Improving sales forecast accuracy means closing the gap between the number a team calls and the revenue that closes. The front-line manager owns it, RevOps measures it, and the work takes a quarter to show: one quarter to set the rules and trace the slips, the next to see the number tighten.
Most advice on accuracy is about tools and methods: better data capture, weighted pipeline, AI models trained on history. Those help, and there is a bigger, simpler lever. A forecast is a list of deals, each of which a rep believes will close, and the number misses when those beliefs are not backed by anything the buyer has done. This guide fixes that, one deal at a time. It is part of our deal risk guide, which covers where deal risk comes from.
What forecast accuracy is and who owns it
Forecast accuracy is how close the number called at a fixed point in the period comes to what closed. Pick the point and keep it: for a quarter, many teams use the commit called at the start of week 4 or week 6, compared with closed revenue at quarter end.
Measure it two ways. The headline is the percentage difference between the called number and the closed number. The more useful one is deal-level: of the deals in commit at the snapshot, how many closed in the period, and what happened to the rest? A team can hit its number with half its commits slipping and other deals arriving unforecast. The headline looks fine and the forecast told leadership nothing.
The manager owns accuracy because the manager decides what goes in commit. RevOps owns the measurement and the definitions. Reps own the evidence on their deals.
Why forecasts miss
Salesloft's guide sorts forecasting failures into people, process and tools, which is a fair way to start. People bring wishful thinking and sandbagging. Process means inconsistent definitions and reviews that check the number and skip the deals. Tools means incomplete CRM data. Gong argues that CRM data is a poor foundation because it depends on rep entry, and reports from its own analysis of emails and calls that deals with a "red flag" in their email exchanges were less likely to be won.
- PeopleWishful thinking and sandbagging
- ProcessInconsistent definitions, reviews that skip the deals
- ToolsIncomplete CRM data
- PeopleWishful thinking and sandbagging
- ProcessInconsistent definitions, reviews that skip the deals
- ToolsIncomplete CRM data
All three failures meet in one place: a deal sits in commit without the buyer evidence that commit should require. The rep is hopeful because nobody asked for evidence. The process lets confidence decide the category. The data is thin because the conversation that would have produced the evidence never happened. Fix the evidence and you work on all three at once.
The procedure: seven steps
Fix the snapshot and start measuring. Choose the week you compare against (week 4 of the quarter is common) and record two things every quarter: called versus closed, and commit-to-close rate by deal. What good looks like: one number and one list, the same way every quarter, owned by RevOps.
Define the categories by buyer evidence. Commit means the signer has heard the case, a named buyer stated the cost of the problem, the approval steps are mapped with names, and the close date is tied to something on the buyer's calendar. Best case means the signer is identified and the cost is confirmed, with one item open. These are our operating rules; if you run MEDDPICC or another method, use its fields. What good looks like: two reps given the same deal would put it in the same category.
Ask the same questions every week. In the forecast call, ask five questions of every committed deal: who signs and when they last heard the case, what the buyer said the problem costs, the approval steps with names, any other option raised, and the buyer's date. Deals that fail move down a category on the spot. What good looks like: every answer quotes a buyer or names a dated event.
Record why every slip slipped. When a committed deal misses the period, the rep and manager write one line: which piece of evidence was missing at the snapshot. Use fixed reasons (signer never heard the case, no stated cost, steps unmapped, alternative unaddressed, date was the rep's, buyer event moved) so they can be counted.
Coach the gaps while the deal is live. Every deal that moves out of commit gets a named buyer conversation and a date. The two or three largest get a 20-minute coaching call with the rep that week, where the rep talks the deal through and leaves with an opening line said out loud. How to coach a stalled deal shows that call.
Fix close dates at the source. Many slips trace to a date the buyer never owned. Coach reps to ask for the buyer's date and what depends on it before a deal enters commit. How to coach reps on close dates has the questions.
Review the slip reasons each quarter. Count them. The top reason is next quarter's coaching focus for the whole team, and the rep-by-rep counts show who needs help with which step. What good looks like: the top reason changes from one quarter to the next, because the old one got fixed.
