A forecast can look reasonably accurate overall and still create a planning problem.
Imagine a product that is forecast too high month after month. The error in any one month may not seem dramatic, but purchasing and production plans are repeatedly built on demand that does not arrive. Over time, inventory builds, working capital is tied up and the risk of ageing stock increases.
The opposite pattern is just as damaging. If the forecast is repeatedly too low, the business may be left reacting to shortages with expediting, disrupted production plans and missed customer demand.
This is the problem that **forecast bias** helps reveal. It looks beyond the size of the forecast error and asks whether the errors are consistently pointing in the same direction.
Accuracy does not show the whole picture
Forecast accuracy measures how close a forecast was to actual demand. It is useful, but it normally treats an error in either direction as an error of similar size.
For example, if demand is forecast 20 units too high one month and 20 units too low the next, the individual errors still matter. However, when viewed together, there is no obvious tendency to consistently overstate or understate demand.
Bias looks for that tendency.
Using the convention in this article, bias is calculated as **forecast minus actual demand**:
- A positive result means the forecast was higher than actual demand: over-forecasting. - A negative result means the forecast was lower than actual demand: under-forecasting.
The sign convention needs to be agreed and clearly labelled. Some businesses report the calculation in the opposite order, which reverses the signs. The useful question is not whether positive or negative is "good"; it is whether everyone reviewing the measure understands which direction represents over-forecasting and which represents under-forecasting.
Why one period is rarely enough
A single forecast period can be misleading.
Actual demand changes for many legitimate reasons: a customer may place an unusually large order, a promotion may perform differently than expected, or a planned order may move between months. One error does not automatically mean the forecast process is biased.
Bias becomes more meaningful when it is reviewed across multiple periods. If the forecast is repeatedly higher than actual demand, that is a pattern worth investigating. The same applies when actual demand repeatedly exceeds the forecast.
This distinction matters because planners should not try to correct every individual error. Some variation is normal. Persistent bias is a better exception signal because it suggests that something in the forecasting process may need attention.
When the product manager is always a little bullish
Consider a product manager who is responsible for a particular product range. They know the market well, are close to their customers and are confident about the product's future.
When the monthly forecast comes around, they tend to be a little optimistic.
Actual demand might be 1,000 units, but they forecast 1,100. The following month, actual demand is again around 1,000, but the forecast is 1,100 again. Nothing looks particularly alarming in any individual month. The product manager may even have a reasonable explanation each time — a new customer, an expected promotion, additional market demand or an opportunity they believe is coming.
But if this happens month after month, a pattern starts to emerge.
The forecast is consistently higher than actual demand, creating a **positive forecast bias** under the convention used in this article.
The product manager does not necessarily have to be doing anything wrong. Their judgement may be based on genuine commercial expectations that simply aren't translating into realised demand. The problem for the supply planner is that those expectations are becoming embedded in the supply plan.
Purchasing may buy more stock, production may schedule additional runs and inventory may gradually build — even though the underlying demand is not increasing as expected.
This is one reason bias is useful. It can reveal a **systematic human influence on the forecast** that may be difficult to see when reviewing individual forecast changes.
The conversation with the product manager is then much more useful than simply saying *“your forecast is wrong.”* The question becomes:
> “We've noticed that the forecast for this product has been consistently above actual demand. What is driving the difference?”
That opens the door to understanding whether the forecast contains a genuine forward-looking expectation, an overly optimistic assumption, or simply a forecasting habit that has developed over time.
A forecast can be inaccurate without being biased
It is possible to have poor forecast accuracy but little forecast bias.
Consider an item with volatile demand. In one month the forecast is too high, in the next it is too low, and the errors are broadly similar over time. Accuracy may be poor because demand is difficult to predict, but the forecast is not consistently leaning in one direction.
That calls for a different conversation from a consistently biased forecast. The issue may be genuine demand variability rather than a systematic flaw in the forecast.
Conversely, a forecast may appear acceptable on an overall accuracy measure while still showing a consistent tendency to be too high or too low. Smaller errors in the same direction can accumulate into a significant inventory or service issue.
This is why bias is most useful alongside accuracy, not instead of it. Accuracy shows the size of the problem. Bias shows the direction and whether there may be a repeatable pattern behind it.
What persistent over-forecasting does to the supply plan
When forecasts are consistently high, supply plans tend to create more inventory than the business needs.
Depending on replenishment lead times and order quantities, the impact may show up as excess finished goods, unnecessary raw-material purchases or production that consumes capacity without supporting actual demand. The longer this continues, the greater the exposure to working capital, storage costs, ageing stock and eventual obsolescence.
Over-forecasting can also lead to purchasing commitments that are difficult to unwind. By the time the bias is visible in stock levels, orders may already be placed or production may already be scheduled.
The right response is not necessarily to simply reduce every forecast. First understand why the pattern exists. The cause could be assumptions used in the forecast, the statistical model, manual forecast adjustments, or a misunderstanding of what demand should be included.
What persistent under-forecasting does to the supply plan
Under-forecasting creates a different kind of pressure.
Supply arrives too late or in insufficient quantities because the plan was built on demand that was too low. The immediate result may be stockouts, but the wider effect can include rushed purchase orders, premium freight, production schedule changes and poor customer service. In some cases, demand is lost altogether because the product is not available when the customer wants it.
A business may accept some of this risk for low-value or unpredictable items rather than carry substantial inventory. That is a commercial trade-off. But it should be a conscious decision, not the accidental result of a forecast that is consistently too low.
Use bias to decide where to investigate
Forecast bias is most valuable as an exception-management measure.
Rather than reviewing every item in detail, planners can use persistent bias to identify products, customers, locations or forecast segments that deserve closer attention. Then reduce the issue to a practical question: what is repeatedly causing the forecast to sit above or below actual demand?
Useful questions include:
- Is the pattern present over enough periods to be meaningful? - Is the bias concentrated in a particular item, location or customer group? - Have manual adjustments consistently pushed the forecast in one direction? - Does the statistical forecast make assumptions that no longer match the demand pattern? - Are supply decisions being affected by the bias through excess stock, shortages or unnecessary expediting?
The purpose is not to chase a perfect forecast. That is rarely realistic. The purpose is to spot repeatable directional errors early enough to improve the planning decision.
Key Takeaways
- Forecast bias shows whether forecast errors consistently lean high or low over time. - Positive and negative bias must be clearly defined because calculation conventions can differ. - Review bias across multiple periods; one period may simply reflect normal demand variation. - Accuracy and bias answer different questions and should be reviewed together. - Persistent over-forecasting can create excess inventory and unnecessary commitments, while persistent under-forecasting can lead to shortages, expediting and lost sales. - Treat bias as an exception signal that prompts investigation, rather than a reason to react to every forecast error.