A forecast becomes reliable when every stage has an observable exit criterion, not when you refine the percentages. A deal only moves stage on a fact verifiable from the outside: a person met, an approval obtained, a date accepted by the customer. Weighting comes last, calibrated on your own history rather than on standard rates. And the forecast is separated from the pipeline review, otherwise nobody admits a deal is stalling.
In most companies we open up, the forecast is off by a third, sometimes by half, and everyone knows it. The usual reaction is to demand more rigour from the sales team, or to refine the probability percentages. Both answers miss the cause.
A wrong forecast is almost never a honesty problem. It is a definitions problem: as long as a deal changes stage on an impression, the number aggregates impressions.
The design flaw: stages that describe your activity
Look at the names of your sales stages. Nine times out of ten they describe what the salesperson did: first contact made, demo delivered, proposal sent, follow-up done.
Those stages measure your effort, not the customer’s intent. A proposal sent to someone with no budget who has never met their decision-maker sits at the same level as a proposal sent to a company that has already voted the spend. The forecast adds them together, and it is wrong.
The change that fixes most of the problem is to redefine each stage by what the customer did. Not “proposal sent” but “proposal received and reviewed with the decision-maker”. Not “demo delivered” but “the three people who must approve have seen the product”.
The exit criterion, and its test
Every stage needs a single written exit criterion, verifiable by someone who was not in the conversation.
The test is simple: if a colleague reads the deal record, can they say without calling the rep whether the stage has been cleared? “The customer seems interested” fails the test. “The finance director attended the meeting on the 12th” passes it.
Four criteria cover most complex cycles: the final decision-maker has been met, the problem has been quantified using the customer’s own data, the technical or security constraint has been cleared, and a go-live date has been accepted by the customer. Those are facts, they can be observed, and they are not negotiated in a meeting.
That work assumes data kept properly, failing which the criteria exist on paper and not in the tool, the subject of your revenue operations foundation.
Why weighting comes last
The temptation is to start there, because it is the visible part: assign 20 %, 50 %, 80 % probability per stage and multiply.
Laid over poorly defined stages, weighting produces false precision. It turns uncertainty into a number with two decimals, which makes it harder to contest: nobody argues with a calculated figure. You have made the error more credible without reducing it.
Once exit criteria are in place, weighting becomes useful, and it is computed from your own history. Take the last twelve months of deals and look, for each stage, at what percentage actually signed. The result almost always surprises, and it differs a lot from standard percentages, particularly on early stages, which are usually wildly overrated.
The signature date, the other half of the problem
A forecast has two unknowns: the amount and the date. Companies work on the first and ignore the second, although date slippage does more damage than losses.
The rule that helps most: the signature date does not belong to the salesperson, it belongs to the customer’s process. It is not inferred from an intention, it is inferred from a dated event on their side: a committee that meets, a budget year that starts, an existing contract that expires.
A deal whose signature date is “end of quarter” with no customer event justifying it is a deal without a date. The most honest treatment is to take it out of the quarter, and it is often the least popular decision of the year.
Separating forecast from pipeline review
This is the organisational point that unlocks the rest, and it costs nothing.
The forecast is an exercise in commitment: a number given to leadership, used to make decisions. The pipeline review is an exercise in mutual help: on this specific deal, what needs doing this week. The two cannot share a slot, because nobody admits a deal is stalling in the meeting where they are asked to commit to it.
So hold the forecast monthly and the pipeline review weekly, and say so explicitly to the team. We detail the second ritual in running a pipeline review that moves deals.
Measuring reliability, not just the number
A forecast is steered like any indicator: you have to measure its error. Three numbers are enough, taken each quarter.
The gap between committed and actual, as a percentage. The share of committed deals that actually signed in the committed quarter. And the share of signed revenue that was not in the forecast at all, which is the opposite symptom and just as troubling: a badly kept pipeline, where deals close without having been tracked.
Those three numbers across four quarters tell you whether your criteria are improving. Without them you correct blind, and you blame conditions for what belongs to your definitions.
Where to start this week
Take the ten deals currently at the most advanced stage of your pipeline, and ask one question of each: which observable fact justifies it being there?
In most teams, three or four do not survive the question. They are at that stage because the rep believes in them, not because the customer did something. Moving them back is unpleasant in the moment, and it is the only action that improves the reliability of your number immediately.
Frequently asked questions
- Why is a sales forecast always wrong?
- Almost never through dishonesty. Most often because the stages are defined by what the salesperson did, not by what the customer did. 'Proposal sent' describes your activity and says nothing about buying intent. As long as exit criteria describe your side of the desk, the forecast measures effort.
- What makes a good stage exit criterion?
- Facts observable from the outside, verifiable by someone who was not in the conversation: the final decision-maker has been met, a technical validation has passed, a budget has been confirmed in writing, a go-live date has been accepted by the customer. If the criterion starts with a verb describing your action, it is not a criterion.
- Should deals be weighted by probability?
- Yes, but last, and with your own historical rates rather than standard percentages. Weighting laid over poorly defined stages produces false precision: it turns uncertainty into a number with two decimals, which makes it harder to challenge.
- How often should you forecast?
- Once a month for the committed number, separately from the weekly pipeline review. Merging the two pushes salespeople to defend their deals instead of asking for help: nobody admits a deal is stalling in the meeting where they are asked to commit to it.