How to Build a Sales Forecast From CRM Data

Your CRM holds everything a forecast needs. It also holds a large amount of material that will wreck one. The build is mostly a series of decisions about which fields carry signal, which fields carry rep opinion, and how to convert closed history into rates you can apply to deals that are still open.

What CRM data do you actually need?

Five extracts: open opportunities, twelve months of closed opportunities, account attributes, stage change history, and activity timestamps.

Pull open opportunities with amount, stage, close date, owner, and create date. Pull closed opportunities with the same fields plus the final outcome and the actual close date. Join both to account records so every deal carries a segment, because enterprise and SMB behavior share nothing except a column header.

Stage change history is the extract teams skip, and it is the one that makes the model work. Without it you cannot measure how long deals sit in each stage or how often they move backward. Activity timestamps do a different job. They tell you whether a deal is alive.

Does messy CRM data disqualify you from forecasting?

No. Consistent bad data still predicts. Inconsistent clean data does not.

Every revenue leader believes their data is uniquely bad and that this is the reason they cannot run the business the way they want. It is a myth. Everyone has bad data, and garbage in does not have to mean garbage out, as long as the garbage is consistent.

If your reps systematically skip stage three, the model learns that stage two leads directly to stage four. If half your reps skip it and half do not, and the split changes when territories change, the model learns nothing. Spend your cleanup budget on consistency across teams and time periods rather than on perfecting individual records.

Which CRM fields should you distrust?

Stage probability first, then the rep-entered close date. Amount deserves particular attention. Pipeline amounts run above what deals actually close for. For example, an average deal size of $80,000 in open pipeline against $40,000 in closed-won. If you forecast on pipeline amount without applying a realization ratio from your own history, you will overstate every period.

How do you turn closed history into conversion rates?

Group opportunities by the stage they reached, then measure the share of each group that ended in closed-won.

Calculate rates by segment, not blended. A single blended win rate hides the fact that your enterprise motion converts slowly at high value while your SMB motion converts quickly at low value. Blending them produces a number that describes neither.

Those figures are an illustration of the calculation, not a benchmark. Run it on your own closed set and the shape will differ. Watch for stages where the conversion rate barely moves from the stage before it. That gap tells you the stage has no exit criteria and is doing no work in your process. More on the mechanics of this build sits in our guide to how to create a sales forecast.

How do you assemble the forecast from those rates?

Apply segment rates to open pipeline, subtract dormant deals, then add the revenue that will be created and closed inside the period.

Start with open opportunities carrying an in-period close date. Apply the conversion rate for the stage each deal currently occupies, and multiply by your realization ratio so the amount reflects what similar deals actually closed for.

Then cut the dead weight. Any opportunity with no change to stage, close date, or amount in twelve months should come out of the number regardless of the stage it sits in. Across ORM's customer base, more than 10 percent of open pipeline meets that description. A change to stage, close date, or amount counts as meaningful activity. Nothing else does.

The last layer is the one most CRM-driven builds omit. Some share of the revenue in any period comes from opportunities that do not exist in CRM yet. Measure how much revenue was created and closed inside the same period historically, and carry that forward as its own line. Skipping it understates every quarter for teams with short cycles. Definitions for the terms in this build are in the sales forecasting glossary entry.

How do you keep the model honest over time?

Recalculate rates monthly on a rolling window and compare the forecast to actuals by segment every period.

A rate library built once decays. Win rates move when a competitor enters and applies pricing pressure. Cycle length stretches when buyers hesitate. The model has to pick up those shifts quickly, which means the underlying rates need to refresh on a schedule rather than during an annual planning exercise.

Track the variance by segment rather than in total. A company-level forecast that lands within a point of plan can hide an enterprise segment that came in 20 percent short and an SMB segment that overdelivered by the same amount. Measuring forecast accuracy at the segment level tells you which part of the model needs attention.

Reps who consistently change close dates deserve a flag in the same review. A close date change is the strongest single indicator that a deal is slipping, and it is visible in CRM the moment it happens.