Crystal W

I know if a team has good instrumentation just by looking at the spec. Bad sign: tons of events, each with zero or one property tracked. You're missing all the context that explains *why* users do anything.

When users churn because they 'have too much product,' a pause button beats a cancellation flow every time. Canceling is a permanent solution to a temporary problem. Match the fix to the actual constraint.

For a free product, week-1 retention should be at least 60% — then flatten. That's what we saw at Gojek in the early days. If you're still with friends and family, it should be close to 80%.

The step right before conversion is usually where retention problems live — not in the product itself. Fix the setup moment, not just the conversion moment.

When conversion is poor and users haven't experienced the product yet, it's almost certainly a copy problem. Change the copy to reflect the pain, not the solution — so users can see themselves in the use case.

Run experiments even with 30 users. The underlying trends you get won't change much at larger scale — only the precision does. Zero data points is always worse than 30.