A one-page category definition to copy
Write the rules once, share them with every rep and manager, and keep them on the forecast call screen. A version to adapt:
FORECAST CATEGORIES (buyer evidence required)
COMMIT
[ ] Signer named, and has heard the business case (date: ____)
[ ] Cost of the problem stated by a named buyer (quote: ____)
[ ] Approval steps mapped, with a name on each step
[ ] Close date tied to a buyer event (event: ____)
[ ] Any other option the buyer raised has been addressed
BEST CASE
[ ] Signer identified
[ ] Cost of the problem confirmed by a buyer
[ ] No more than one commit item still open (which one: ____)
PIPELINE
Anything else. A deal without a stated cost stays here.
Rule: a deal moves up only when every box in the new category is ticked
with evidence. A deal moves down the week a box stops being true.
Two habits make the sheet work. Each ticked box carries its evidence in the CRM, so a manager can check it in the forecast call without asking. And the rules belong to RevOps and the sales leader together, so a rep cannot argue a deal up by finding a manager who reads them loosely.
A worked example: one quarter's slips
This example is illustrative. A team of seven reps had 24 deals in commit at the week-4 snapshot. Fourteen closed in the quarter. The manager and reps recorded a reason for each of the ten that slipped.
| Slip reason (missing evidence at snapshot) | Deals | What it means for next quarter |
|---|---|---|
| Signer never heard the case | 4 | Commit requires the signer on a call; coach champions to bring the case up |
| Close date was the rep's | 3 | Ask for the buyer's date and its cause before commit |
| Approval steps unmapped (procurement, legal) | 2 | Map the paper process by the proposal stage |
| Buyer event moved (reorganization) | 1 | Not coachable; leave it out of the coaching plan |
Nine of the ten slips were visible at the snapshot as a missing piece of evidence. Under the step-2 rules, none of those nine would have been in commit. The forecast would have been smaller and closer to what closed. Next quarter the team focuses its coaching calls on getting the signer to hear the case, and the manager checks the count again at quarter end.
How to calculate forecast accuracy, with numbers
The calculation takes two lines. Take the number called at the snapshot and the number that closed, and divide the difference by what closed.
Take the illustrative quarter above. The manager calls a commit of $1.2 million at the start of week 4. The quarter closes at $1.0 million. The miss is $200,000, which is 20% of what closed, so the forecast was 80% accurate on the headline. Some teams divide by the called number instead; either works if you never switch.
The deal-level view tells the manager more. Fourteen of the 24 committed deals closed, a commit-to-close rate of 58%, and part of what closed came from deals that were in best case or not forecast at all. The ten slips are the reasons recorded in the table above.
Track both numbers on one line per quarter, along with the top slip reason. After three or four quarters the line shows whether the rules in step 2 are tightening the forecast, or whether new deals are still covering for slipped ones.
Forecasting methods, and what each one can fix
Teams usually forecast in one of three ways, and many blend them.
| Method | How it works | What it can fix | What it cannot fix |
|---|---|---|---|
| Category (commit, best case, pipeline) | Reps and managers place each deal in a category, and the commit adds up to the call | Inconsistent judgment, once the categories have evidence rules | Deals placed on hope when the rules are loose |
| Weighted pipeline | Each stage carries a probability, and deal values are multiplied by it | Rep optimism, because the stage sets the weight | Deals sitting in the wrong stage |
| Historical or AI model | A model trained on past deals predicts which will close and when | Patterns no person would spot across hundreds of deals | A missing buyer conversation on a live deal |
The three share one weakness: each depends on the stage or category being true. A weighted pipeline with deals in the wrong stage multiplies the wrong number. A model trained on deals that were staged loosely learns loose staging. That is why the steps above start with evidence rules for the categories. Whichever method you use, it gets more accurate when the inputs describe what the buyer has done.
Coaching the two rep habits behind most misses
Two habits push a forecast off, in opposite directions.
Optimism. The rep believes the deal will close because the champion is enthusiastic. In the forecast call, it shows up as answers about how the buyer feels with no quote about what the buyer did. Coach it with the five questions from step 3, asked the same way every week, so the rep learns to come with evidence. Do not argue the rep's read of the champion. Ask what the signer has said.
