A sales pipeline is just a funnel with feelings attached. Strip the feelings out and you can see exactly which stage is quietly bleeding revenue.
I took a full export of 4,200 opportunities and rebuilt the funnel stage by stage — entry, qualification, demo, proposal, close — then asked the only two questions that matter: where do deals die, and what do the survivors have in common?
The raw CRM export was messy — duplicate opportunities, stages skipped, timestamps missing. I modelled it in Snowflake with dbt into one clean row per opportunity, with the full stage history and time-in-stage attached, then surfaced it in Looker.
With a clean stage history, the funnel became a simple set of conversion ratios — and the time-in-stage data let me test behaviours, not just outcomes. The standout test: split every deal by how quickly its first demo happened, then compare win rates.
Insight one: the funnel leaks hardest at the demo. Half of qualified deals never get one booked.
“The deals that won didn't have bigger budgets. They had earlier demos.”
Insight two: speed compounds. The faster the first demo, the higher the close rate — and it isn't close.
- 01Wire an alert that flags any qualified deal without a demo booked inside five days — close the leak in real time.
- 02Study the top two reps' sequences to see what's coachable versus what's just talent.
- 03Model expected revenue per stage so forecasting stops being a gut-feel spreadsheet.