Hi! In Fact Table Metric (Retention) is that real to measure retention not from experiment exposure but starting from custom event (like trial started or purchase made)? I didn't find Retention metric description in documentation or Ask-AI.

Hi Andrey, the default timestamp that the retention metric is built on is experiment exposure ("What % of users are retained X days after experiment exposure") e.g. you have any matching row in the event table at least X days after exposure. If you want to create a retention metric with respect to some timestamp other than the user's first exposure to the experiment, you need to do that manually in your SQL or create that in your data warehouse before hand. At some point we plan to let you pick a different event as the benchmark timestamp rather than just experiment exposure, as well as potentially making it easy to combine 3d, 7d, 14d retention of the same type in one metric. So those potential future improvements are why we started to split it out. Not currently in progress, but we have our eye on them if there's customer demand.

Thank you very much for clarification. Am I right that to create such metric I need to create Legacy Metric and write SQL manually for that?