Affiliate Traffic in GA4: Why Your Partner Channel Reports as Referral
A partner manager sends a screenshot. The network reports 41 sales last month at 6 percent commission. GA4, filtered to the affiliate campaign, reports nine key events. Finance wants to know which number to book.
Neither number is wrong. They count different things at different moments, and the gap has a shape once you know what survives the click. The bigger problem usually sits further upstream. Most affiliate traffic never reaches the Affiliates channel at all, so the question starts with a missing row rather than a disagreement about counts.
What the click looks like before it reaches you
A visitor clicks a link on a publisher's page. That link rarely points at your product page. It points at the network, which sets its own cookie, records the click against a publisher ID, and redirects to your landing page with an identifier attached.
Your tag then reads two inputs: the URL the visitor arrived on, and the referrer the browser sends with the request. Everything GA4 records about that session comes from those two inputs plus cookies already sitting on your domain.
Three outcomes follow, and they look nothing like each other in a report.
- Tagged link with utm_medium=affiliate. The session lands in the Affiliates channel, with the network or the publisher as the source.
- Untagged link. The only signal is the referrer, so the session lands in Referral with the network's redirect domain as the source. That value is usually a hostname nobody on the team recognises, not a publisher name.
- Referrer stripped. A redirect that runs through JavaScript, a referrer policy set to no-referrer, or a link attribute that suppresses it leaves GA4 with nothing. The session is recorded as Direct.
| What happens | What GA4 records | Usual cause |
|---|---|---|
| Redirect with campaign parameters | Affiliates channel, network or publisher as source | Destination template in place |
| Redirect with no parameters | Referral, network hostname as source | No template at program level |
| JavaScript or no-referrer redirect | Direct, medium (none) | Referrer suppressed in transit |
| Purchase in a later session | Direct session, event credited to the affiliate source | Last non-direct click, cookie still present |
The Affiliates channel has one condition
GA4's default channel group includes an Affiliates channel, and the rule behind it is narrow: medium equals affiliate. A session with medium referral goes to Referral even when the visitor arrived from a partner you pay commission to. That is why a program can run for a year and never produce a row called Affiliates.
The default channel group cannot be edited. You can tag the links so sessions match the rule, or build a custom channel group that gathers network hostnames under one name.
The second option is a patch rather than a fix. Grouping two network hostnames into a custom Affiliates channel makes the report readable, while every session underneath still carries no campaign data. You cannot break down by publisher, and you cannot join to commission, because the source is a hostname.
Tag the link at the network, not on your page
The parameters have to sit on the landing URL after the redirect. Most affiliate platforms let you set a destination template per program or per publisher, and most support a sub-ID that the network stores against the click and echoes back in the conversion postback.
A working template looks like this:
https://example.com/product?utm_source=network-a&utm_medium=affiliate&utm_campaign=partner-program&utm_content={subid}
Keep the medium fixed at affiliate so the channel rule fires. Put the publisher or the placement in utm_content rather than in the source, otherwise one program produces forty sources and none of them can be compared. If you need publisher and placement separately, use utm_content for the publisher and a second parameter for placement, registered as a custom dimension.
Then confirm it. Open the landing page in a fresh tab with the parameters in the URL, watch DebugView, and check that session_start carries campaign_source and campaign_medium. If your site sits behind a redirect that rebuilds the query string, test the real path instead of the direct one.
The sub-ID is the only reliable join key
Commission lives in the network. Revenue lives in GA4, or in your shop platform, or in BigQuery. The sub-ID is the one value both sides can hold, so treat it as a key rather than a dimension.
Two ways to use it. The cheap way: capture it on landing into an event-scoped custom dimension and attach it to the purchase event. That works for spot checks and shows up in the Events report. The durable way: send it with every event into the BigQuery export and join orders to commission rows in SQL.
Do not register it as a user-scoped dimension and then group reports by it. A click ID takes a new value on every session, which is the definition of high cardinality, and Analytics condenses the tail of a high-cardinality dimension into a row called (other). You end up with a report that cannot answer the question it was built for. Use it as a filter, or leave it in BigQuery.
Two clocks and one model
The network counts a conversion inside a window the advertiser sets, commonly 30 days, sometimes 60 or 90. GA4 counts a key event inside its own lookback, which defaults to 90 days for most events and 30 days for the acquisition events first_open and first_visit.
