When tagging breaks for a few hours, the events that should have fired during that window are gone forever. The events that fire after the outage will be misattributed, because GA4 has no way to know that those users had earlier touchpoints. There is no fix for the lost data, but there are steps to limit the bad downstream attribution.

Diagnose the scope first

Before doing anything, figure out: what was broken (loader, specific tag, all tags), how long (15 minutes, 12 hours, two days), and what fraction of your traffic was affected (regional outage, all of it).

A 15-minute outage during a low-traffic window is barely worth mentioning. A 12-hour outage during a campaign launch is a real incident with downstream consequences.

Annotate the gap in GA4

GA4 has annotations on Realtime and Acquisition reports. Add one for the outage window with a description ("tagging server outage 2026-04-15 14:00-18:00 UTC"). Anyone looking at reports for that window in the future sees the annotation and knows to discount the data.

Annotations are visible across the whole property, so the entire team sees them. Skip them and someone will eventually use the affected data for a decision.

Backfill from the warehouse if you have one

If you wrote events to BigQuery in parallel with GA4, the warehouse may still have the data even though GA4 missed it. Replay these events through the GA4 measurement protocol with the original timestamps. GA4 will accept events up to 72 hours old.

For events older than 72 hours, the GA4 measurement protocol rejects them. Backfill is no longer possible; only annotation remains.

Reset attribution windows

In GA4 Admin > Attribution Settings, the default attribution model is Data-Driven. After a meaningful outage, the model has skewed inputs. There is no manual reset; GA4's model adapts over time as new data flows.

For the first few weeks after the outage, expect attribution reports to show patterns that do not reflect reality. Communicate this to stakeholders so they do not make budget decisions on the noisy data.

Prevent the next one

An outage is the prompt to add the monitoring you should have had. At minimum: an uptime check on your tagging URL, an alert on event volume drops in GA4, and a runbook for the on-call person who gets the alert. Take the post-outage hour to set this up; you will not have time later.

For ongoing observability, the tag audit habit catches drift before it becomes an incident.