside the lookback windowThe time after an ad interaction in which a conversion can still be credited to it.. The attribution and deduplication guide explains how those rules work; this tutorial only rules them out as a cause.
The opposite symptom, where channels together claim more conversions than the business took, has its own troubleshooting page. If a missing-sales report turns up conversions that look manufactured rather than lost, stop here and follow the tutorial on investigating suspected affiliate fraud.
Before you start
- The partner's evidence. Dates, the pages or placements involved, any sub-identifiers they pass, and ideally order references or the times of the customers' purchases. Without something specific the trace has nothing to match.
- Access to the platform's records. Clicks, conversions with their status, the reason for any rejection, and the click and order identifiers carried on each one.
- Access to the order system. Order references, times, payment methods, status, and whether the order carried a stored click identifier.
- The current programme terms. What they say about validation, deduplication, the lookback window, declined consent and adjustments to commission after the fact.
- Tag Manager access, if the conversion fires through a container. Edit rights let you run preview mode, which the debugging and preview tutorial walks through.
- A record of recent changes. Releases to the site, checkout, payment methods, consent banner and container over the period the partner names. Most gaps start on a date something changed.
The steps
Collect the partner's evidence and pin down the claim
Ask the partner for what they can see: the dates the gap covers, which pages, links or sub-identifiers sent the traffic, and any order details their audience shared with them. Ask what makes them think sales are missing, such as clicks holding steady while conversions fell, or readers saying they bought. Write the claim down in one sentence, with its dates. A claim of "sales are low" is a performance question; a claim of "clicks are normal but conversions stopped on a date" is a tracking question, and it is the one this tutorial answers.
Match the claim to your own orders
Take the partner's dates, references or times and look for the orders in the order system. For each, note whether it exists, whether it carries a stored click identifier, and whether the platform shows a conversion for it. Where the partner has no order references, compare the shape instead: their clicks by day against platform conversions by day for that partner, and against orders in the same period. The gap is real if clicks continue and conversions stop or fall while orders do not.
Rule out rejections and lost credit
For each order the platform did record, check its status. Rejected conversions have a reason in the platform or in your validation record; check that the reason is one the terms allow and tell the partner which it was. For orders that carry a click identifier but sit with another partner or channel, check the deduplication rule and the lookback window in force on those dates. If the outcome matches the rules, the order is not missing; it was decided. Explain it to the partner with the clause it relies on, and set those orders aside.
Replay the journey on the paths customers use
Click the partner's real link, not a test link you built, and buy through every path a customer might take: guest and account checkout, each payment method including digital wallets and buy-now-pay-later options, a discount codeA code a customer enters at checkout to receive an offer, also used to identify sales from a creator, partner or channel that links cannot capture., a mobile browser and a desktop one, and a journey that leaves and comes back later inside the window. The affiliate programme launch checklist asks for the same test journey before launch; here you run it on the live programme. Note for each path whether the click identifier survives to the order and whether the platform records the conversion.
Check links, redirects and stored identifiers
Where the replay loses the conversion, follow the identifier. Does it appear on the landing page URL? Does a redirect for a region, a new domain, a consent page or a link shortener drop the query string? Is the identifier read from the URL and stored, and is it still there at checkout and on the order record? The tracking guide's links table lists what breaks each element. Fix nothing yet; record exactly where the identifier disappears.
Check consent states
Repeat the failing path with each consent choice the banner offers. As the privacy and consent guide and Google's consent mode documentation describe it, the site sets default consent states before any tag loads, the banner updates them, and tags adjust; with consent mode in its basic form, tags do not fire until consent is given. Check that the affiliate tracking follows the choice the customer made, no more and no less. A conversion lost because the customer declined is correct behaviour. A conversion lost after the customer agreed, because the banner's update never reached the tag or the stored identifier, is a fault.
Check the postback and the pixel
Find the order the replay created and check whether the postback sent it to the platform with the click identifier and order reference. If the conversion fires through Tag Manager, run preview mode on the confirmation step: the debugging tutorial explains how the event timeline shows which tags fired on each event, which did not, and which trigger condition failed. Preview runs on the workspace you start it from, so to see what customers currently receive, preview from a workspace with no pending changes. If the postback runs through a server container, the server-side tagging basics explain where it sits; check that it receives the event and the consent state. A pixel that fires without a postback, or the reverse, tells you which half failed.
Name the cause and fix it
Ask why the identifier or the event was lost, and keep asking. The five whys stop the trace at a cause you can fix rather than a symptom: the postback did not fire, because the payment method returns to a different page, because the trigger listens only to the old page, because the release that added the method did not include a tracking test. Fix the cause, then rerun the failing path and every other path to confirm the fix did not break something else. Record the date the fault started and the date it was fixed; those two dates bound the orders to re-credit.
