How to Set Stage Probabilities From Your Own Deal History
How to set stage probabilities: define each stage by what the buyer did, then ask how many deals that reached it closed. Your own history beats any template.
To set stage probabilities, define each stage by something the buyer did, then ask one question of your own history per stage: of the deals that reached this stage, how many eventually closed as won? The answer is that stage's probability. If you do not have the history yet, write down a conservative guess, label it as a guess and replace it once a quarter of real outcomes exists.
Everything else in this post is about doing that honestly.
How to set stage probabilities you can defend
A stage probability is a claim about a group of deals, not about any one of them. It says: deals that look like this, at this point, have historically closed at about this rate. That claim only holds if the word "stage" means something fixed. So the work has two steps, and the order matters. First fix what each stage means. Only then count.
Step one: define every stage by something the buyer did
A probability attached to a feeling is worthless. If a deal can enter a stage because the seller is hopeful, the probability on that stage will drift upwards until it describes nothing. So each stage needs an exit that is a fact, something that happened and can be checked.
The guide gives examples of the kind of fact it means:
- Discovery call held
- Decision maker engaged
- Proposal reviewed with the buyer
- Verbal agreement
- Paperwork in procurement
Your stages may be named differently. What matters is that each one passes a simple test: could a colleague look at the record and confirm that it happened? If the answer is no, the stage is a mood, and any number you attach to it will be too.
You also need two clearly defined end points: a won stage and a lost stage. Without them, you cannot count how many deals reached a given stage and went on to close, which is the only calculation this method needs.
Step two: ask one question per stage
For every stage, ask: of the deals that ever reached this stage, how many eventually closed as won? That ratio is the stage's probability.
The question is deliberately dull. It does not ask what the seller thinks, what the deal feels like or how the last call went. Individual judgement about individual deals tends to be optimistic, because the person closest to a deal is the person most attached to it. The stage history takes the mood out of the arithmetic. It will be wrong about any single deal and surprisingly accurate across a larger number of them.
| Source of the number | Who sets it | Reliability |
|---|---|---|
| Your own history, one question per stage | You, from the record | Best: it describes your business, and it can be corrected |
| A conservative written guess | You, labelled as a guess | Acceptable as a starting point, if you replace it |
| The seller's feeling about the deal | The person closest to it | Poor: reads as optimism |
| The software's default | A template | Poor: describes nobody's business |
Why not use the defaults that came with the software?
Tools often ship with tidy defaults that rise in even steps through the pipeline, ten, twenty-five, fifty, seventy-five, ninety. They look authoritative and they are no use to you, because they were not drawn from your deals. Percentages that came from a template drift further from reality each quarter they go unchecked, and nothing about them tells you they have drifted.
A probability you calibrated badly is still better than one you borrowed. Yours can be corrected, because you know where it came from and what to compare it against. A borrowed number gives you nothing to compare.
What to do if you have little or no history
Most founders reading this will not have a clean record of deals by stage. That is fine, and the method has a path for it.
- Set conservative guesses for each stage.
- Write down, somewhere visible, that they are guesses.
- Put a date next to them, the point by which you will replace them.
- When a quarter of real outcomes exists, replace each guess with the answer to the one question.
Be realistic about the early period. With a dozen open deals, a weighted pipeline is a polite fiction, and early probabilities will be set from thin history. That is not a reason to skip them. It is a reason to treat the first few quarters as calibration rather than prediction, and to name any large deal individually rather than letting a probability stand in for a judgement about it. The thinking behind that is in weighted pipeline versus commit.
Who should not set the number
Do not let the seller set the probability on their own deals. This is not about honesty. It is about position: the closer someone is to a deal, the more optimistic they will be about it, and stage history is duller and much more accurate. Let the seller sell, and let the record do the estimating.
It also helps to make the forecast safe to be honest in. If an inflated number is rewarded and a missed one is punished, the stage a deal sits in will be chosen to look good. Praise accurate calls, including accurate bad news, and the stage data stays clean enough to count.
What a stage probability can and cannot tell you
It helps to be clear about the limits, so that nobody asks the number to do more than it can.
- It cannot predict one deal. No deal closes for sixty percent of its value. It closes, or it does not. The probability describes how a group of similar deals behaves, not how this one will.
- It cannot see time. A probability says whether, not when. A deal that is real and qualified can still close next quarter instead of this one, and most forecasts miss on timing, not on loss. That is why close dates and age sit beside probabilities in the review.
- It cannot stand in for a large deal's detail. One large deal can outweigh everything else and make an average meaningless. Name it, and discuss it as itself.
- It is only as good as the stage definition. If entering a stage is a matter of enthusiasm, the history behind the number is a history of enthusiasm.
None of this makes the probabilities less useful. It tells you where to use them: across the whole open pipeline, as a way of asking whether you have enough for the coming quarter.
Warning signs that a stage has drifted
Stages rarely break all at once. They drift, one hopeful move at a time. These are the signals worth looking for in the weekly review:
| Signal | What it usually means |
|---|---|
| Deals jump into late stages without the factual step having happened | The stage is being entered on hope |
| Dates keep changing while stages stay put | The record is being edited to look steady |
| Almost nothing is ever marked lost | The pipeline only grows, so every stage looks better than it is |
| Probabilities have not changed in a year | They are not being compared with outcomes |
Each of these makes the history you count from less trustworthy. The remedy is the same in every case: restate the factual exit for the stage, correct the record, and recalibrate.
Keep the probabilities alive
Probabilities set once and never revisited are one of the most common ways a forecast decays. Two habits prevent that.
- Recalibrate every quarter. Compare what each stage was said to convert at with what actually converted from it, and adjust to match reality, even when reality is unflattering. The full comparison is in how to calibrate a sales forecast.
- Watch deal age. A deal that has been open for about twice the length of your typical sales cycle is no longer an average deal at its stage. Its stage probability no longer describes it, so it needs a decision: a concrete next step with a date, or close it out. See how to handle slipping deals.
One sign that this upkeep has lapsed is worth memorising: stage probabilities that have not changed in a year. A forecast whose numbers never move is not stable. It is not being compared with anything.
A quick checklist before you trust the numbers
| Check | If the answer is no |
|---|---|
| Is every stage defined by something the buyer did? | Redefine it before counting |
| Is there a clear won stage and lost stage? | Add them, or you cannot count conversions |
| Do I know where each probability came from? | Trace it, or relabel it as a guess |
| Are guesses dated for replacement? | Add a date |
| Did I compare last quarter's probabilities with what happened? | Do it now, and adjust |
Taken together, these steps are an afternoon's work, and after that the upkeep is small. If you want to see where the probabilities sit in the wider picture, the starting point is what a sales forecast is for.
Frequently asked questions
Can I use the default percentages while I get started?
You can, but be clear that they are borrowed and not yours. A better start is a conservative guess that you label as a guess and date for replacement, because that tells everyone the number is provisional.
How many deals do I need before the numbers mean anything?
The guide does not give a threshold. It says that with a dozen open deals the weighted pipeline is a polite fiction, so treat early quarters as calibration and rely on named deals for the near term.
Should different salespeople have different probabilities?
The method sets probability by stage, not by person or mood. Keep it that way, and keep the stage definitions factual so that everyone places a deal in the same stage for the same reason.
How often should I change them?
Review them every quarter, in writing, against what actually converted from each stage.
The full method, with the checklists, is in The Sales Forecast Playbook, a free 33-page playbook.