Forecast Accuracy

Hiring, ad spend and cash decisions are made against the forecast. A forecast that is regularly too high commits the business to costs the revenue never covers. One that is regularly too low holds back spending the business could have afforded. Either way, people stop trusting the number and start adjusting it by feel.

A forecast is also a message to people outside the sales team. A board, an investor or a bank reads it as a promise, and a founder who misses by a third twice in a row has to explain why the next number should be believed.

Accuracy is useful beyond the number itself. When the same deals are over-counted every quarter, the pattern points to a specific fault in stage definitions, deal hygiene or the sales cycle, which can be fixed.

• Freeze the forecast on the first day of the period, so it cannot be quietly rewritten later.
• Replace default stage probabilities with the real win rate from each deal stage over the last few quarters.
• Remove any deal with no activity for a set number of days from the active forecast.
• Forecast a range, such as committed, likely and best case, and score each band separately.
• Review every large miss and write down the reason, so the pattern becomes visible.

Forecast accuracy compares a prediction with the result. The simplest method is to take the revenue you said would close, subtract what did close, and divide the gap by the actual result. Say a team forecast £200,000 for the quarter and closed £150,000. The gap is £50,000, which is a third of the result, so the forecast was a third too high. Use the same method every period so the numbers can be compared.

Two different things can be wrong. Error is the size of the miss. Bias is the direction. A forecast that is wrong by 10 per cent in both directions is noisy. One that is 10 per cent too high every quarter is biased, and bias is far easier to correct.

• Rewriting the forecast mid-period. A number that moves toward the result is not a forecast. Freeze it and score it.
• Scoring only the total. A total can look right while the committed band is wildly off and best case rescues it. Score each band.
• Treating one miss as the lesson. A single quarter can be luck. Look for bias across several periods.
• Blaming the reps. If every rep over-forecasts, the cause is usually the stage definitions or the default percentages.
• Forecasting deals that cannot close in time. A deal that needs a procurement process of 90 days does not close in a 60-day window.

• Deals sit in a late stage for months and keep counting as likely.
• Stage definitions are loose, so two people put the same deal in different stages.
• Default stage percentages in the CRM are used instead of the percentages the business has actually seen.
• Deals are forecast to close in a quarter that is shorter than the usual sales cycle.