Inventory Optimization SAP IBP

[a_khomne]

Hello Experts,

Can anybody justify the calculations to find out the following by changing some inputs to algorithm

• Demand Variability Safety Stock
• Supply Variability Safety Stock
• Service Variability Safety Stock

I know the basics as well as significance but don't know how exactly it is calculated.(mean to say that how service level variability is calculated by the algorithm)

It will be really helpful if someone can explain the output values considering the inputs with some example.

Thanks and Regards,

Amol Khomne

[ayanbishnu1981]

Hello a_khomne

I will try to answer the question to the best of my ability. As you may already know that the IO tool in SAP IBP is actually aquired by SAP from a company callled Smartops Corporation. In a nutshell the algorithim behind single / multi stage IO is a propietory and classified code and to people like us its a black box.I completely understand that in some cases it becomes difficult for us to explain the logic of IO to our clients. Hence I would provide all the insight that I have about this algorithm.First of all Mr. Sridhar Tayur & Martin Barkman (founder of smartops) are brilliant scholars who has multiple PHD's in the area of supply chain and inventory optimisation and they globally accredited as one of the gurus in this field and thier expertise unquestionable.But that said below are some guideline based on which the IO algorithm has been developed

• First of all the pareto rule is considered. or better known as 80/20 rule with multilple linear variables with dynamic weightage.
• It also involves EOQ (Economic Order Qty) and EMQ (Economic Manufacturing Qty) method as well for assortments
• Reorder qty determination function is alo a part of the whole logic
• It includes Poisson Distribution method as a part of the algorithm

The IO algoritm is still an evolving method where more constant and dynamic pattern based variables are taken into consideration using AI and ML which will increase accuracy of the result but at the same maing the logic more complex in nature.

What all I mentioned above are just the tip of the iceberg and its developing as we move forward.

I am not sure if I have answered your question or made it even more confusing.

Regards

Ayan

[a_khomne]

Hi ayan,

Thanks for the response.

So from your answer we can conclude that in many scenarios we will not be able to explain the calculations behind all this algorithms.

But if we can not explain about algorithm calculations then how can we decide that which input values are best for our algoriths (more to the business) such as TSL, PBR and all.

Because if we maintain TSL as 1 then Safety stock calculated are really high (beyond the bounderies we can say).

How can we decide then that which TSL will be the best for our client?

If you have some idea and can bring some light to this then please.

Thanks

Amol Khomne

[vitor_camargo1]

Thank you abishnu1981 for your great answer !

It is really hard to explain the results once the client requested.

However, I guess that the inventory optimization uses the Normal distribution instead of the Poisson because it considers the CV (coefficient of variations) as an input, which is not required for Poisson distribution. Are you sure about it?

I really appreciate discussing such as these topics that don't have formal answers to improve our skills!

Regarding,

Vitor Camargo

[ayanbishnu1981]

Hello a_khomne

To answer your question about TSL and PBR, i will try to be as brief as possible.

A TSL value of 1 will always result in very high value for safety stock as we are in a way telling the tool that we want fulfill allincoming orders without any miss. TSL value of 1 should be avoided. in case you mantain it as null system treats te value as .9999.TSL value should be based on your cost of the SKU and priority of that customer group.a high TSL for an expensive SKU means holding for SS and hence holding more working capital. Generally what i recommed to my clients are as follows.

• Lets prioritise customer group with a ABC classification based on sales volume and order repeat frequency.
• Similarly do a kind of XYZ classification of SKU based on the cost of the SKU and its contribution to the revenue margin
• for combination of customer group A and SKU X maintain a high TSL between .85 to .95
• for combination of customer group B and SKU Y maintain a moderate TSL between .75 to .90
• for combination of customer group C and SKU Z ( where most of the customer and SKu will fall) maintain a TSL between .75 to .80

PBR also has relationship with TSL. TO determine the PBR for a particular LOCPROD combination we need see what is the production cycle or replenishment cycle for that PROD. Also we need to ascertain the AVG reorder frequency at a agg level lets say customer group. Once we have those nos, then PBR value for that loc prod should be between the replenishment cycle and reorder cycle.

But it is easier said than done. My personal opinion is simulate, re simulate and resimulate with different values within the range mentioned above.

Regards

Ayan

[ayanbishnu1981]

Hello vitor.camargo

Your statement about normal distribution & consideration of CV (coefficient of variations) as an input is absoluetly correct. But normal distribution alone is not enough in the IO optimizer. Poisson distribution is a discrete probability distribution while the normal distribution is a continuous probability distribution. In simple words, Poisson distribution considers time interval and rarity of events while normal distribution shows measurement of identical independent variable.

Hence thier is no doubt that the model has normal distribution, but normal duistribution alone is not sufficient enough to develop the entire IO algorithm as we get new rare event and factors coming in where the use of Poisson distribution is needed.

[a_khomne]

Hello Ayan,

Thanks for the explanation. It is helping me a lot.

Now I am trying with all the possible values in between the ranges you specified with the simulation.

Thanks and regards,

Amol Khomne

[vitor_camargo1]

Hello ayanbishnu1981,

I got your point, it makes a lot of sense.

Thank you !