Sandbagging. The rep keeps a strong deal out of commit to protect against a miss. It shows up as deals that meet every commit rule sitting in best case, then closing early. The fix is the same rulebook used in the other direction: if every box is ticked with evidence, the deal goes in commit, whatever the rep's comfort. A manager who rewards early downgrades and accepts honest commits gives reps no reason to hide either way.
Both habits fade when the category is decided by evidence the manager can see.
A 90-day plan
Days 1 to 15. Pick the snapshot week, write the category rules, and walk every rep and manager through the one-page sheet. Record last quarter's slips from memory and CRM history as a baseline, with reasons.
Days 16 to 60. Run the five questions in every forecast call. Move deals that fail. Record a reason for each slip as it happens, and give the largest gaps a coaching call the same week.
Days 61 to 90. Measure the snapshot against close for the quarter, count the slip reasons, and set the top reason as next quarter's coaching focus. Compare commit-to-close with the baseline. Expect the change to show fully in the second quarter, once the rules have been in place from the first week.
Where it goes wrong
- Changing the definitions mid-quarter. If commit loosens in week 10 to make the number, the snapshot comparison means nothing. Hold the rules until quarter end.
- Treating accuracy as a tool purchase. A forecasting model trained on history learns your team's habits, including bad ones. It can predict slippage. The rep still has to have the conversation that prevents it.
Applying the rules to every rep
The procedure needs evidence checked on every committed deal every week, and a coaching call on every gap that matters. That is more than a manager with eight reps can do well, and quality matters: across 1,246 pharmaceutical reps, managers' coaching skill predicted goal attainment, while frequent coaching from less skilled managers was linked to worse results (Dahling et al. 2016, via our Game Winning Kickoffs study).
Our AI sales experts were built for that load. A rep brings a committed deal to one of them on a call, and the expert stages it by buyer evidence against the same category rules, names what it is missing and plans the conversation that fills it, in the team's own method. Every rep's deals get checked against one rulebook, whichever manager they report to. Bryce shows the manager which gaps recur across the team. More in drive deal execution. If last quarter's commit missed by more than you can explain deal by deal, talk to us.
FAQ
How do you measure sales forecast accuracy?
Compare the number called at a fixed point in the period, such as the week-4 commit, with the revenue that closed by period end, as a percentage difference. Then measure it by deal: what share of committed deals closed in the period. The deal-level number shows whether the forecast was right for the right reasons.
Why are sales forecasts inaccurate?
Sales forecasts are inaccurate mainly because deals sit in commit without buyer evidence behind them. Typically the signer has not heard the case, no buyer has stated a cost, approval steps are unknown, or the close date is the rep's. Wishful thinking, loose definitions and thin CRM data all feed that one problem.
What is a good sales forecast accuracy rate?
A good sales forecast accuracy rate is one that beats your own past quarters, measured the same way each time. Accuracy depends on deal size, cycle length and how many deals make up the number, so a team closing a few large deals a quarter will swing more than one closing hundreds of small ones. Pick a fixed snapshot, measure it the same way every quarter, and judge progress against your own trend.
How can AI improve sales forecast accuracy?
AI forecasting tools read activity and conversation data and flag deals that look like past slips, which helps a manager decide where to look. They improve accuracy most when the flag leads to a coaching conversation that gets the missing buyer evidence, so the deal either earns its place in commit or leaves it.
How do you stop deals slipping out of the forecast?
Require buyer evidence for commit, especially a close date tied to something on the buyer's calendar. Ask the same evidence questions of every committed deal each week, move deals that fail, and coach the rep on the conversation that fills the gap the same week. Record why every slip happened and coach the top reason next quarter.
Who should run the forecast call?
The front-line manager runs the forecast call for their team, using the category rules RevOps and the sales leader own. RevOps brings the snapshot and the slip list. The sales leader joins for the deals that move the number. Keeping the rules with RevOps and the call with the manager stops either one from bending the definitions to fit the week.