How a referral exclusion works
When a visitor arrives on your site from another website, analytics reads where they came from and records it as a referral. If that arrival also looks like the start of a new visit, it can overwrite the source that brought the person in originally. A referral exclusion tells analytics to ignore a named domain when it does this, so a return from that domain does not start a fresh session with a new source attached.
In Google Analytics 4 the list lives in the admin area of the data stream, under the tag configuration settings, usually labelled as unwanted referrals. You add the domains you never want treated as a source. The traffic itself is not deleted and the visitor is not blocked; only the source attribution changes. Whatever brought the person to the site in the first place stays in place.
Why referral exclusions matter
The classic case is payment. A visitor clicks an ad, browses, then leaves your site for a payment provider to complete the transaction and comes back to a thank-you page. Without an exclusion, that return is credited to the payment provider, and your successful order is filed under a source that never sold anything. The campaign that actually earned the sale shows nothing.
This bites hard on Nepali sites, where checkout regularly hands off to a hosted gateway page before returning. It also applies to any booking engine, appointment system or login provider that lives on a separate web address. Once the credit lands on the wrong source, bidding decisions and budget decisions follow it.
Where referral exclusions go wrong
The commonest mistake is using the list as a cure rather than a bandage. If your own domain keeps appearing as a source, the real fault is usually a missing tag, a second measurement ID or an unconfigured cross-domain setup, and excluding the domain hides the symptom while the underlying self-referral problem carries on distorting sessions.
The second is excluding too much. Adding a partner site, a directory or a marketplace that genuinely sends you customers destroys real reporting, and it is very hard to notice afterwards because the traffic simply reappears as direct.
The third is forgetting that exclusions are not retroactive. Historic reports keep the credit they were given, so change the setting and note the date rather than expecting old data to correct itself.
What to do about it
Open your traffic acquisition report, sort by referral source and read the list honestly. Payment gateways, your own hostnames, login providers and internal tools belong on the exclusion list. Anything that is a real website with a real audience does not.
For payment and booking flows, add the exclusion and also configure the hand-off properly where the platform allows it, so identity survives the round trip. Then place a test order yourself and confirm the conversion is credited to the original channel rather than to the gateway. If several domains are involved, treat it as part of a wider cross-domain tracking job rather than a one-line fix, and record what you changed in your GA4 setup notes.