Most revenue teams report one accuracy number and treat it as the answer. The number is usually forecast divided by actual, expressed as a percentage, computed once at quarter close. It answers a narrow question and stays silent on the two failures that cost the most money: consistent direction and late arrival.
Each metric below detects a different failure. Picking the right set is a matter of deciding which failures you cannot afford to miss.
What does each accuracy metric actually detect?
Precision, direction, and timing are three separate properties, and no single metric covers more than one of them. Reporting a single figure guarantees two blind spots.| Metric | Formula | Detects | Blind to |
|---|---|---|---|
| MAPE | Mean of abs(forecast - actual) / actual | Average percentage miss | Small-denominator blowups |
| WAPE | Sum abs(error) / sum actuals | Dollar-weighted precision | Direction of the error |
| Signed bias | Mean of (forecast - actual) | Systematic over or under calling | Size of individual misses |
| Hit rate | Percent of periods inside a tolerance band | Consistency against a threshold | Magnitude outside the band |
| Day-one error | Signed error of the week-one forecast | Whether you predict or report | Late-quarter execution |
Why does MAPE break at the segment level?
MAPE divides by the actual, so any segment that closed a small number produces a huge percentage from a small dollar miss. An enterprise segment that forecast $1.2 million and closed $400,000 shows 200 percent error. A mid-market segment that forecast $9 million and closed $8 million shows 12.5 percent. The enterprise miss is smaller in dollars and dominates the average.This is why MAPE-based dashboards keep flagging the same low-volume segments quarter after quarter. The metric is reporting the denominator, not the forecasting.
WAPE fixes this by pooling the numerator and denominator before dividing. Every dollar of error counts the same regardless of which cell it came from, which matches how a CFO experiences a miss.
When is signed bias the metric that matters most?
When the error points the same direction more than four quarters running. At that point the error is not noise, it is an assumption baked into the process, and it compounds into hiring plans and board commitments built on revenue that was never arriving.Signed bias is also the metric that separates two problems people confuse. A rep or segment missing by 20 percent high in one quarter and 20 percent low in the next has an absolute error of 20 percent and a bias of zero. That is a qualification problem. A rep missing by 8 percent high six quarters running has a smaller absolute error and a real bias. That is a calibration problem, and the fixes have nothing in common.
Keep the sign. The moment you convert to absolute values you have destroyed the only information that identifies systematic error. More on the distinction sits in our forecast accuracy glossary entry.
Why is quarter-close accuracy the least useful number you report?
Because knowing the number in the last week of the quarter changes nothing. The quarter has already happened. Every decision that could have altered the outcome, on pipeline generation, discounting, or resource allocation, closed weeks earlier.Day-one error is the number with operating value. It measures whether you understood the shape of the quarter before the quarter began, which is the only window in which action is still possible. Most teams find their quarter-close accuracy respectable and their day-one accuracy poor, which is a precise description of a process that reports rather than predicts.
Track both and publish the gap between them. A shrinking gap is the clearest evidence that a forecasting change worked. Around 90 percent accuracy on new and expansion business is common when teams build the forecast manually, though the manual version degrades as conditions change during the quarter. ORM targets 95 percent without manual adjustment, holding from day 1 through day 90.
What about accuracy metrics that measure the wrong thing entirely?
Hit rate against plan and pipeline coverage are performance metrics wearing an accuracy costume. Both get reported in accuracy reviews and neither belongs there.Hitting plan measures sales performance against a target set before the period started. A team can hit plan four quarters running while forecasting badly every time, because the misses landed in a favorable direction. The reverse is equally common.
Pipeline coverage is worse, since it looks quantitative enough to pass as a prediction. Across ORM customers coverage runs from 1.4x to 5x, with most near 3.5x. A company at 4x can miss badly if the coverage is concentrated in the wrong stage, owned by the wrong reps, or aged past usefulness. Coverage is an input. Treating it as a conclusion is covered in the 3x pipeline coverage rule is wrong.
How should you set the tolerance band?
Derive it from your own historical variance rather than importing a target. Compute the standard deviation of your signed quarterly error over the last eight quarters. A band of plus or minus one standard deviation is a realistic starting tolerance, and it will be wider for a company with a small number of large deals than for one closing hundreds of small ones.Importing someone else's 5 percent tolerance into a business closing twelve enterprise deals a quarter guarantees a permanent red status that everyone learns to ignore. The point of a band is to trigger investigation when something has actually changed, and a band calibrated to your own deal mix does that. Definitions and formulas for the underlying measures are in our sales forecasting best practices guide.
Frequently Asked Questions
Is MAPE a good metric for revenue forecasting?
At the company level yes, at the deal or small-segment level no. MAPE divides error by actuals, so a segment that closed a small number produces an enormous percentage from a small dollar miss. Use WAPE when cell sizes vary.
What is the difference between MAPE and WAPE?
MAPE averages the percentage error of each item, giving every item equal weight regardless of size. WAPE divides total absolute error by total actuals, so large deals influence the result in proportion to their dollars. WAPE is the safer default for revenue.
Why track signed bias when you already track absolute error?
Absolute error tells you how far off you were. Signed bias tells you which direction you are consistently off, which is the only one of the two that identifies a systematic problem you can correct. A team can have low absolute error and severe bias.
Is hit rate against plan a forecast accuracy metric?
No. Hitting plan measures performance against a target set months earlier. Forecast accuracy measures whether your prediction matched the outcome. A team can hit plan every quarter with terrible forecasting, and the reverse is also true.
How many accuracy metrics should a revenue team report?
Three. One precision metric such as WAPE, one direction metric such as signed bias, and one timing metric such as day-one error versus quarter-close error. Anything beyond that gets ignored, and anything less leaves a failure mode invisible.
See how ORM turns these insights into action
ORM builds custom revenue forecast models for B2B SaaS companies. Not dashboards. Prescriptive analytics that tell you what to do next.
Schedule a Demo