Forecasting from the CRM you actually have: a method that survives messy data
Perfect pipeline data does not exist. A forecasting approach that uses stage history, conversion rates and deal age to produce a number leadership can plan on — and improves the data as a side effect.
18 Dec 2025 · 2 min read · by HeiSyn
Most forecasts are the sum of deal values multiplied by a probability someone typed in, and everyone knows it. A better forecast does not need better guesses from reps; it needs the CRM's own history — how long deals sit in each stage, how many move forward, how long they take to close — turned into rates and applied to the pipeline as it stands.
Build the conversion table
From the last four to eight quarters of closed deals: for each stage, the share of deals that reached it and went on to close, and the median time from that stage to close. Cut by segment if segments behave differently. This table is the forecast engine, and it comes from data reps already produced without trying.
Weight by stage and age
Each open deal's contribution is its value times the stage's historical close rate, discounted as the deal ages past that stage's median. A deal sitting in Proposal for three times the normal duration is not worth its stage rate; the discount says so without a rep having to admit it.
Three numbers, not one
Commit: deals with a signed verbal or a contract in redlines. Likely: the weighted pipeline for the period. Upside: what would close if every stalled deal revived. Present all three each week with the change from last week, and the forecast becomes a conversation about specific deals rather than a single contested number.
Let the forecast fix the data
Deals that the model discounts heavily are exactly the ones a manager should ask about, and the answer usually cleans the record: close it, move it back, or update the date. Over a quarter, the pipeline reflects reality more closely because the forecast kept pointing at where it did not.
Questions people ask next
- What if our history is too short for conversion rates?
- Use what you have and say so. Even two quarters of closed deals give usable stage rates for a first model; label the forecast as low-confidence and tighten it as data accrues. A rough model grounded in history still beats reps' probabilities.
- Should reps see the model's discount on their deals?
- Yes. When a rep sees that a deal stuck in proposal for two months is being counted at a fraction of its value, the conversation about whether it is real happens naturally. Hidden discounts breed distrust; visible ones breed cleaner pipelines.
- How often should the forecast be recomputed?
- Nightly, automatically, with the weekly snapshot saved so changes can be reviewed. A forecast computed by hand once a month is out of date by the meeting; a nightly one lets the weekly discussion focus on what moved.
Want this built for your team?
A free scoping call: we map your motion, name the number that should move, and quote a fixed-scope build.
