The traffic log
What is in the traffic log, what you can filter on, and the difference between the individual requests and the traffic overview that stays.
The traffic log is the part self-hosting normally leaves out: which requests arrived at your measurement address, and what your measurement server answered.
The columns
| Column | What it holds |
|---|---|
| Time | When the request arrived. |
| Address | Which measurement address, if you have more than one. |
| Method | GET or POST. |
| Path | The path that was requested, including its parameters. |
| Status | The HTTP status your measurement server returned. |
Under Inspection you see one request expanded.
Filtering
You can filter by address — or across all addresses at once — and by minimum status. That second one is the filter you will use most: set it to 400 and you keep only what went wrong.
A status of 200 means your measurement server accepted the request. It does not mean the measurement also arrived at Google or Meta; what leaves your server from there is decided in Tag Manager, not here.
Two things that are not the same
The individual requests are a window. The log rotates away once it grows too large — that is a limit on size, not on time, so a busy measurement server holds a shorter period than a quiet one.
The traffic overview is what stays. Once a day the log is condensed into one row per combination of address, event, browser, device and consent. That summary does not rotate away, and it is what you still see months later. Exactly how long depends on your plan; that is on the pricing page.
Put briefly: the individual requests are there to find a problem today, the traffic overview to answer a question three months from now.
Consent in the figures
Every request carries one of three values, read from what GA4 already puts in the query string. Nothing extra is collected from the visitor.
- Granted — the measurement was allowed to use storage, carries a client ID and belongs to a session.
- Denied — the measurement did arrive, but without a cookie. Which is exactly what consent mode is supposed to do.
- Unknown — this request says nothing readable about it. The library itself, a site that never implemented consent mode, or a parameter in a shape Google does not document.
Unknown is deliberately not a middle state between granted and denied. It means this cannot be determined, and the percentage on screen is calculated over the other two — so that “cannot tell” never quietly becomes a refusal.