Our beta cohort decayed 91% in 8 weeks.

We invited 120 people to a closed beta on February 3. Here is the week-by-week active count, defined as "opened the app at least once in the past 7 days":

• Week 1: 120 (100%)

• Week 2: 94 (78%)

• Week 3: 71 (59%)

• Week 4: 52 (43%)

• Week 5: 38 (32%)

• Week 6: 24 (20%)

• Week 7: 15 (13%)

• Week 8: 11 (9%)

We surveyed the 11 who stayed and the 109 who left. The 11 who stayed all had one thing in common: they had imported data from another tool in their first session. The 109 who left mostly had not.

We had treated data import as an advanced feature, buried in settings. It was actually the activation event. We moved it to step 2 of onboarding and made it mandatory-ish (skippable but with a warning).

Next cohort of 100: week 8 active count was 41, not 9.

What is your real activation event — not the one you assumed, the one the data shows?