``` Tracking signal = ( running sum of signed forecast errors ) / MAD ```
Reading the number
Signed error here is forecast minus actual, so a positive value means the forecast ran high.
| Quarter | Forecast | Actual | Signed error | Running sum | MAD to date | Tracking signal |
|---|---|---|---|---|---|---|
| Q1 | $4.6M | $4.5M | +$0.1M | +$0.1M | $0.10M | 1.0 |
| Q2 | $5.1M | $4.8M | +$0.3M | +$0.4M | $0.20M | 2.0 |
| Q3 | $5.6M | $5.1M | +$0.5M | +$0.9M | $0.30M | 3.0 |
| Q4 | $6.0M | $5.4M | +$0.6M | +$1.5M | $0.375M | 4.0 |
Plus or minus 4 is the standard control limit. A signal inside the band means errors are canceling the way random errors should. A signal outside it means the forecast is producing a predictable overstatement or understatement.
Why the running sum is the point
Random error cancels. A forecast that misses 8% high, then 7% low, then 6% high, then 9% low keeps its running sum near zero no matter how large the individual misses are, and tracking signal stays flat. That forecast has an accuracy problem, and MAD or MAPE will show it.
A forecast that misses 3% high four quarters running has a smaller accuracy problem and a bigger structural one. It is off by a fixed lean you could remove with a calibration factor, and the running sum is the only summary that surfaces it.
What a tripped signal means
A tracking signal outside the band says an assumption inside the forecast is wrong in a fixed way. ORM's read on why forecasts fail is that something in the business or the market changed and the forecast is still built on old assumptions, and a persistent one-directional lean is what that looks like in the data.
Reset the model against recent actuals rather than applying a permanent haircut, then keep the signal running. Track it every period alongside forecast accuracy so the lean is caught in quarter two rather than quarter four, and fold the result into your sales forecasting review. For the process changes that follow a tripped signal, see sales forecasting best practices.
Frequently Asked Questions
What is the tracking signal formula?
Tracking signal equals the running sum of signed forecast errors divided by mean absolute deviation over the same periods. The signed errors keep their direction, so misses in opposite directions cancel while misses in the same direction accumulate.
What tracking signal value should trigger a review?
Plus or minus 4 is the conventional control limit. Inside that band the misses look like noise. Outside it, the forecast is leaning in one direction often enough that the lean is structural and should be corrected rather than tolerated.
Why use tracking signal instead of just looking at forecast bias?
Bias for a single period tells you nothing, because any one quarter can miss in either direction. Tracking signal accumulates the direction across periods and scales it by typical error size, so it separates a run of same-direction misses from ordinary variance.
What does it mean when tracking signal equals the number of periods?
It means every miss ran the same direction. When no error cancels, the running sum equals the sum of absolute errors, so the ratio to mean absolute deviation reduces to the period count. Four straight high quarters produce a tracking signal of exactly 4.0 regardless of the dollar amounts.
Put these metrics to work
ORM builds custom revenue forecast models that turn concepts like tracking signal into prescriptive action for your team.
Schedule a Demo