What it means
Almost every operational decision starts with a demand number. Purchasing, production scheduling, warehouse space, seasonal hiring and cash planning all flow from an estimate of what will sell, so the quality of that estimate sets the quality of everything downstream.
Forecast accuracy converts a vague sense of being roughly right into a tracked number. The usual approach measures the absolute error, meaning the size of the miss regardless of direction, and expresses it as a share of actual demand.
Direction still matters commercially even though the metric ignores it. Over-forecasting ties up cash in stock that may be discounted later, while under-forecasting causes stockouts, expedited freight and customers who quietly buy elsewhere.
Most businesses measure accuracy at several levels, because the picture changes with the lens. A total company forecast can look excellent while individual product lines are wildly wrong, since one item's over-forecast cancels out another item's shortfall.
Sensible targets vary enormously by category. Stable repeat-purchase goods often achieve 85% to 95% at product level, whereas fashion, new launches and heavily promoted lines sit far lower, so the honest benchmark is the trend in your own numbers rather than a universal figure.
Alongside accuracy, most planning teams also track bias, which is the average error including its sign. Accuracy tells you how big the misses were and bias tells you whether they lean consistently in one direction, and a forecast that is 85% accurate with no bias is a very different problem from one that is 85% accurate and always low.
In practice
Real-world examples.
Example
A frozen food producer forecasts 40,000 units of a ready meal and ships 44,000. The error of 4,000 units against actual demand gives accuracy of 90.9%, comfortable enough that the planner leaves the model alone.
Example
An electronics retailer forecasts 5,000 units for a headphone launch and sells 3,500. The error of 1,500 units is 42.9% of actual demand, so accuracy is 57.1% and the buyer is left with three months of surplus stock.
Example
A garden centre forecasts 20,000 bags of compost for spring and sells 26,000 after an unusually warm March. The error of 6,000 bags is 23.1% of actual demand, an accuracy of 76.9%, and the shortfall cost several weekends of lost sales.
Think of it
“Forecast accuracy shows how well you predict what customers will want-your prediction skill.
Formula
Calculation
Forecast error = Actual demand - Forecast
Forecast accuracy = (1 - (Absolute error / Actual demand)) x 100
A beverage company forecasts 12,000 cases of one flavour for October and actually sells 10,000. The absolute error is 2,000 cases, the error rate is 2,000 / 10,000 = 20%, and accuracy is 100% - 20% = 80%.
Rolling the measure up across a range works the same way, using the average of the individual error rates. If three products record error rates of 10%, 20% and 30%, the mean absolute percentage error is (10 + 20 + 30) / 3 = 20%, giving portfolio accuracy of 80% even though no single product was 80% accurate.Case study
Seen in the real world.
Brightloom Beverages is a fictional soft drinks maker, used here purely as an illustrative example. Its planning team reported 94% forecast accuracy every month and was quietly proud of it, until a new operations director asked for the same measure at product level and got 71%.
The gap was arithmetic rather than dishonesty. Two flavours were consistently over-forecast and two were consistently under-forecast, and at company level the errors cancelled out into a flattering total that told nobody anything useful.
Brightloom began measuring accuracy by product, fed the promotional calendar into the forecast, and lifted product-level accuracy from 71% to 82% over three quarters. Stock written off at the end of shelf life fell from $410,000 to $260,000 a year, a saving of $150,000, and out-of-stock complaints from its two largest retail customers stopped appearing in quarterly reviews.
The planners also began publishing bias alongside accuracy, which showed that one flavour had been over-forecast in eleven of the previous twelve months. The lesson in this fictional example was not really about statistics: Brightloom's forecasts improved once the people who knew about promotions, listings and competitor activity were in the room when the numbers were set, and once the measure being reported was specific enough to point at a product rather than a company.
Watch out
Common mistakes.
- Judging accuracy only at total company level, where over-forecasts on one line hide under-forecasts on another and the headline number flatters everyone.
- Allowing the sales team's target to double as the demand forecast, which turns an operational estimate into a negotiation and builds in a permanent upward bias.
- Measuring the error against the forecast rather than actual demand, which produces a different and usually kinder percentage than the standard calculation.
Questions
People also ask.
What is a good demand forecast accuracy?
It depends heavily on the product, but stable repeat purchases often reach 85% to 95% at item level while new or promoted lines are far less predictable.
Can accuracy be negative?
Yes, when the absolute error is larger than actual demand, which happens when a forecast is more than double what sells, and most teams simply report it as zero.
Does a more complex model always improve accuracy?
No, gains usually come first from cleaning the data, removing one-off events and adding known promotions, before any change of statistical method.
From the founder's library
Take it further with the book.
Build your financial confidence beyond this definition. Shihan's full-length guide, Accounting Fundamentals, takes the same plain-English approach and turns it into a complete, practical playbook for non-finance managers, business owners and students - with chapter-end quiz answers and presentation slides included.
25% off with code MMHQ25, applied at checkout. Priced in USD - checkout may show the equivalent in your local currency.
View the book and save 25%