What are the signs that MAPE is not a reliable forecast metric?

Why MAPE breaks down as a forecast metric

MAPE only works well when actual values stay comfortably above zero and demand is reasonably stable. Once actuals get very small, the percentage error can explode even when the absolute miss is minor. That makes the metric look worse than the operational impact really is, which is why low-volume and intermittent items are a poor fit for it.

It is also sensitive to the direction of error. If over-forecasts hurt more than under-forecasts, MAPE can hide that asymmetry because it treats percentage miss size as the main signal. In practice, that means two forecasts with the same MAPE can have very different business consequences.

What bad MAPE signals usually look like

A reliable forecast metric should help you distinguish noise from genuine planning failure. If MAPE jumps wildly from one period to the next, especially across items with low or uneven demand, the metric is probably reacting to denominator effects rather than forecasting quality. The same warning applies when one or two small actuals dominate the average.

Another sign is that MAPE rewards the wrong behaviour. Teams may appear to improve by avoiding large percentage misses on tiny values, while still making decisions that create poor inventory, staffing, or replenishment outcomes. If the metric does not align with the cost of being wrong, it is not a trustworthy planning measure.

When the error distribution is skewed, MAPE can also obscure bias. A forecast that consistently runs high may create stock or capacity waste, while a forecast that runs low may create shortages, missed service targets, or emergency work. If the metric does not surface which side of the error is hurting you, it is not giving you the full operational picture.

What to use instead when MAPE is misleading

For intermittent demand or many low-volume items, look for metrics that behave better around zero and that preserve decision usefulness. Absolute-error measures are often easier to interpret because they stay tied to units rather than percentage swings. If the decision is about service levels, inventory risk, or capacity planning, the metric should reflect that business cost, not just forecast math.

It also helps to segment the portfolio. High-volume stable items can sometimes be assessed differently from slow-moving items, seasonal items, or SKUs with sporadic sales. A single aggregate metric across all of them usually hides more than it reveals, so separate treatment is often more honest and more actionable.

Practitioner Guidance

What to verify: Check whether the forecast set contains zeros, near-zeros, or intermittent demand before trusting any MAPE result. If those conditions are common, treat the metric as a diagnostic input, not a performance verdict.

Decision rule: If the business cost of over-forecasting differs materially from the cost of under-forecasting, do not rely on MAPE alone. Add a metric or review step that makes the asymmetry visible so the forecast can be judged against the actual planning risk.

Practitioner takeaway: MAPE is most useful when demand is stable, positive, and reasonably balanced, but it becomes misleading when small denominators or asymmetric costs dominate the picture.

Related resources from NHI Mgmt Group

• What do security teams get wrong about using MAPE as a single performance metric?
• What are the signs that access graph queries are failing to give security teams reliable answers?
• What are the signs that a mobile app security platform is not giving teams reliable results?
• What are the signs that web application security testing is not giving reliable results?