Free Tool · Interactive

Forecast Decision Headroom Calculator

Before you spend 3 weeks sharpening a forecast, find out whether a sharper number would change the decision at all. Enter the threshold, the forecast and your honest typical error, and the answer updates as you type. Nothing is sent anywhere, the maths runs in your browser.

The Decision
The tipping point at which the decision changes. Build the line above 10,000 units. Add a shift above a capacity limit. Reorder below a stock level.
What the forecast currently says.
How Wrong It Usually Is
Your honest average miss on this forecast, not your best month. If you track MAPE, that is the percentage option.
Units, dollars, pallets, hours. Used only to word the answer.
Reset to example numbers
Decision headroom
4.0

Distance to the threshold, divided by your typical error.

What the number is made of
Distance to the threshold 4,000 units
Typical forecast error 1,000 units
That error as a share of the forecast 7.1%
Headroom 4.0
1.0 and belowAccuracy is decisive
1.0 to 2.0Thin, worth a check
Above 2.0Accuracy buys nothing

More accuracy buys nothing here

100 plausible outcomes
100 plausible outcomes against the decision thresholdEach dot is one plausible outcome: the forecast of 14,000 units plus a normal error with a standard deviation of 1,253 units. The decision threshold is 10,000 units. Outcomes on the forecast's side of it keep the decision; the rest flip it. The decision flips in 0 of 100 plausible outcomes.8,00010,00012,00014,00016,00018,000Actual outcome, unitsDecision thresholdForecast100 plausible outcomes against the decision thresholdEach dot is one plausible outcome: the forecast of 14,000 units plus a normal error with a standard deviation of 1,253 units. The decision threshold is 10,000 units. Outcomes on the forecast's side of it keep the decision; the rest flip it. The decision flips in 0 of 100 plausible outcomes.10,00012,50015,00017,500Actual outcome, unitsDecision thresholdForecast
  • The decision holds
  • The decision flips

The decision flips in 0 of 100 plausible outcomes.

The typical error is read as an average miss, so the outcomes spread with a standard deviation 1.25 times as large: 1,253 units.

The thinking behind this tool: Stop Optimizing Forecast Accuracy →
It works through 2 cases with the same threshold and the same error that reach opposite conclusions about where to spend effort.

One forecast is arithmetic. A forecasting process is a different problem: knowing which of your forecasts feed real thresholds, and which exist because they always have, usually retires more work than it creates. Start a conversation →