Service level type-Order fill rate

[former_member696146]

Hello Experts,

Can you please clarify on my following query ?

Whether with time series data is it possible for me to use the service level of type,order fill rate ? Looking at the definition of Order fill rate, the understanding is that ,the service level calculations considers the customer orders delivered in full from available inventory.But only with order series data, the individual customer orders might be loaded into IBP system from execution system and the IO module should be in position to calculate the service level.Do i have the right understanding ?

Thanks,

Praveen

[lev_degtyarov · Accepted solution]

Hi Praveen,

IO algo calculates on time-series buckets. No orders data.

But overall concept of service level and safety stock is based on BIG data. This is statistics.

That means some times you can go below desired service level, some times above. But on average, let say if you look at bigger number of time buckets (> 20 weeks) it is expected you will achieve desired service level.

Here is an example for fill-rate service level and also non-stock out probability. Bot types of service levels are available in IBP.

Of course you can also use fill-rate or non-stock out probability metrics for particular order.

Regards,

Lev

[former_member696146 · Accepted solution]

Hello Lev,

You had given me a very detailed and wonderful explanation and this helps a lot .Thank you for your efforts.My query is exactly on the point that you had mentioned in the statement :

"But overall concept of service level and safety stock is based on BIG data. This is statistics"

For instance, based on the actual quantity and forecasted value (of the past),i am arriving at forecast error cv (or measure of forecast accuracy) .IBP software can propose me a safety stock numbers considering the probability of stockout ( normal distribution).This is understood and agreed.Similarly in case of "Available in full" service level type,how will the system propose me a safety stock considering only my target service level ? Am i required to load any past service levels or NSP information and service levels achieved or any history data for modelling out the normal distribution curve (something of the kind,I do not have experience here),with which system will calculate the required safety stock numbers ? Even for lead time variability the past figures are helpfull in determining lead time error CV with which the safety stock can be determined.I am hoping that i am able to put across my concerns in right way.Please let me know if you need more explanation on the same ?

Thanks,

Praveen

[lev_degtyarov · Accepted solution]

Praveen,

sorry, didn't understand you questions completely. You mix a lot of things in one post.

But I will try to answer.

ONE. Yes, typically you can load historical data for:

- actual customer service level

- actual available on-hand stock

- actual safety stock

- actual lead time, lead time CV

All these key figures can be used as part of analysis. For example a) you can compare IBP calculated safety or IBP on-hand with your actual stock level data b) historical service level can help you to take a decision on target customer service level.

TWO. If you do not calculate actual service level in external system, you can also customize IBP and calculate it directly in IBP:

- load actual available on-hand stock (it should be custom KF)

- load customer requested quantity aggregated as time-series per on weekly buckets (it should be custom KF)

- create custom calculated key figure\-s for actual service level calculations

THREE. As of service level type: there are 2 approach how to measure it - fill rate and non-stockout probability. IBP understand both types. You can choose one of it depending on how customer measure it. Check your LOCATIONPRODUCT master data, attribute SERVICELEVELTYPE. Possible values are 'A' - Available in Full (Non-stockout probability) (used by default in case 'null' value) or 'F' - Fill Rate

Then you define target service level in TARGETSERVICELEVEL key figure. After that you run IO which calculates safety stock for given target service level and given type (A or F).

Regards,

Lev

[former_member696146 · Accepted solution]

Hello Lev,

Let me put my question with more clarity.

My understanding on your explanation is as follows :

One : The Key figures you had mentioned is purely for analysis purpose.

Two : This point is more on procedure for calculating the service levels .Steps we need to adopt for calculating service levels in IBP system (alternate could be BW system)

Three : Two different service level types available and we need to opt for one based on customer requirement.

I am clear on these points as you had clearly explained to me and thank you again on your explanations.

My Query :

I understand IO is able to calculate safety stock with the user inputting the target service level and also service level type.For IO to calculate the safety stock based on the service level target requirements ( e.g could be 90 % ,95 % or 99%),do we need to load any history data (say sales order or service level history etc). Based on which values(or inputs) are the normal distribution curve modeled ?.Currently only actual sales history and also the forecast history from the past are the only datas loaded in the system(specifically for the purpose of forecast error CV calculation).

