Gartner found that fewer than half of sales leaders and sellers have high confidence in their organisation's forecasting accuracy. That statistic gets quoted as a warning. It is more useful read as a description: most forecasts are built in a way that reliably produces a wrong number.
The four reasons forecasts miss
1. The stages do not mean anything
A forecast is arithmetic performed on stage data. If two reps place comparable deals in different stages, the arithmetic is operating on noise. This is the root cause behind most of the others, and it is fixable in a week by writing exit criteria that describe buyer behaviour.
2. Close dates are aspirations
Ask a rep when a deal will close and you get the date they hope for. Ask what the buyer committed to and you often get silence, which is the actual finding.
The test is simple: does the close date correspond to something the buyer said, or to the end of your quarter? Pipelines where a suspicious number of deals close on the final day of the quarter are showing you an artefact of the process rather than a pattern of buyer behaviour.
3. Nobody removes anything
Deals enter pipelines easily and leave reluctantly. Marking something closed-lost feels like an admission, so deals get pushed instead: one more quarter, one more follow-up. Within a year a meaningful share of the pipeline is deals that ended months ago and were never buried.
4. The incentive points the wrong way
If a rep is criticised for a deal that slips, they will start calling deals late, and you will lose the early warning that made the forecast useful. If pipeline volume is what gets praised in board meetings, the pipeline will grow and the conversion rate will quietly fall.
Whatever you reward in the forecast review is what the forecast will optimise for, which is worth thinking about before designing the review.
Why probability percentages make it worse
Most CRMs ship with a probability field attached to each stage. Discovery is 20%, proposal is 60%, and the forecast is the sum of value multiplied by probability.
The problem is that an individual deal has no 60% outcome. It closes for $100k or it closes for nothing. Weighted arithmetic is only meaningful across a large sample, and a team running forty deals a quarter does not have one. What the weighting actually does is produce a number between two possible realities, which then gets reported as though it were a prediction.
It also hides the interesting question. "We have $840k weighted" invites no follow-up. "Eleven deals are committed, worth $410k, and here is what has to happen on each" invites exactly the conversation the review exists for.
Commit, best case, pipeline
Three categories, defined by evidence rather than by feeling.
| Category | What it means | Test |
|---|---|---|
| Commit | You would bet your quarter on it | The buyer has confirmed a signature date and the steps before it. Nothing outstanding could move that date. |
| Best case | Plausible, not yet evidenced | Active buying process, clear next step, one or two open questions that could still go either way. |
| Pipeline | Real, but not this quarter | Everything else. Counted for coverage, excluded from the call. |
The number you commit to is the commit column. Best case is the upside you describe separately. Coverage across the whole pipeline is a different measure again, covered in what 3x actually means.
The weekly process
Thirty minutes, same time each week, same four questions per deal in commit and best case. The consistency is what makes it work, more than the content of any single session.
- What changed since last week? If the answer is nothing, that is the most important thing said in the meeting.
- What did the buyer do, specifically? Seller activity does not count as movement.
- What has to be true for this to close, and by when? Anything vague here is a deal that belongs in best case.
- What would make you move it out? Naming the kill condition in advance makes it far easier to act on later.
Two rules keep it honest. Deals can move down as easily as up, and nobody is punished for moving one down: the punishment belongs to the person who leaves a dead deal in commit until week eleven. And every deal in commit gets a named next step with a date, because a commit without one is a hope with a category label.
What to tell the board
When you first apply real criteria, the forecast will drop. Often materially, because some of what was committed was never evidenced. That conversation is easier if you have it before the number changes rather than after.
The framing that works: the previous number measured optimism, the new one measures evidence, and the gap between them is information you now have and did not have before. Most boards respond better to a lower number with a defensible method than a higher number that missed twice in a row.
From then on, report variance rather than the raw call. A team that lands within 8% every quarter is far more valuable to plan around than one that occasionally nails it exactly.
Source: Gartner, State of Sales Operations Survey.
Want an outside read on whether your commit column would survive the four questions? That is most of what the GTM teardown does, and it costs nothing.
Frequently asked questions
What is a good sales forecast accuracy for B2B SaaS?
Why do sales reps sandbag or inflate forecasts?
Should we use probability percentages on deals?
How often should we run a forecast review?
Zachary works with founder-led B2B SaaS companies between seed and Series B, as a fractional VP of Sales and Marketing and through the team that builds their revenue operation on Throughline.
Get a free GTM teardown