Forecast Accuracy
Forecast accuracy describes how closely forecast values match corresponding actual results under a stated error measure, scope, and time horizon.
There is no universal forecast-accuracy formula. Absolute error, percentage error, MAE, MAPE, WAPE, and bias answer different questions and have different limits.
Measure forecast accuracy by comparing like-for-like forecast and actual observations with a method suited to the data. State the horizon, level of aggregation, error measure, and treatment of zero or near-zero actuals.
Turn repeated forecast errors into evidence
One variance explains one observation. Accuracy and bias emerge from a defined series of comparable observations.
- 01Forecast
The version locked before the actual result was known.
- 02Actual
The corresponding recorded result for the same scope and period.
- 03Error
The signed or absolute difference under a named method.
- 04Repeated observations
Comparable errors across periods, categories, or horizons.
Assess size of error and persistent direction separately.
What is Forecast Accuracy?
Forecast accuracy is an evaluation of how close prior forecast values were to corresponding actual results. The comparison should use the forecast version that existed before the outcome, the same period and organizational scope, and a clearly defined measure of error.
Different measures emphasize different properties. Absolute error preserves the units of the forecast. Percentage error supports scale comparison but becomes unstable when the denominator is zero or close to zero. Mean absolute error summarizes average magnitude. MAPE averages percentage errors and inherits their zero-value problem. WAPE compares total absolute error with total actual magnitude and can hide offsetting behavior across categories if used alone.
Accuracy and bias are related but distinct. A forecast can have moderate absolute error without persistent bias, or it can repeatedly overforecast or underforecast even when total error looks acceptable. Horizon also matters: a one-month forecast and a twelve-month forecast should not be judged as if uncertainty were identical.
Error versus accuracy
Forecast error is the difference between forecast and actual. Some teams transform error into an accuracy score, but that transformation is a policy choice and can produce misleading results, including negative or undefined percentages. Reporting the selected error measure directly is often clearer.
Accuracy versus bias
Absolute measures ignore direction. Bias retains the sign of each error and shows whether forecasts tend to run high or low. Review both when repeated directional error would change decisions.
Category versus total
A total can look accurate when category errors offset each other. Revenue may be above forecast while payroll or other costs are also above forecast. Review the level at which the decision is made, then reconcile it with the total.
Zero and near-zero values
MAPE and similar percentage measures divide by actual or forecast values. Zero and small denominators can make the result undefined or disproportionately large. Use an alternative measure or a documented exception policy for those observations.
Why it matters
Accuracy review creates a feedback loop between assumptions and recorded results. It helps a team find where timing, volume, price, hiring, cost, or classification assumptions need investigation.
The result is evidence about a forecasting process, not a character score for the person who prepared it. Targets and tolerances should reflect the decision, horizon, volatility, and cost of error rather than an unsupported universal benchmark.
Common error measures
Error = Actual - Forecast Absolute Error = |Actual - Forecast| Absolute Percentage Error = |Actual - Forecast| / |Actual| x 100 MAE = Sum of absolute errors / Number of observations WAPE = Sum of absolute errors / Sum of |Actual| x 100
Illustrative multi-period review
A team locked monthly cost forecasts of $80,000, $82,000, and $85,000. Actual costs were $84,000, $79,000, and $91,000. The signed errors are $4,000, -$3,000, and $6,000. The absolute errors are $4,000, $3,000, and $6,000, so MAE is about $4,333.
The positive signed total suggests an underforecast tendency in this small sample, but three observations are not a universal conclusion. The team should review the categories and assumptions behind the differences.
How RunwayCal helps
RunwayCal's Budget vs Actual and variance views can support like-for-like review of planned and recorded values where those states are available. The variance remains an observation that a person interprets in context.
RunwayCal does not publish a universal forecast-accuracy score, infer causes automatically, or guarantee that future forecasts will match actual results.
Common mistakes
- 1Calling one percentage formula the universal definition of forecast accuracy.
- 2Using MAPE when actual values are zero or close to zero without a policy.
- 3Reviewing only the total when category errors offset each other.
- 4Revising the historical forecast after actual results are known.
Get the Financial Clarity Newsletter
Practical tips on cash flow, runway, and financial decisions for founders, business owners, CFOs, investors, and board members. Free, weekly, no spam.
Compare the plan with what actually happened
Use a traceable variance review to improve assumptions without turning one score into the answer.
Explore Budget vs Actual