Hi team, I have a subscription-based app and I’m trying to measure the impact of experiments on subscription renewals. The problem is that users start their free trial during onboarding — before most feature flags are activated — so I can’t use “users who started a trial” as the denominator within any specific metric. I’ve discussed some options with support already, but none of them seem to solve the issue. Do you have any suggestions for how to approach this kind of setup? How can I measure the effect of experiments on renewals if trial start happens before feature exposure? Thanks in advance!

Normally people with this issue will use 'activation metrics'

Thanks, Graham, This is indeed a good solution if we’re only looking at metrics related to trial subscriptions. Let me give a bit more context, because I think your approach might not work for my experiment. We’re aiming to improve metrics for all new users, who can fall into one of three categories: free users, trial subscribers, and non-trial subscribers. The activation metric approach (e.g. taking a trial) would limit us to only trial users. We could trigger exposure on first app open, which would work without an activation metric. But we have a long onboarding flow (with a paywall inside), and we need to exclude users who don’t complete it — which requires an activation metric. The issue is that this activation point comes after a potential trial start. My suggestion is to allow configuring a lookback window that can reach before the exposure event — for example, in the denominator of the metric. I know there’s a lookback window setting in the current UI, but it doesn’t allow accessing events before exposure. As far as I remember, the legacy metric system used to support negative values for the conversion window, but that’s no longer available.