Low every quarter
Predictable under-forecastRevenue decision guide · Forecast accuracy
Forecast accuracy: find where the error enters
Measure forecast error by role, period and direction, then connect it to the deals and decisions that moved the number.
The operating question
A single accuracy percentage hides the operating problem.
The quarter closes six per cent under plan. On Monday morning the review starts with the deals that slipped, then moves to the final forecast and the actual result. The number is wrong, but the room can't agree where it became wrong.
The Seller remembers calling one deal as Best Case. The Manager remembers moving it to Commit after a procurement meeting. Leadership remembers taking a lower number to the board. Only the final roll-up survived, so all three accounts sound plausible and none can be tested. Forty minutes later the team agrees to inspect the pipeline earlier next quarter. The forecast process itself remains untouched.
That is the problem with an accuracy figure on its own. It tells you the distance between a number and an outcome. It doesn't tell you whose judgment introduced the error, whether the error was consistently high or low, or which evidence was available when the number was submitted.
High, low, high, low
Direction changes each quarterThe obvious fix is to track one percentage
It's reasonable to start with forecast accuracy over time. A team should know whether the number it submits bears any relationship to what lands. The trouble starts when one percentage becomes the answer rather than the first question.
Two teams can report the same headline accuracy while needing opposite responses. One is six per cent under every quarter. The other alternates between high and low. The first team has a predictable bias that can be corrected in review. The second doesn't yet have a stable way to know. Averaging them into the same score removes the only information a Manager can act on.
Your accuracy figure is an average of disagreements nobody wrote down
A forecast isn't one number. It's a sequence of named positions. The Seller submits one view, the Manager agrees or changes it, leadership publishes another, and each position is taken at a different distance from the outcome.
If the system keeps only the latest value, the disagreement disappears. The final number can be measured, but the process that produced it can't. That makes the review feel objective while leaving the operating question unanswered: where did the error enter?
Keep distance, direction and decision point
Measure absolute error to understand how far the forecast missed. Keep signed error to show whether it was high or low. Then separate each forecast point: start of quarter, mid-quarter, final submission, Seller, Manager and Revenue Leader.
Those are not competing metrics. Together they tell you whether the team learns as the quarter progresses, whether one layer consistently corrects another, and whether a late accurate submission simply arrived too late to be useful. Accuracy improves when the review can find the decision that needs changing, not when the score is displayed more often.
Preserve what the number rested on
A submitted forecast needs its target, definition, included deals and supporting evidence sealed with it. When a deal later slips, new information must not be allowed to improve the earlier position. The review should see what the person could know at the time, including what was missing.
That distinction changes the coaching conversation. A person who made a defensible forecast from incomplete evidence isn't the same as a person who ignored evidence already in front of them. A final accuracy percentage makes them look identical.
Where accuracy programs go wrong
Don't turn the measure into a public ranking of Sellers. Once the number affects status, people adjust the forecast to protect the score. Don't compare a final leadership submission with a Seller position submitted weeks earlier. Don't move the target after the quarter and describe the result as accurate.
Most of all, don't let an accurate roll-up hide persistent error underneath. A leadership team can compensate for a biased Manager and still publish a good number. That is useful this quarter and dangerous next quarter if the correction lives only in one person's head.
A note on the evidence
PROOF doesn't yet publish a customer accuracy improvement figure. No percentage belongs on this page until it has a defined baseline, forecast point, period and denominator.
The claim here is narrower and testable: keeping each submitted position, its direction of error and the evidence available at the time lets a team locate where the forecast process diverged from the outcome.
How PROOF approaches it
Keep the evidence, decision and action connected.
PROOF keeps the source, timing and accountable decision together so the team can act while the outcome can still change and learn from what happened afterwards.
See the working product