Attribution also differs in a way that catches people. GA4 credits events with last non-direct click. A visitor who clicks an affiliate link on Monday and returns by typing your address on Wednesday still has the Wednesday purchase credited to the affiliate source, because direct does not overwrite a non-direct source. Google documents that behaviour in its referral help using exactly that scenario.
So a returning visitor can produce a Direct session in one report and an affiliate attribution in another, and both are correct for the question they answer.
Why GA4 reads low against the invoice
Five mechanisms, all pointing the same way.
- Consent. When analytics_storage is denied, GA4 does not collect the event. The network's postback is server to server and needs no analytics consent, so the sale counts on their side.
- Cookie lifetime. Safari caps cookies written by script at seven days, which is shorter than most affiliate windows.
- Blockers. Privacy extensions and ad blockers drop the Google tag more often than they drop a network's own redirect.
- Cross-device. The network matches its own click ID through its own identity graph. GA4 matches a client ID in a cookie, or a user_id you supply.
- Lost storage. The visitor clears cookies, switches browser, or opens the purchase in a private window.
The practical result is that GA4 reads lower than the network in almost every program, and the gap widens as the window lengthens. Report the network figure as the commercial number and GA4 as the behavioural one. Chasing parity consumes a quarter and never lands.
Do not exclude the network domain
The instinct when a partner hostname appears as Referral is to add it to the unwanted referrals list. That is the wrong tool, and it costs you the only signal you have.
Three reasons. The exclusion is not retroactive, so past sessions stay where they are. It caps at 50 conditions per data stream, a small budget once payment providers, identity domains and webmail hosts are on the list. And after exclusion the next session keeps the previous source, because attribution is last non-direct click, so an affiliate arrival can end up credited to organic or to an old campaign.
The list is for domains that act as part of your own checkout: payment processors, identity providers, your own subdomains when self-referral detection misses them, webmail hosts used by password reset flows. Google's help page on unwanted referrals names the payment processor and the website-managed interaction cases directly.
Adding a condition appends ignore_referrer=true to matching events. Applied to the wrong page, that field removes the traffic source for every session that lands there, which is why Google warns against setting it across a whole site.
When you are the publisher
The other half of the market earns from affiliate links instead of paying them, and the measurement problem inverts.
Enhanced measurement records outbound clicks by default for every link that leaves your domain, with link_domain, link_url and an outbound flag on the event. Links to domains you configured for cross-domain measurement do not produce an outbound click, so a partner domain added to that list goes quiet.
If you work with several networks, break the outbound click event down by link_domain rather than by page. That answers which partner receives the traffic. Anything finer, clicks per placement or revenue per article, needs the network's report joined to yours, and the join key is again the sub-ID you placed in the link.
If your affiliate links run through a redirect on your own domain, expect that redirect to appear as a page view and to count toward engagement. It is a real page load, but it is not content, and a site with thousands of outbound redirects reports engagement figures that mean very little.
A reconciliation that survives a client meeting
Three columns, then one paragraph of explanation.
- Network report. Commissionable sales and revenue by publisher. This is the invoice.
- GA4. Sessions, key events and revenue from the Affiliates channel, plus whatever sits in Referral from known network hostnames.
- Consented subset. GA4 key events where analytics consent was granted, divided by the consent rate for the same period. Use it as a scale check, not as a target.
Then say the delta out loud. Consent and cookie lifetime explain most of it. What remains is usually a tagging fault, and a fault has a shape: one publisher reporting as Direct, or a whole network sitting in Referral because nobody set the template.
The afternoon checklist
- Ask each network for a destination template with utm_medium=affiliate and a sub-ID slot.
- Add the template at program level, then confirm one publisher's link inherits it.
- Test a real click in DebugView and check session_start for campaign_source and campaign_medium.
- Open Traffic acquisition and confirm an Affiliates row exists with the expected sessions.
- Capture the sub-ID on landing and send it with the purchase event.
- Pull 90 days from the network and from GA4, then write down the delta and its cause.
None of this repairs last quarter. Attribution already recorded is not rewritten when the template is fixed, so historical affiliate sessions stay wherever they landed. What changes is the next report, and the argument about which number is real turns into a list of known differences with a size attached to each one.
Running a partner program and cannot tie commission to sessions? Book a free tracking audit and we will check the template, the channel rule and the join key.