Funnel Analysis: Overall vs. Stage-to-Stage, Cohorts, Velocity

"Conversion rate" hides three questions. A Data PM keeps them apart.

1. Overall conversion — did it close?

won ÷ (won + lost) on resolved deals. Report it by created-month cohort, because recent cohorts haven't had time to close and will always look worse at the right edge of the chart. Excluding open deals is the first thing to do before believing any drop.

2. Stage-to-stage conversion — where do they drop?

Needs the append-only history, not the current stage. For each stage, count distinct deals that ever entered it; the ratio between consecutive stages is the drop-off. A deal that skipped a stage simply never appears in it.

3. Velocity — where do they stall?

Days between consecutive history rows for the same deal, i.e. LEAD(entered_at) OVER (PARTITION BY deal_id ORDER BY entered_at). A stage whose duration doubles is a stronger RCA signal than a rate that wobbles.

Using the three together (the RCA walk)

1. Confirm the drop is real on resolved, mature cohorts.
2. Slice overall conversion by one dimension at a time (region, source, rep, tier). A cause that touches everything (a price increase) can't explain a change in one cell.
3. In the affected cell, look at stage-to-stage and velocity to see where in the funnel the loss happens. Late-funnel losses with longer negotiation usually mean lead quality; early losses mean targeting.