We defined all our audiences in GA4, how do those map to Croct audiences?
We are a B2B SaaS and over the past three years we have built up a fairly large library of GA4 audience definitions (returning visitors, pricing page viewers, multi-session evaluators, and so on). We are rolling out Croct for on-site personalization this quarter.
My checklist questions:
- Is there any way to import GA4 audience definitions, or do we rebuild them?
- If we rebuild, how hard is the translation for typical behavioral definitions?
- What is the practical difference between an audience in an analytics tool and one used for on-site personalization? I want to set expectations correctly with the team.
2 answers
There is no importer, and the rebuild is the smaller part of the work. The bigger shift is architectural, which answers your third question.
A GA4 audience is a batch-computed list: users accumulate into it after processing. A Croct audience is a CQL condition evaluated in real time on every interaction, as explained in the audience overview. Nobody is "in" an audience ahead of time; the condition either holds for this visitor right now or it does not. The practical consequence is that the audience applies on the same visit where the behavior happens, rather than after a processing delay.
On translation: most GA4 behavioral definitions map to short CQL expressions. A few examples:
user is returninguser has viewed a page with path "/pricing"session's stats' pageviews > 3campaign's source is "google"You have 100+ signals to work with across Events, Location, Marketing, Session, Shopping, User, Web, and Technology, plus custom attributes, so the typical definition takes a minute or two to rebuild. The post on how Croct differs from analytics platforms covers the analytics-audience versus real-time-audience distinction in more depth.
We did this exact migration last year. One tip: do not rebuild everything. Years of GA4 audiences usually include a long tail nobody uses. We listed the audiences actually referenced in reports or campaigns and rebuilt those, which turned out to be about a quarter of the library. Also worth knowing that the audience estimator in Croct shows you the size of a condition as you write it, which made it easy to check that our translations captured a similar population to the GA4 originals.