Self-Referrals: Why Your Own Domain Is in the Source Report
Your own site is listed as a traffic source, sometimes near the top. It looks absurd, and it is describing something real.
A self-referral means a session ended and a new one started while the same person was still on your site, and the new session recorded the previous page as where it came from. The visitor never left. The measurement did.
That matters more than the odd row in a report, because every self-referral is an original source that got overwritten, and the campaign that actually brought that person is no longer credited with what they did next.
Why it happens
A session is tied to a domain, or to a set of domains the measurement setup has been told to treat as one. Cross a boundary the setup does not recognize, and the visit is treated as a new arrival from wherever it came, which is you.
Four situations produce that boundary.
1. A payment or checkout process on another domain
The most common and the most expensive. The visitor goes to a payment provider, completes the transaction, and returns. If the payment domain and your domain are not linked in the measurement setup, the return leg is a new session referred by the payment provider, and the purchase is credited to it rather than to the campaign that brought the person.
This is why a checkout provider frequently appears as a top converting source in accounts nobody has configured for it, and it is a serious distortion rather than a cosmetic one.
2. A subdomain or a second domain you also own
A booking system, a help center, a shop, a landing page tool, a careers site. Anything on a different hostname is a different domain unless the setup says otherwise.
3. Internal links tagged with campaign parameters
Putting campaign parameters on a link from one page of your site to another page of your site starts a new session and overwrites the original source with your own tag.
This one is entirely self-inflicted and it is the easiest to fix, because the fix is to stop doing it. It usually appears on internal promotional banners and on links inside emails that point to a second page of the same site.
4. A redirect or an intermediate hop
Any hop through another host, including a link shortener or a click tracker, is a boundary.
And a related cause worth ruling out separately: if paid traffic is arriving with its identifier stripped by a redirect, the visits land unattributed rather than as self-referrals, which is a different symptom with a similar underlying shape. That is its own subject on this site.
Why it costs you more than a strange row
The first session's source is gone from the second session. So the campaign that brought someone in gets credited with the visit and not with the purchase, and the purchase is credited to your own domain, a payment provider, or an internal banner.
The effect is systematic rather than random. It concentrates on the journeys that go through a checkout, which are exactly the journeys that convert. So the distortion falls hardest on the numbers that decide budget.
And session counts inflate, because one visit is being recorded as two or three.
If your channel numbers do not agree with each other and you are trying to work out which to believe, this is one of the two or three usual causes, and our article on what direct traffic actually is covers the other main destination for sessions whose origin was lost.
Finding yours
Look at your referral report and search for your own domain, including every subdomain and every variant.
Then look for domains you recognize as part of your own process: payment providers, booking tools, help systems, anything a visitor passes through on the way to converting.
Anything on that list that is receiving conversion credit is the priority, because that is where the misattribution is doing real damage.
The fix, described without a menu path
There are two settings involved and both have been renamed and relocated more than once, so this article deliberately does not give you a path.
The first tells the measurement which domains to treat as one property, so that crossing between them does not break the session.
The second tells it which referrers to ignore, so that a return leg from an external service does not overwrite the source.
Both exist in every serious measurement platform, and your own platform's current documentation is where to get their present names and locations. That is not a dodge: naming a setting from memory in a category that renames things quarterly is how an article becomes confidently wrong.
Three practical notes that do not depend on the interface:
Include every variant. The www and non-www forms, every subdomain, and any second domain.
Configuration applies going forward. Historical data is not corrected, so note the date you changed it and read reports across that boundary with the change in mind.
And check afterwards. Give it a few days and look at the referral report again. If your own domain is still there, something is not covered.
The one you fix without any setting
Remove campaign parameters from internal links.
No configuration will save you from this one, because the tagged link is deliberately telling the measurement that a new campaign visit has started. It is doing exactly what you asked.
The rule is simple and it is worth telling everybody who builds links: campaign parameters go on links that bring people to the site from outside, and never on links between pages of the site. That belongs in whatever naming convention your team follows, and it is the single line in it that prevents the most damage.
The order to work in
1. Stop tagging internal links. Free, immediate, and it removes a cause rather than masking it.
2. Configure the domains that belong to you as one property.
3. Configure the external services in your conversion path to be ignored as referrers.
4. Note the date, and check the referral report a week later.
5. Then re-read your channel numbers, because they will have changed, and the change is a correction rather than a performance movement. Saying that in the report is what stops somebody reading a fixed measurement as a good month. Our article on what wasted spend looks like in an account covers the neighboring discipline: a number that improved because you changed how it was counted has not improved.