Quote from Jaguar-APS on 8. 6. 2020, 14:33

Kesten wrote us the following e-mail:

How should "forecast error" be defined? A Google Scholar search for 'error "forecast minus actual"' and for the reverse formulation turn up 51 and 46 hits respectively-almost a tie.

I asked a small sample of senior IIF members (two) whether they prefer to define forecast error as A-F or as F-A, and why? Again, opinion was divided with one preferring the A-F definition as derived from the basic statistical model formulation of A=F+e and the other preferred the more intuitively appealing F-A whereby a positive error means that the forecast was too high.

I'd like to know what you think: Which do you prefer, and why?

Regards,

Kesten

The usual explanations for error e=A-F (actual - forecast) include the following:

in statistical terms, the forecast is an expected value. A deviation in statistical computations is actual minus mean or other expected value. Thus, e = A - F is consistent with standard statistical calculations, actual - mean.

In planning and control settings, the sign of the deviation can be important in the context of a negative feedback control loop.

Also, as pointed out previously, the relationship of

Actual = Forecast + e

is an additional advantage of this definition.

There are other explanations. Finally, all of the above explanations do not preclude the opposite definition, but an additional operation (subtraction) would be necessary to make e = F - A operable in the above settings.

Based on my observations and experience, the A-F has been the most widely used formula. The most important thing though is - commonality / standard that everyone understands. With the Global S&OP being implemented at fairly aggressive rate lately, the most important thing is an alignment on what individual KPIs collected globally mean. Having different ways of calculating forecast error within one set of aggregated measures mean that one will never know what the real situation / performance actually is.