Re-credit the affected orders under the terms
List the orders between the two dates that went through the broken path and carry evidence of the partner's click: a stored identifier on the order, or a click in the platform's logs that matches. Apply the same validation you apply to every conversion, so returns and cancellations are excluded. Then add them through whatever adjustment process you have agreed with your network or platform; agree one now if none exists, and write it into the terms. Tell the partner what happened, which orders were affected, how they were credited and what changed so it does not happen again. Record the adjustment where the month-end will find it; the tutorial on closing the month on affiliate commissions covers matching adjustments to the invoice at month-end.
Add a check that catches the next gap sooner
A gap the partner found first is one your monitoring missed. Add an alert on sudden drops in recorded conversions per partner, as the tracking guide recommends. Add the test journey on every payment path to the release checklist for checkout, payment, consent and container changes. Reconcile platform, order system and analytics totals each period, using the checks in the data quality guide. Keep the trace record with the partner's account so the next question is answered faster. If the gap appeared after a move between platforms, the tutorial on moving an affiliate programme to a new platform lists the tests every path should pass before partners switch.
Where conversions go missing
The table lists the break points a trace most often finds, in the order the chain passes through them. Each row assumes the order exists and the partner's click is real.
| Break point | Symptom | Check | Fix |
|---|---|---|---|
| Partner link | Clicks low or absent in the platform although the partner reports traffic | Click the live link; compare with the link the platform issued | Reissue links; deep-link through the platform's method |
| Redirects | Clicks recorded, no conversions after a site change | Follow the link and watch the query string at each hop | Preserve parameters through every redirect |
| Stored identifier | Identifier on the landing page, absent on the order | Inspect storage at landing, at checkout and on the order record | Store on the advertiser's own domain; pass it to the order |
| Consent | Conversions lost only for some consent choices | Repeat the journey with each banner option | Make tags follow the update, not only the default |
| Payment path | One payment method or checkout type never converts | Replay each path; compare orders by payment method | Fire the postback from the order event, not one page |
| Postback | Pixel fires, platform sees nothing from the server | Check the server's outgoing call for the test order | Repair the trigger or integration; alert on drops |
| Deduplication | Order tracked but credited elsewhere | Check the rule and the window in force on the date | No fix; explain under the terms |
| Validation | Conversion recorded, then declined | Read the rejection reason against the terms | No fix if the reason stands; correct it if not |
The last two rows are not faults. They belong in the trace because a partner sees them as missing sales, and an answer that shows the rule and the clause usually ends the question.
A stationery shop finds its lost wallet orders (illustrative)
Plovrick Paper and a planner reviewer's missing month
Plovrick Paper, an invented online stationery shop, hears from a partner who reviews paper planners that a month of sales is missing. The partner's clicks to Plovrick held steady at about 1,900 for the month, but recorded conversions fell from around 60 to 23. Readers had told the partner they bought.
Plovrick matches the partner's click dates against its own orders. It finds 58 orders that month carrying the partner's stored click identifier. The platform recorded 23 of them and rejected none. All 35 unrecorded orders were paid with a digital wallet, which Plovrick added to its checkout on the 4th of the month.
The replay confirms it. Card and account checkouts record the conversion. A wallet payment returns the customer through a separate confirmation page, and the postback's trigger listens only to the old one, so it never fires. Preview mode shows the trigger condition failing on that page. Plovrick changes the postback to fire on the order event rather than a page, reruns every payment path, and adds a wallet test to its checkout release checklist.
It applies its usual validation to the 35 orders, excluding 3 returns, and submits 32 through the adjustment process agreed with its platform, as its terms provide. It tells the partner which orders were affected and why, and adds an alert for any partner whose recorded conversions fall sharply while clicks hold.
Common mistakes
- Paying on the partner's list without checking. Some named orders were rejected under the terms or won by another partner. Match first, then re-credit only what a fault on your side lost.
- Testing with a link you built. Partners' real links pass through redirects and deep links your clean test link never sees. Replay the journey from the partner's page.
- Testing only the main checkout. Wallets, guest checkout, codes and mobile browsers are where gaps hide.
- Treating declined consent as a fault. Where the customer said no, the conversion should not be tracked. Fix only losses that happen after consent was given.
- Previewing the wrong workspace. Preview shows the draft you started it from. Debug the version customers receive before blaming a change that has not been published.
- Fixing the report instead of the cause. A manual credit without a repaired trigger means the same partner calls again next month.
- Re-crediting without validation. Restored orders still have to pass the checks every other conversion passes.
- Leaving the partner to find the next gap. An alert on per-partner drops and a test on every release would have found this one first. The fundamentals hub and the networks and in-house comparison show who owns each part of the tracking chain, and so who has to watch it.
