Start where the leak is largest
Map the funnel end to end and quantify drop-off at each step. The biggest absolute loss — not the easiest test — is where you begin.
Qualitative first, quantitative second
Session recordings, on-site polls, and sales call transcripts surface objections that analytics never will. They generate hypotheses worth the traffic you have.
- Test big swings, not button colours — small effects need huge samples.
- Run tests for full business cycles to avoid weekday bias.
- Track downstream quality, not just form fills.
Make wins permanent
Every validated learning should end up in a design system, a copy guideline, or a brief template. Otherwise the same lesson gets relearned next quarter.
