Outlier Correction - Median vs Ex-Post methods

Hi APO-heads,

I understand mathematically how both these methods behave, but does anybody have general rules about which products respond better to either method?

For example -

- Seasonal products should use ex-post method; or

- products with high random variability should use median method

- etc

Also does anybody have experience tinkering with the customising parameters for outlier correction in the IMG (setting the tolerance to outer limit and setting MAD calculation to use all periods in the horizon, not just to current period)? Did these have a positive effect and under what circumstances?

Thanks all,

Sam