Sales Forecast Accuracy: Formula, Cohorts, and Limitations

Sales forecast accuracy compares a forecast made at a defined point in time with the results that later occurred. It is not one universal metric: teams use different formulas, time horizons, and levels of aggregation. A useful report states those choices so leaders can tell whether forecast quality changed or the calculation changed.

Choose a formula and freeze point

For a single period with a nonzero actual, absolute percentage error can be calculated as the absolute difference between forecast and actual divided by actual. Some teams express accuracy as one minus that error. This can produce a negative accuracy when the error exceeds actuals, and percentage error is undefined when actuals are zero. For a set of periods, weighted absolute percentage error divides the sum of absolute errors by the sum of actuals, but it too needs care when total actuals are small.

Also report forecast bias: the signed difference between forecast and actual. A forecast can be close on average while consistently optimistic or pessimistic. State whether the forecast was taken at month start, quarter start, or a fixed number of days before close, and compare the same horizon across periods.

Compare like with like

• Separate new business, renewals, and expansion when their sales motions differ.
• Compare by segment or team only when the sample is large enough to be stable.
• Keep closed-lost, pushed, and open opportunities governed consistently.
• Show the forecast and actual values alongside any accuracy percentage.

Diagnose the error, not just the score

A miss can come from deal slippage, stage probabilities, opportunity amount changes, missing opportunities, or market events. Review what was known at the forecast date, not only what is visible after the outcome. Cohort deals by the period they were expected to close to expose repeated timing changes.

Worked example

If the quarter-start forecast is $1.2 million and actual bookings are $1.0 million, the absolute error is $200,000. Absolute percentage error is 20% of actuals; under a one-minus-error convention, reported accuracy is 80%, while signed bias shows the forecast was $200,000 high. A different denominator or aggregation method will produce a different headline. Record the formula beside the chart so readers can reproduce it.

Treat accuracy as a process signal, not a ranking of individual sellers. Pair it with forecast bias, close-date movement, pipeline coverage, and documented judgment. If the definition changes, calculate a comparable historical baseline or clearly mark the break in the series.