Open your CRM's opportunity stages and look at the probability next to each one. Discovery 10%. Qualified 20%. Solution 40%. Proposal 60%. Negotiation 80%. Verbal 90%.
Where did those numbers come from? In most companies, nobody knows. They were set when the CRM was implemented, often by copying a template, and they've never been checked. Yet, they drive the weighted pipeline number that shows up in forecast calls and board decks.
Checking them takes an afternoon and one SQL query, as long as you have one thing: a snapshot of the pipeline at a fixed point in time.
What You Need
For each of the last several quarters, you need the list of deals that were open on the first day of the quarter, with the stage each one was in that day. Most CRMs keep field history, and most warehouses that sync a CRM keep daily snapshots. If you have neither, start saving a snapshot today and come back in two quarters.
The first-day snapshot matters. If you look at deals in their current stage, you'll only see the deals that are still open or the stage they died in, which tells you nothing about how likely a deal was to close when it was at that stage.
Then join each snapshot deal to how it ended: won at any point, and won in the same quarter.
The Query
I simulated eight quarters of first-day snapshots, about 3,400 deal-quarters, with the usual 10/20/40/60/80/90 probabilities in the CRM and different true win rates underneath.
-- for every deal open on the first day of a quarter: what share were eventually won, and won that same quarter? select stage, any_value(crm_probability) as crm_says, count(*) as deals, round(avg(won_ever::int), 2) as won_eventually, round(avg(won_this_quarter::int), 2) as won_same_quarter, round(sum(amount * won_this_quarter::int) / sum(amount), 2) as won_same_quarter_by_value from snapshots group by stage order by stage; stage crm_says deals won_eventually won_same_quarter won_same_quarter_by_value 1-discovery 0.1 1129 0.07 0.02 0.01 2-qualified 0.2 882 0.16 0.05 0.05 3-solution 0.4 644 0.28 0.14 0.14 4-proposal 0.6 421 0.47 0.31 0.29 5-negotiation 0.8 236 0.57 0.47 0.47 6-verbal 0.9 94 0.84 0.80 0.76 Two separate problems show up.
The CRM's probabilities are too high even for eventual wins. A deal in Negotiation is won 57% of the time, not 80%. A deal in Solution is won 28% of the time, not 40%. This is common. Stage probabilities are usually set by optimists, and they're rarely lowered.
The forecast needs a different number entirely. A stage probability answers "will this deal ever close?" A quarterly forecast needs "will this deal close this quarter?" Those differ a lot for early stages. A Solution-stage deal on the first day of the quarter has a 28% chance of eventually closing, but only a 14% chance of closing before the quarter ends. Early-stage deals take time.
The last column weights by deal value. If it differs a lot from the deal count version, big deals behave differently from small ones, and you should calibrate by size band too.
Does it help?
The real test is whether the measured rates forecast better than the CRM's. I learned the rates from the first four quarters and tested them on the last four, which is the honest way to check. Testing on the same quarters you learned from will always look good.
-- weighted pipeline with the CRM's probabilities vs the measured ones, against what actually closed with rates as ( select stage, sum(amount * won_this_quarter::int) / sum(amount) as measured from snapshots where quarter <= 4 -- learn from the first four quarters group by stage ) select s.quarter, round(sum(s.amount * s.crm_probability) / 1e6, 2) as crm_weighted_m, round(sum(s.amount * r.measured) / 1e6, 2) as measured_weighted_m, round(sum(s.amount * s.won_this_quarter::int) / 1e6, 2) as actually_closed_m from snapshots s join rates r using (stage) where s.quarter > 4 -- and test on the last four group by s.quarter order by s.quarter; quarter crm_weighted_m measured_weighted_m actually_closed_m 5 4.60 1.82 2.01 6 5.19 2.11 2.37 7 4.35 1.71 1.91 8 4.61 1.85 1.98 The CRM-weighted pipeline says about $4.6 million every quarter. About $2 million actually closes. The CRM number is more than twice the truth. The measured rates come in about 10% low, close enough to be useful, and the error is consistent enough that you could see it and adjust after a quarter or two.
A weighted pipeline that's off by more than 2x isn't just inaccurate. It's actively harmful, because people make hiring and spending decisions based on it.
Things that will trip you up.
Stage skipping. If reps often jump deals straight to Proposal, the deals in Proposal on the first day are a mix of real proposals and deals that never had discovery. That's fine for calibration. You're measuring what Proposal means in practice, not in the sales playbook.
Re-opened and recycled deals. If lost deals get re-opened later, decide whether a re-open is the same deal or a new one, and be consistent.
Changing processes. If the stage definitions changed a year ago, only use quarters after the change. Fewer quarters of the right data beat more quarters of the wrong data.
Small numbers. With 94 deals in Verbal across eight quarters, the 84% is reasonably solid. If a stage has 20 deals, the measured rate could easily be off by 15 points. Pool more quarters, or merge thin stages.
What to change
Update the CRM probabilities to the eventual win rates. They're used in lots of places, and the eventual rate is what the field implies.
Use the same-quarter rates for the quarterly forecast. Keep them in a small table in the warehouse, not in the CRM, and recompute them every quarter from the latest eight quarters.
Show both numbers in the forecast meeting. "Weighted pipeline at the CRM's probabilities is $4.6 million. At measured rates, it's $1.8 million." The first time the meeting sees that, the conversation changes.
Recalibrate on a schedule. Win rates drift as products, markets, and teams change. Put a calendar reminder at the start of each quarter.
None of this needs a forecasting tool. It needs a pipeline snapshot, one query, and the willingness to replace a number everybody's used to with one that's right.
I'm Malek Damiri, head of revenue operations in Atlanta. I write about pipeline, forecasting, and the numbers behind them at malekdamiri.com.