Forecast error measure -When to use MAD or MAPE ?

[former_member696146]

Hello Experts,

Can you please provide your guidance on when to use MAD or MAPE for calculating forecast error measure ? Can we say MAD measure is to be used when Low and intermittent volume of data are to be processed and MAPE to be used when processing high volume of data ? If yes,then what is the definition of low,intermediate and high volume of data ? Can you please help me understand the same ?

BR,

Praveen

[former_member696146]

Hello Experts,

My question is relevant to forecast error calculation in IO module of SAP IBP.

BR,

Praveen

[lev_degtyarov · Accepted solution]

Hello Praveen,

see below:

Average Demand Interval - this is average number of periods between sales. If you have history of 52 weeks out of which sales are available only in 10 periods, then ADI = 52/ 10 = 5.2 = 6 weeks.

Switching ratio - this is percentage, that measure number of cases when you switch from non-zero to zero values and also backward in from zero to non-zero values your sales history

Forecast Timing accuracy - it shows you a percentage of periods when both non-null forecast & non-null sales are exited.

Regards,

Lev

[former_member696146 · Accepted solution]

Hello Lev,

Thank you on your explanation.Now i am able to understand the relevance of "Average Demand interval" to that of "Sales frequency" and that makes a lot of sense.But I am not completely able to get the point about the 'Switching ratio" and "Forecast timing accuracy".How is this related to Sales frequency ? Can you please clarify ?

BR,

Praveen

[Private_Member_467521]

Hi praveen_ibp,

Here is the conceptual explanation

MAPE
The MAPE (Mean Absolute Percent Error) measures the size of the error in percentage terms. It is calculated as the average of the unsigned percentage error, as shown in the example below:

Many organizations focus primarily on the MAPE when assessing forecast accuracy. Most people are comfortable thinking in percentage terms, making the MAPE easy to interpret. It can also convey information when you don’t know the item’s demand volume. For example, telling your manager, "we were off by less than 4%" is more meaningful than saying "we were off by 3,000 cases," if your manager doesn’t know an item’s typical demand volume.

The MAPE is scale sensitive and should not be used when working with low-volume data. Notice that because "Actual" is in the denominator of the equation, the MAPE is undefined when Actual demand is zero. Furthermore, when the Actual value is not zero, but quite small, the MAPE will often take on extreme values. This scale sensitivity renders the MAPE close to worthless as an error measure for low-volume data.

MAD
The MAD (Mean Absolute Deviation) measures the size of the error in units. It is calculated as the average of the unsigned errors, as shown in the example below:

The MAD is a good statistic to use when analyzing the error for a single item. However, if you aggregate MADs over multiple items you need to be careful about high-volume products dominating the results

Topic originally published in forecastpro.com

Cheers,

Luis

[former_member696146]

Hello Luis,

Thank you for your explanation.

Can you please give the definition of low,intermediate and high volume of data ? How do i categorize datas to be analysed as low,intermediate or high volume of data ? Please help me understand the same.

BR,

Praveen

[lev_degtyarov]

Hello Praveen,

If you asking about application "Manage Forecast Error Calculations. Inventory optimization" in Fiori then see below the key differenceies.

Regards,

Lev

[former_member696146]

Hello Lev,

I am aware that forecast error CV would remain the same, whether MAD or MAPE is used.Can you please help me to understand as what is the forecasting error measure commonly used across manufacturing industries with regards to IBP IO? MAD or MAPE ?

On the point share by you,definitely the third point (stable or fluctuating demand),would be one of the key pointers in deciding the forecast error measure to be used for determination of forecast error CV calculation.Thank you very much for this hint.

BR,

Praveen

[lev_degtyarov]

Hello,

MAPE is more often used in industries.

MAPE is easy to understand rather than MAD as it was mentioned above Luis De Azevedo Freitas

Regards,

Lev

[former_member696146]

Hello Lev,

Following is my understanding :

Please see attachment given mape-vs-mad.png

Is point 4 one of the main influencers for deciding the forecast error measure whether to go for MAPE or MAD ? Can you please share your opinion,since i am not able to decide as whether the available dataset with me is categorized as low,medium or high volume ?

BR,

Praveen

[lev_degtyarov]

Hello Praveen,

you can use sales frequency classification in Forecast Error profile. Create a new profile for only sales frequency and run it before CV calculation.

If most of your data combinations has 'INTERMITTENT' pattern (like in spare parts, drogerie retail, etc) then go with MAD, otherwise MAPE. Usually I stay with MAPE based calculations.

Regards,

Lev

[former_member696146]

Hello Lev,

I was trying to get into some kind of understanding on the screenshot shared by you.I believe few keyfigures (Demand interval ,switching ratio or timing accuracy) which determine the "sales frequency" are calculated according to certain logic.I was able to find only IO average historical demand and IO average historical demand CV as part of standard keyfigure given by IBP but not the other.Can you please help me with the calculations with regards to same ?

Appreciate your efforts on clarifying query of mine.Thank you.

BR,

Praveen

[former_member1095518]

Hi Lev,

Thanks for the explanation! now I am trying to find out why the constant 1,25 in the CV calculation with the MAD

JM

[lev_degtyarov]

Hello Marin,

here is one of the mathematical approximation for STDEV = 1.25 * MAD.

• https://blog.arkieva.com/relationship-between-mad-standard-deviation/

Also consider well-know equation that CV = ST.DEV ÷ Mean , therefore:

CV = 1,25 * MAD ÷ Mean

In IO module for the mean the following was taken: max (mean forecast, mean sales).

Best Regards,

Lev

[Private_Member_467521]

Hi praveen_ibp,

Real world says MAPE can apply easily to both high and low volume products.

However, when the ACTUAL is very low, it makes MAPE have extreme values, as the ACTUAL is the denominator in the equation. So, if you start from a conservative (naive) forecast, you will have to determine what are acceptable levels of deviation on MAPE, usually based on the ugly face your boss give you, when you show him the forecast using MAPE - just kidding. If the values are extreme because ACTUAL is too low, you will have to recalibrate it based on your forecast deviation history. Doesn't matter how much math you put into the model, you always have the historic information to guide you and you don't want to be in a spot to explain why numbers are so far off.

I am afraid you won't find any lecture that give you the definition you are looking for, just because that is inherent of your own model.

Cheers,

Luis 🙂

[former_member696146]

Hi Luis,

Please help me with my query :

1)Can MAPE be also used for low volume products since i am getting varying information on the same?
2)Also what is the general forecasting error measure used commonly across industries? MAD or MAPE ?

Please give your expert opinion on the same.

BR,

Praveen