Sales forecast accuracy: measure the miss and its direction

01

Two numbers from the same errors

A forecast error is the actual result minus the forecast. For this check, divide that error by the forecast and multiply by 100. Take the absolute value of each result and average them to find the average size of the miss as a percentage of the forecast. The sign disappears, so misses of plus 3 percent and minus 3 percent both count as 3. A commonly used measure, mean absolute percentage error or MAPE, uses the actual result as its denominator instead (Hyndman and Athanasopoulos).

Keep the sign and average again. A positive average means actual results were above the forecasts on average. A negative average means they were below. An average close to zero can conceal misses in both directions, so count how many quarters fell on each side as well.

02

A published forecast that is close and almost always low

Many listed software companies publish guidance, a revenue range for the coming quarter, when they report their results, and a quarter later they report what came in. Guidance is the company's own forecast, published with its financial results, not a sales team's commit. In the record below the ranges sat low against the outcome: no quarter fell below the bottom of its range, so a miss in one direction was the expected result. It also arrives weeks into the quarter it covers, between 11 and 44 days in for these companies and a median of 31, so the number is given after part of the quarter has already been seen. That makes it a clear picture of a forecast that is close and tilted, and a poor benchmark for anyone's sales forecast.

Datadog shows the pattern. On 7 August 2025, 37 days into its third quarter, it guided revenue "between $847 million and $851 million". Three months later it reported revenue of $886 million for the quarter, 4.1 percent above the top of the range.

Across 16 software and security companies and 124 company-quarters that ended between September 2024 and September 2026, reported revenue came in above the top of the guided range in 113 company-quarters and inside it in 11. No quarter fell below the bottom. With the sign removed, the average miss was 2.54 percent of the midpoint of the range. With the sign kept, it was +2.53 percent. Revenue came in above the midpoint in 120 of the 124 company-quarters.

An average miss of 2.5 percent alone hides the direction. If your own record has the same pattern, report its size and direction to the board, then investigate the reason.

03

Guidance against reported revenue, company by company

Read each row's two averages together: the first keeps the sign, the second removes it.

04

Three patterns to look for in your own record

The first pattern is a forecast that lands on the same side every quarter. Datadog came in above the top of its range in all 8 quarters, by 3.9 percent on average, with ranges about 0.6 percent wide.

The second is a forecast close to the middle that still leans. Salesforce came in inside its range in 4 of 8 quarters and below the midpoint in 2, and its average absolute miss against the midpoint was 0.6 percent. A forecast this close to the middle can still lean one way, so the count of quarters on each side matters as much as the two averages.

The third is a wide range. Fortinet gave ranges about five times wider than the average across the 16 companies, and its revenue landed inside the range in 4 of 8 quarters. Its average absolute miss against the midpoint was still 3.4 percent, larger than for most of the others.

05

How to run the check on your last eight quarters

1. For each of the last eight quarters, or twelve if you have them, write down the forecast as it stood on the day you reported it, and that date. Use the number you gave, not a later revision. Choose forecasts made at a comparable point before each quarter ends, because a forecast made later has more information. If you gave a range, write down the low, the high and the midpoint.

2. Next to it, write the actual result and calculate the signed percentage miss as 100 times (actual minus forecast), divided by the forecast. For a range, use its midpoint consistently. If a forecast or midpoint is zero, the percentage is undefined. Record the monetary miss separately for that quarter.

3. Average the percentage misses twice, once with the sign and once after removing it. The first gives the average direction, and the second gives the average size. If you divide the signed average by the absolute average, a result near +1 means most of the total error by size is positive, near -1 that most is negative, and near zero that positive and negative errors largely cancel. If every miss is zero, both averages are zero and the ratio is undefined. Report the perfect match instead.

4. Count the quarters above, below and exactly on the forecast or midpoint. A run of quarters on the same side is a pattern that the averages alone do not show.

5. Report the average absolute miss, the signed average and the count of quarters on each side. Give the forecast date or lead time alongside them so the reader knows when the numbers were set.

06

Where the miss comes from

The two averages show the size and direction of the historical miss. To investigate what contributed to it, compare the deals behind each forecast with what happened, including changes in timing and deal value. Deal qualification can help the team examine what each customer still needs before buying. Opportunity qualification with MEDDICC trains a team to do that, and the questions for each MEDDICC element can show that a deal is further from a decision than the forecast suggests. The MEDDICC scorecard scores one opportunity at a time, so it can be run on each deal behind a quarter's number.