You call a number at the start of the quarter. The quarter ends 30% away from it, sometimes above, usually below. Leadership stops believing the forecast, so hiring, spending and cash planning all run on gut feel instead.
Sales forecast accuracy is fixable, but not by asking reps to try harder. Here is how to measure it properly, what a good number looks like, and the changes that actually improve it.
How to measure sales forecast accuracy
Use one formula and stick with it, so the number is comparable quarter to quarter:
Forecast accuracy = 1 − (|actual − forecast| ÷ actual)
If you forecast $500,000 and closed $400,000, accuracy is 1 − (100,000 ÷ 400,000) = 75%. The variance, the gap as a share of actual, is 25%.
- Healthy: variance within 10%.
- Watch: 10 to 20%.
- Critical: more than 20%, in either direction.
Measure the forecast you called at a fixed point, such as week two of the quarter, not the one you updated in the final week. A forecast that is only accurate on the last day isn't a forecast.
Why sales forecasts are inaccurate
Missed forecasts almost always trace back to what is in the pipeline, not to the forecasting method:
- Stages without exit rules, so "late stage" means different things to different reps.
- Stale deals that still count, because nothing removes them.
- Close dates set by hope. The rep picks the end of the month; the buyer never agreed to it.
- One person's judgement in place of a method, so accuracy depends on who is calling it.
- AI-written fields nobody checks. If your CRM updates close dates or stages from call transcripts automatically, the forecast inherits every misheard "maybe next month".
How to improve sales forecast accuracy: seven changes
- Write exit rules for every stage based on what the buyer has confirmed. Start with our sales pipeline stages template.
- Age deals out. No activity in 30 days moves a deal out of the forecast until it moves again.
- Make close dates buyer-confirmed. A close date the buyer hasn't agreed to is a guess, so label it as one.
- Weight by stage, using your own history. Multiply each deal by the close rate of its stage over the last two or three quarters.
- Split commit from best case. Commit means the buyer has said yes to terms; best case is everything else that could land.
- Review the same way every week, with the same view and the same questions, so changes are visible.
- Require approval for AI updates to forecast fields such as stage, amount and close date.
Is your forecast built on a clean pipeline?
Run the free Client Leak Check. Two minutes, no CRM access required.
Run The Client Leak Check →Track accuracy by rep and by stage
A company-wide accuracy number hides the cause. Track it by rep to see who calls high or low consistently, and check which stages the misses came from. If most misses were deals forecast from mid stages, the exit rules for those stages are too loose.
What changes when the forecast is fixed
With clients abroad, these changes took forecast variance from 34% to under 10% within 90 days, and forecast accuracy from 44% to 88%. Neither came from new software. They came from a pipeline where every deal had a real stage, a real date and a reason to be there.
A worked example: one quarter, before and after
Say a team calls $600,000 in week two. They have 40 open deals. Twelve haven't had a logged activity in over a month, and nine have close dates the buyer never confirmed. The quarter closes at $410,000. Accuracy: 1 − (190,000 ÷ 410,000) = 54%.
Now apply the rules before calling the number. The 12 stale deals leave the forecast. The nine unconfirmed dates move to best case. The remaining deals are weighted by each stage's historical close rate. The called number drops to $450,000. If the quarter still closes at $410,000, accuracy is 90%, and the variance is under 10%.
Nothing about the selling changed. The forecast simply stopped counting deals that were never going to close this quarter. That is why accuracy improves first by removing pipeline, not adding it.
Forecast accuracy vs forecast bias
Accuracy tells you how far off you were. Bias tells you which way. A team that is 85% accurate but always over-calls has a different problem from one that swings above and below.
- Consistently over: the pipeline is inflated. Look at stale deals and loose late-stage rules.
- Consistently under: reps are sandbagging, or deals close that were never in the CRM. Check where closed deals actually came from.
- Swinging both ways: there is no method yet. Fix the weekly review first.
Track both numbers each quarter. Bias usually points straight at the fix.
If you have the right amount of pipeline but still miss, read how to calculate pipeline coverage and why B2B sales forecasts keep missing.