BR,

Praveen

[lev_degtyarov · Accepted solution]

Praveen,

See below regarding historical data that you may use...

Decisions are up to you what to load and what to customize.

As of me I always do 2, 4. In rare cases I do 1, 3 if BW is not able to calculate required data.

Step 1. Setup target service level (by default you DON'T NEED to load history there)

if you know what service level you want, then just put it in key figure TARGETSERVICELEVEL or in Master data 'Demand Stream'. And also setup service level type in master data 'Location Product'. You don't need to load historical data for this.

But quite often customers say they don't know what is their service level. And this case you need to estimate\calculate it. It can be done in IBP with some customization (historical orders and on-hand stock load are required for this). But typically customer calculates their actual service level externally, like you sain in BW for example

Step 2. Calculate forecast error coefficient of variation (by default you MUST load history there)

You can setup forecast error CV in 2 ways:

a) load historical sales and historical forecasts and run calculations in IBP. This is the best option as IBP IO has some very good features when calculating it (bias consideration, outliers exclusion, intermittency, distribution curve - normal\gamma, etc.)

b) you calculate forecast error cv externally, let's say in BW. Then you setup calculated values in IOFORECASTERRORCV.

Step 3. Actual Lead Time and Lead time coefficient of variation (by default you DON'T NEED to load history there)

I already explained it to you here: Lead time error and production lead time error calculation in IO module

Step 4. Historical actual oh-hand stock and safety stock (by default you DON'T NEED to load history there)

It is not considered by IO optimizer. You can load this data and use it for reporting, analytic and comparison purposes. E.g.:

• What is my current safety stock?
• What is IBP safety stock?
• What is delta between them?

Regards,

Lev Degtyarov

[former_member696146 · Accepted solution]

Hello Lev,

Following is going to be my approach :

Step 1: Target service level would be inputted manually,calculation of the historical data would happen in S/4 HANA system.

Step 2 : Procedure going to be adopted is on similar lines as you had suggested.

Step 3 : Your explanation was precise and we would be following the similar approach

Step 4 : Also your suggestion is good and we would adopt the same as well.

So based on your explanation,Can i understand this way ?

Apart from the history data requirements as explained by you in the comment section,we would not need any other history data for safety stock calculations to manage the "Order fill rate" requirement of the customer.In other words, the 'Z factor' which is used for determining the safety stock numbers does not require any history data to be loaded and is a standard followed globally.

Thank you very much for your explanation till now.

BR,

Praveen

[lev_degtyarov · Accepted solution]

Hello,

your understanding is correct

regards,

Lev

[former_member696146 · Accepted solution]

Hello Lev,

Thank you for your detailed explanation.Appreciate the same.

BR,

Praveen

[lev_degtyarov]

Hello!

In standard configuration IO is not calculating customer service level. Key figure TARGETSERVICELEVEL is an input parameter for IO.

Usually we measures actual service level outside IBP (e.g. in BW). Then we source actual service level from BW to IBP (e.g. actual SL = 93%). Based on actual service level value user may decide: to use actual as target (e.g. target SL = 93%) or user will use changed value (e.g. target SL = 95%).

By the way you can also create a custom IBP master data for orders and custom IBP key figure for on-hand available stock levels. Then load data via CPI-DS interfaces. Then it is pretty easy to calculate actual SL directly in IBP IO. Be careful with data volume you want to load.

So, finally it is up to u & customer to decide do you want to measure actual SL in IBP or in external system like BW. I like the latest.

BR,

Lev

[former_member696146]

Hello Lev,

Thank you very much for your wonderful suggestions.

I wanted to understand the relevance of using "Order fill rate" for the time series based data ?

Let us assume customer wants to maintain the target service level as 95 % and i assume system(IO algorithm) would propose me a safety stock of say 100 pieces.The question is that how does the system suggest the 100 pieces without having the customer order informations in hand (weekly demand forecast is alone the demand input that i am proposing to suggest to the business) since ,time series data is used and not order series data. That is where my doubt is whether Order fill rate is relevant for time series data ? Does this make sense at all.Is there some thing else that i am missing ? Please give your expert comments on the same.

Thanks,

Praveen