Recursive BOM in IBP and it's impacts on TS Supply Optimizer

[eddyizaguirre3]

Hello community,

I'm having an issue where my optimizer is pre-building too much of a finished good and some it's component on two levels.

My inventory costs are layered according to the level of the components, in a way that for example.

Holding cost FG > Holding cost Comp Level 1 > Holding cost Comp Level 2.. and so forth

After examining the Production source Item and exploding the BOM to it's lowest level I found some recusiveness in the model. Meaning a finished good as a component of it's component.

This is mostly due to the use of scrap materials.

My logic tells me that IBP and most time series models in general are not suited to handle recursive BOMs the way PP-DS does, for example, and most likely this problem should be solved by taking it out of the model.

However, I wanted to confirm my assumption is correct before suggesting to remove this.

Have anyone ever experienced this behavior on the optimizer?

Thanks and happy beginning of the year 2021 to everyone!

[carsten_schumm · Accepted solution]

Hello Eddy,

cycles in the network and recursive BOM are supported in the optimizer. Sometimes such setups can yield to strange result. E.g. transporting back and forth to avoid inventory holding.

Do you use subperiods of supply for the FG? If yes, please try a run without it, to check if that may cause the unwanted inventory.

Additionally please make sure that the sum of the inventory holding costs for all components is less than the inventory holding costs for the FG. Otherwise it may still be cheaper from a total cost perspective to store only the FG.

Kind regards, Carsten

[eddyizaguirre3 · Accepted solution]

Hi Carstem,

Thank you for your response.

I'm not using subperiods of supply on this case.

That's why I was asking. If my BOM is recursive and the recursiveness takes place in the same location, It might break the rule that the Holding cost FG > Sum (Holding cost Components), because one cost needs to be higher than the other.

Wouldn't it depend on the component coefficients? Am I right to think that another rule that needs to be in place to avoid pre-building is that

Component coefficient (Comp used in FG) > Component coefficient (FG used in Comp)? See case 2 of my example

Cheers.

[carsten_schumm · Accepted solution]

Hello Eddy,

yes of course the coefficients must be considered in the cost setup. The coefficients itself must not be order - only the product of coefficient and holding cost is relevant.
Since FG is no real component for production of SFG it is also fine that FGs holding costs are higher than SFGs.
Goal seems to be to hold SFG in that example if excess is needed.

When you don't use subperiods of supply there must be another intend for the optimizer to hold more inventory than the inventory targets.

Do you have large initial inventories? Or some resource or lot-size limitations that require a pre-build to satisfy the demands? Any adjusted values for these products?

Regards, Carsten

[eddyizaguirre3 · Accepted solution]

Hi Carstem,

You were right. At the end the recursiveness was not the cause of the pre-building. All the rules we had discussed were there.

Rather, I had a Minimum Capacity Usage on my first two periods on one of resource of the FG that I failed to spot.

Still I am glad I asked because now I know which types of input could impact a recursive BOM.

Thank you for the insights :)!

[ayanbishnu1981]

Hello Eddy

Your requirement seems more like a combo planning scenario where the combo FG1 has a component which itself is similar Finished goods e.g FG2 and FG2 has a component which which can gain be a 3rd FG lets say FG3 or even SFG1 which is a semi finsihed goods.

The reasons for which optimizer decides to prebuild is not just based on inventory costs like holding, but it also takes in to consideration safety stock violation cost, cost of production and and available capacity. Based on what you mentioned is that optimizer is prebuilding a lot for FG1, FG2 and FG3 as well.

So please try out the below options based on your requirement feasibility.

1) Use aggregate constraints and specify a max value for production in the KF MAXAGGPRODUCTION (Maximum Aggregate Production Receipts) or use the KF MAXAGGINVENTORY (Maximum Aggregate Inventory) for specifying a upper threshold limit for production or inventory on a particular period. optimizer will never cross the MAX limit under any circumstances.

2) Try to model this requirement without using production source item MDT for FG1 and FG2. If you do that you have create a custom KF to generate the dependent req for FG2 based on the req in FG1 and same for the next level of FG and then update the total requirement ( custom kf3 dependent Demand + custom kf4 independent demand ) into the standard input KF INDEPENDENTDEMAND@PERIODPRODLOC.

Let me know how that works. Please keep me posted.

Regards

Ayan Bishnu

[eddyizaguirre3]

Thank you Ayan for your reply,

It is a very interesting way of solving that problem. Luckily, my case is simpler than that. See my comment to Carstem

Cheers

[ayanbishnu1981]

What you mentioned in the comment about recursiveness is right to some extent. and based on the cyclic reference and inventory cost dilema the MILP mostly will generate some strange results, which afre correct mathemetically but wont make any sense from a real business scenario.

Also as mentioned by carsten.schumm for ssub period of supply, are you using safety days in optimizer based on SUBPERIODOFSUPPLY and lotsizing procedure as 1 in LOCPROD MDT ? If yes then can i assume that you are running optimizer at weekly mode and you have maintained NOOFSUBPERIOD as 7 for all weekly bucket ?

Also are you execuitng optimizer at some other level like month whereas you still have ur input/output KF at TWK ?

[eddyizaguirre3]

Thanks Ayan. I found the issue!

[ramguevara]

I think the scenario you are describing is called co-production. I have no experience in this area, but seems there is support for this but specific caveat that its not allowed at the same time (by time series ratios). Here is the link describing it. Not sure if it answers but hope it helps

https://help.sap.com/viewer/feae3cea3cc549aaa9d9de7d363a83e6/2011/en-US/feed3835e25f485c8160fbad2945...

[eddyizaguirre3]

Hi Ram,

Thank for your kind response. Unfortunately the scenario I am describing is not a co-production one. It's just the case when You have a Finished good that is a component of it's component. To put it simply

Imagine you have a finished good and it's BOM in IBP.

in the production source item master data. You would have a Product ID (The component), and a source ID as key.

In that case

Product ID Source ID

CP1 FG1_1001

But in the same master data you also have

Product ID Source ID

FG1 CP1_1001

Cheers,