# Sending Storyblok experiment data to GA4, or is the built-in reporting enough?

Asked by henrik_d on 2025-04-30. Tags: storyblok, ga4, analytics, ab-testing.

We run experiments on Storyblok content through Croct and all our funnels live
in GA4, where I spend most of my day as the growth analyst. I want experiment
results next to those funnels, but I do not want two sources of truth that
disagree with each other.

What exactly does the GA4 connection send over? And is there anything the built-in
dashboards do that GA4 fundamentally cannot, so I know which numbers to treat as
canonical when they differ

## 1 answer

### Accepted answer from Croct Bot (2025-04-30)

The two views serve different jobs, so the clean setup is: GA4 for funnel
context, built-in reporting for experiment decisions.

What the GA4 integration streams: the audiences a visitor fits, the experiences
that impacted them, the experiments they participated in, and the variants they
were assigned. It is a one-time, no-code project connection. Once flowing, you
can segment your existing funnels and retention reports by Croct audience or
variant, which is exactly the "results next to my funnels" view you want.

Why the numbers can differ, and which to trust: GA4 applies its own processing
rules. It drops events beyond 100k per user per day and groups high-cardinality
dimension values under "(other)". At normal traffic these rarely bite, but they
are the reason GA4 counts can drift from the source.

The [built-in reporting](https://docs.croct.com/reference/analytics/experiment/overview) has no such processing. It is unsampled, covering 100% of
the data, and the statistics are Bayesian, giving you probability-to-be-best per
variant rather than raw counts to interpret yourself. You can drill down by
device, audience, time window, and custom events. GA4 has no notion of the
experiment's statistical state, so the win/loss call should always come from the
built-in dashboards; GA4 is where you explore what the winning variant did to
downstream funnel steps.