How to Calibrate a Sales Forecast Every Quarter
How to calibrate a sales forecast: compare commit, stage probabilities and close dates with what happened each quarter, then adjust the numbers to fit.
To calibrate a sales forecast, compare it with what happened at the end of every quarter, in writing, using three comparisons: commit against actual closed, each stage's probability against what actually converted, and how many deals closed in the period their date said. Then adjust the probabilities to match reality, even when reality is unflattering.
A forecast that is never compared with outcomes never improves. It just accumulates confidence.
How to calibrate a sales forecast every quarter
Calibration is the step that turns a forecast from a guess into an instrument. The idea is the one at the heart of the method: a forecast does not have to be right, it has to be wrong in a direction you know, by an amount you can measure. You cannot know the direction or the amount until you have looked at the outcome and written down the gap.
The job is small. The guide describes it as an honest afternoon a quarter. It is cheap precisely because the data already exists: the forecast you made, the deals that closed, and the dates and stages in between.
The three comparisons
Three comparisons are enough. Together they are your forecast's track record, and they tell you how much to trust next quarter's.
| Comparison | The question | What it tells you |
|---|---|---|
| Commit against actual closed | Over or under, and by how much? | Whether your commit criteria are too loose or too tight |
| Stage probability against conversion | Did each stage convert at the rate it claimed? | Which probabilities need changing |
| Close-date accuracy | How many deals closed in the period their date originally said? | How far to trust dates, and how much slippage to allow |
Write the answers down. A comparison held in somebody's head is a feeling about the comparison.
Adjust the probabilities, even when it hurts
After the second comparison, change the numbers. If a stage was said to convert at one rate and converted at another, replace the claim with the observed rate. The guide is firm that this applies even when reality is unflattering, and that is where the temptation is: a lower probability makes the pipeline look worse and the plan harder. But a probability you leave high to protect a number is a probability that will mislead every decision made from it. Where the numbers came from in the first place is explained in how to set stage probabilities, and the quarterly check is how guesses become history.
Triangulate with three independent estimates
No single method deserves blind trust, so the method asks you to use three that fail differently. Each can reveal a flaw in the others.
- The trend. What you have actually closed in recent months, projected forward.
- The weighted pipeline. What the open deals are worth on calibrated probabilities.
- The commit. What the people closest to the deals will stand behind.
When the three roughly agree, you can act with some confidence. When they diverge, the divergence is the finding.
| What you see | What to ask |
|---|---|
| Commit well above trend | What is different this time? |
| Weighted pipeline far above commit | Are there stale deals or optimistic stages? |
| Trend well above both | Is the pipeline under-recorded? |
The difference between the first two figures is explained in weighted pipeline versus commit. The key habit is to put them next to each other every time, and to treat disagreement as the thing worth discussing.
Count the lost reasons
Calibration is also the time for the quarterly count of lost reasons. The spread of reasons shows where the company is weak, and the largest category goes to whoever owns that fix. The method is in how to record why deals are lost. Make sure that no decision is examined as a message problem, not dismissed as bad luck.
Check coverage for next quarter
Calibration looks backwards, but one question looks forward: is there enough pipeline entering for the following period? If there is a gap, make it somebody's job today. This is the same early-warning logic as in how to set a pipeline coverage target, and the derived ratio should itself be re-derived if your win rate has moved.
Reading each comparison in practice
The three comparisons are simple to state and easy to do badly, so it helps to be specific about what each one needs.
Commit against actual closed
List the deals that were in commit at the start of the period. Next to them, list what actually closed. The gap is the number, and its direction matters as much as its size. A commit that runs over by a similar amount each quarter is a measurable bias you can discount. A commit that swings in both directions is the real problem, because nothing can be steered by it.
Stage probability against conversion
For each stage, take the deals that reached it and ask how many went on to be won. Set that beside the probability you were using. Stages with a large difference are the ones where the definition may also be loose, so check whether entry was factual or hopeful.
Close-date accuracy
Take the deals that closed and ask whether they closed in the period their original date said. A low share means timing is where your forecast is losing its accuracy, which points you back to the slip rule and to taking dates from the buyer, as covered in how to handle slipping deals.
Why the track record is itself a result
Lacking a record, you cannot judge how far to rely on the number. Both extremes cost money. Trusting it too much means a hire or a budget that the cash cannot support. Trusting it not at all means the forecast is quietly replaced by gut feel, and the company loses the one instrument it had for turning evidence into decisions.
A written record changes the conversation. Instead of arguing about whether the forecast is any good, you can say how it did last quarter, in which direction it erred and by roughly how much, and let the next number be read with that in mind.
Signs that calibration is overdue
You can tell from the forecast itself when it has not been compared with reality for too long.
- The percentages on your stages are a year old.
- Commit lands exactly on target every quarter, which usually means deals that qualify are being held back.
- Wins cluster in the final days of each period.
- Almost nothing is ever marked lost, so the pipeline only grows.
- Dates keep shifting while stages stay where they were.
Each of these shows a figure that has come loose from its deals. Calibration is the remedy.
A quarterly checklist
| Step | Done when |
|---|---|
| Compare commit with actual closed | The gap is written down |
| Compare each stage's probability with conversion | You know which stages were off |
| Check close-date accuracy | You know what share of deals closed in the period their date said |
| Adjust the probabilities | The numbers match what happened |
| Compare trend, weighted pipeline and commit | Any divergence is explained |
| Count lost reasons | The largest category has an owner |
| Check next quarter's coverage | Any gap is someone's job today |
What good calibration looks like over time
The early quarters will be humbling. Probabilities start out as conservative guesses, and the first comparison will show how far off they were. That is expected, and the right response is to treat the first few quarters as calibration rather than prediction.
Over time, two things happen. The gap between commit and actual closed narrows, or at least becomes consistent, which is nearly as good, because a forecast that is reliably a little optimistic is one you can discount and steer by. And decisions begin to change because of the forecast. The measure of the whole exercise is the quarter after the review, when a decision went differently because of it.
Making the review safe to be honest in
Calibration produces numbers about how wrong people were, so it can feel like blame. Keep it about the method. If the commit missed, the question is whether the criteria were right, not who was careless. Praise accurate calls, including accurate bad news, and people will report what they believe next time.
Frequently asked questions
How often should I calibrate?
At the end of each quarter, in writing. The weekly review keeps the forecast current, and the quarterly comparison keeps it honest.
What if I do not have a quarter of history yet?
Start with conservative guesses, label them as guesses and date them for replacement. Replace them as soon as a quarter of real outcomes exists.
What if my forecast is consistently high?
Then it is a useful instrument. A forecast that is reliably a little optimistic can be discounted and steered by, provided you know by how much. The danger is a forecast that is wrong unpredictably.
Do I need special software?
No. Everything described here can be done with a spreadsheet and a disciplined half hour a week, plus an honest afternoon each quarter.
The full method, with the checklists, is in The Sales Forecast Playbook, a free 33-page playbook.