TIL our churn analysis was completely wrong because we were looking at monthly instead of weekly cohorts

Was talking to our data guy at lunch last week and he pointed out our churn numbers looked too good. Turns out we were grouping new signups by month which squished all the early dropoffs together. When he rebuilt it as weekly cohorts we saw the REAL picture - about 15% of people bail within the first 7 days. Month-level was hiding that completely. Has anyone else had a reporting metric totally misrepresent reality because of how you grouped the data?

3 comments

3 Comments

roberts.leo10d ago

Month-level was hiding that completely" - it's like how checking your weight once a month might show a smooth line down but hiding those weekends where you ate like garbage and bounced back up.

lunah1210d ago

Sure, but is that really a huge problem though? Like, so what if you had a bad weekend, the trend is still going down. Month-level data gives you the big picture, and that's what matters for long-term progress. Obsessing over every little daily bounce is a good way to drive yourself crazy and quit. And honestly, who actually weighs themselves daily and doesn't lose their mind over a totally normal water weight fluctuation? Seems like overthinking things to me.

reese_hayes719d ago

Wait, people actually do that? They only check their weight once a month and think that's enough to see what's really going on?