Sales & PartnershipsFree
Sales Conversion Rate Analysis by Stage
Conversion by stage computed on deals that have actually resolved, ranked by how many you lose at each one rather than by percentage.
River writes a sheet, a chart and a short document. The sheet is conversion by stage, by segment and by rep, computed only on opportunities that have finished, won or lost, with deals lost and dollars lost sitting next to every rate. The chart is the funnel drawn per segment rather than for the whole business. The document names the one stage worth working on next quarter and the two changes to make there.
The ranking is by absolute loss, not by rate, and that single change usually moves the answer. A stage converting at 53 percent looks like an emergency until you count how many deals reach it. Rank by deals lost and a different stage wins; rank by dollars lost and a third one does. All three columns appear on the sheet, because the disagreement between them is the finding rather than a rounding problem, and which one a dashboard sorts on is usually an accident of whoever built it.
Written for the sales manager preparing a quarterly review, the revenue operations analyst who has been asked why win rate moved, and the founder looking at a funnel chart for the first time. Run it at quarter end with four quarters of history behind it, before anybody has committed next quarter's coaching time to a theory. The review itself is the deal review pack, what happens on the calls inside the worst stage is the call coaching pack, and rolling the resulting number upward is the forecast submission pack.
Why a conversion rate points the wrong way
A conversion rate is a percentage, and a percentage does not tell you what a fix is worth. Medical research makes the same point. The Cochrane Handbook notes that even if relative effects are similar across subgroups, absolute effects will differ according to baseline risk. In a funnel, that means the same percentage loss costs more where more deals arrive. Two stages losing a third of what enters them are not the same problem when one sees 705 deals and the other sees 96.
Adding segments together hides the reversal that matters. Bickel, Hammel and O'Connell examined Berkeley graduate admissions and found the combined figures showed a clear but misleading pattern that disappeared once each department was looked at on its own. A funnel behaves the same way, so every rate on the sheet is computed per segment as well as overall, and the two are reported side by side. Where they disagree, the segment figure names a change and the total hides it.
Take an illustrative company, Thornbury Systems: four quarters, 1,000 opportunities created and 820 finished. Security review converts worst at 53.1 percent, so it gets the attention. Discovery loses the most deals at 207. Proposal loses the most money at $6.44 million. Once losses that were never winnable come out, Proposal holds $6.44 million of addressable loss and security review holds $1.26 million, which ranks it fifth of six. A quarter of coaching time was heading to the stage the arithmetic puts fifth.
How it works
Paste your history
Opportunity history with the stages each deal entered and when, plus your stage order.
Set aside open deals
Deals still open are left out of every rate, with how many and from which stages reported.
Count the losses
Deals and dollars lost at each stage, sitting beside the conversion rate rather than instead of it.
Price the fix
What each stage is worth fixing, after the losses that were never winnable come out.
What you get
- Conversion by stage computed only on deals that have finished, whether won or lost
- Deals lost and dollars lost beside every rate, with the three rankings shown together
- The same funnel drawn per segment, so a pattern the totals hide shows up
- Losses split into the ones that were winnable and the ones that never were
- Conditional stages flagged, since a stage most deals skip cannot be compared to one they enter
- A ceiling in dollars on fixing each stage, so two candidate projects can be compared
Common questions
Our CRM already reports conversion by stage.
It reports a rate, and a rate on its own cannot tell you what a fix is worth. It also often counts open deals as if they had failed. In the illustrative example that combination understates late stages by nearly ten points and hands the team the stage where the least money sits.
Why leave open deals out?
Because they have not converted or failed yet, so counting them as failures measures how long deals take rather than how many convert. In the illustrative example, Negotiation reads 73.0 percent with its 48 open deals included and 84.8 percent without them, which is the difference between the second-worst stage in the business and the second best.
What counts as a loss that was never winnable?
One where no selling could have changed the outcome. A deal that entered Discovery with no confirmed budget and nobody named to sign, or a security review failed on a control the product does not have. Both belong on the sheet, and neither belongs in the number you are trying to move.
Is a stage most deals skip even comparable?
No, and it gets flagged rather than ranked alongside the rest. Security review in the illustrative example is enterprise only, so 96 deals reach it against 705 at Discovery. Comparing their percentages side by side is what produces the wrong answer, which is why deals entered sits next to every rate.
The segment cut contradicts our overall numbers.
That is the useful outcome, and it is common enough to have a name: Simpson's paradox. In the illustrative example, Proposal loses 31.5 percent overall, which hides mid-market at 45.4 percent against inbound SMB at 18.3 percent. One number describes a stage and the other tells you what to change. Losses per deal are the post-mortem.
How many quarters of history does this need?
Four is comfortable, two is workable, one is not. What matters is the number of finished deals rather than elapsed time: you need enough deals that have finished at each stage for the rate to mean anything, and late stages finish slowest. The sheet prints the deal count next to every figure.
What do we actually do with the winning stage?
Two changes, both specific. In the illustrative example, route inbound SMB out of Discovery until budget and a signer exist, and find what mid-market proposals lose on that SMB ones do not, which may well be price. That second question is the three-year comparison.
Sources
Built by
Chandler Supple
Co-founder & CEO, River
Chandler has spent the last six years working in AI, building tools for millions of users and Fortune 100 companies. In 2025, he co-founded River to build a new kind of operating system for AI agents.
Sales Conversion Rate Analysis by Stage
Describe what you need, add any files you have, and your workspace opens with the work already underway.