You do not need a data warehouse to find where your discounting stops paying off. You need one export of closed opportunities and a willingness to look at it honestly.
The columns that matter
At minimum, one row per closed opportunity with:
- outcome — won or lost
- list value and booked value (annualized) — so we can compute the discount
- segment and close date — so we can see where and when the curve bends
Everything else (rep, approver, competitor present, product tier) makes the picture sharper, but the four above are enough to start.
The four moves
- Compute discount per deal as
1 − booked / list. This is price realization's mirror image. - Bucket into bands (0–5%, 5–10%, …) and compute the win rate in each band, with confidence intervals so you do not over-read thin buckets.
- Find where win rate stops climbing. The first band whose win rate the deeper bands never beat is your win point.
- Sum the discount given beyond it on won deals. That is your pricing upside to pursue — a prioritized list, not a refund.
The traps
- Don't trust thin bands. A 78% win rate from nine deals is noise. Confidence intervals keep you honest.
- Watch quarter-end. If late-quarter deals are systematically deeper, that is a process leak, not a value signal.
- Segment before you conclude. A blended curve can hide that Enterprise is fine and MidMarket is bleeding.
This is exactly the math Pricekeel runs for you — and it is deliberately the kind of thing you could check by hand, because a pricing recommendation you cannot explain is one finance will not trust.