Declaring it in a meeting
A workshop produces a confident sentence about what customers must experience. It may even be right. Without the cohort comparison and the intervention, there is no way to know, and the resulting onboarding is built on a guess that is now hard to question because senior people said it.
Importing another company's number
Published thresholds from well-known products describe those products. Their method is transferable and their conclusion is not. A specific count of connections, uploads, or actions is the answer to somebody else's question.
Mistaking a symptom for a cause
The commonest analytical error. Committed users perform more actions of every kind, so almost any early action will correlate with retention. Only the intervention separates the actions that create value from the actions that mark people who were staying anyway.
Demonstrating instead of delivering
A guided tour with sample data shows the product working. It does not show the user their own problem being solved, which is the entire content of a real realisation.
Assuming a single moment for everyone
An owner running a small team and a manager overseeing eight reps are buying different things. One aha moment for both usually means serving neither well.