Automatic reorder point calculation in mrp type v2

[yesasvi]

Hi

Do you know on how the SAP system calculates the automatic reorder point calculation for the mrp type v2.

It says uses integrated forecast sytem-is there a formula I can use.

Does the reorder point conside Planned delivery time+ goods receipt time instead of RLT?

Also how does the safety stock calculation takes place.

Yesasvi

[SAP4ME]

- RLT is for availability check only, and not for any other calculations.
- For automatic safety stock and reorder point calculations, the pre-requisit is that material has historic consumption values, has forecast values (forecasting carried out), and has service-level-% specified.
- I am guessing that automatic reorder point = sum of forecast values for the period Planned delivery time + goods receipt time.
- I am also guessing that automatic safety stock = (max forecast value - above reorder point) * service level.
- Until you build up historical data, you may use dynamic safety stock functionality (coverage profile).

[yesasvi]

Let my try and do the math of it and will let you know if the formula works.

[LTJ]

Hi,

here is some material we made for our end users regarding consumption driven parts. It might be of help?

Regards,
LTJ

Constant Model – D – Without Initialization

There are two ways to run any model – with or without initialization.

First, let’s try it without initialization. This means you do not have an X in the initialization field on the forecasting view, and thus makes the periods for initialization field irrelevant to this example.

The data needed for this example is: Alpha factor, Delta Factor, Previous Basic Value, Previous MAD, and of course Lead Time and Service Level. Let’s use an alpha of .5 and also a delta of .5. The rest of the figures are stated above.

First, let’s remember the D is a constant model, meaning the forecast we calculate will be the same value for all future periods.

Basic Value

So let’s calculate the forecast for the new month. The formula for this is:

(Alpha * consumption for period –1) + (1 - alpha) * Previous basic value

or

(.5 * 200) + (1 - .5) * 170

100 + 85 = 185 – which is then the forecast for each of the next 6 months, since we have 6 in the forecast periods field above.

Let’s look at our usage and forecast table:

Period Usage Forecast Error Total
6 185
5 185
4 185
3 185
2 185
1 185
-1 200 170 – Previous BV 30
-2 300
-3 150
-4 220
-5 180
-6 250

Error Total

For this model without initialization the error total is simply = Consumption Qty – Basic Value, for the previous period. In this case 200 – 170 = +30.


MAD

Previous MAD was = 20. Since we now have period –1’s usage, pretending that we just finished that month, we can see how accurate the previous basic value was. In this case the error total is +30. so that means the new temporary MAD is also 30. Now we can adjust this value by the smoothing factor called Delta.


The formula for the new MAD is (Delta * Temp MAD) + (1 – Delta) * Previous MAD

or

(.5 * 30) + (1 - .5) * 20

15 + 10 = 25

So you can see that a delta factor of 1.00 would always return the newest MAD, and not take into account the previous MAD at all.

Safety Stock

Now that we have our MAD, we can calculate Safety Stock.

The formula for a purchased part is: (SQRT (Lead Time * 5/7 ) / 21.4) * MAD * K

SQRT is the square root. Warning - when converting the Lead time into workdays, you must round up before dividing by 21.4 and doing the square root calculation. So in our example – (10 * 5/7) = 7.14 Round up to 8 and then divide by 21.4 and take the square root = .6114

MAD is the new MAD value that we just calculated = 25

K is a the service level factor. Below is a table that converts the service level desired into a the K factor. For our example we used a service level of 95%, so we use a K factor of 2.06.

Serv. Lev K Serv. Lev K
50 0 92 1.76
55 0.16 93 1.85
60 0.31 94 1.95
65 0.49 95 2.06
70 0.65 96 2.23
75 0.84 97 2.37
80 1.05 98 2.56
85 1.3 99 2.91
90 1.6 99.5 3.2
91 1.68 99.9 4



So the formula reads with the values: SQRT((10 * 5/7) / 21.4) * 25 * 2.06

Or SQRT(8 / 21.4) * 25 * 2.06 or 31.49 or 31


Reorder Point

Now that we have our Safety stock and our forecasted value for period 1, we can calculate our Reorder Point (ROP).

The formula for ROP is forecasted daily usage * Lead time + Safety Stock

Or

Period 1 Forecast * (Rounded up Lead Time in workdays / 21.4) + Safety Stock

So for our example it is:

185 * (10 * 5 / 7) / 21.4 + 31.49

(10 * 5 / 7) = 7.14, so round up to 8.

then 185 * 8 / 21.4 + 31.49 = 100.64 or 101.

Tracking signal

Tracking signal is used to measure how accurate the forecast is. If it exceeds the tracking limit entered on the forecasting view, exception messages will be created for you to review the forecast. To calculate the tracking signal, it is = | Error total / MAD |

Which is the absolute value of the error total / MAD. In our example it is = | 30 / 25 | = 1.2


Constant Model – D – With Initialization

Now let’s see how to do the constant model with initialization.

Basic Value

Everything changes with initialization. We no longer will use the previous Basic Value and the previous MAD from last month’s calculation. We still will use alpha, delta, lead time, service level, and now we will also use historical periods and initialization periods.

Let’s assume we use the same alpha and delta values of .5 from the previous example. As stated earlier, historical periods is 6 and initialization periods is 2.

So, first let’s find out what the forecast should be. However, in order to come up with a forecast for period 1, we have to calculate what the forecasts would have been in the previous periods (-1, -2, -3 for example) based on the criteria we have entered. Creating past due forecasts is called an ex-post forecast in R/3. The screen shot below shows how the results are a little different with initialization, mainly that there is now data in the ex-post column.



Here’s a refresher of our consumption data:

Period Usage Forecast
1
-1 200 ??
-2 300 ??
-3 150 ??
-4 220 ??
-5 180
-6 250


We have already chosen 6 historical periods and 2 initialization periods. The 2 initialization periods tell us that we will use the 2 earliest periods, in this case periods –5 and –6 as a starting point. What this means is we will take the average of the initialization periods to come up with our forecast for period –4. So looking at our data, we can see that (250 + 180) / 2 = 215 is our forecast for period –4.

Period Usage Forecast Error Total
1
-1 200 ??
-2 300 ??
-3 150 ??
-4 220 215 +5
-5 180
-6 250

So, now we have to figure out what the forecast is for period –3. In order to do this, we now will use the alpha to calculate it, just like without initialization, which is:

(Alpha * consumption for period –4) + (1 – Alpha) * Forecast for period –4

or

(.5 * 220) + (1 - .5) * 215

110 + 107.5 = 217.5

If we continue this logic up to period 1, the table now looks like this:

Period Usage Forecast Error Total
1 220.94
-1 200 241.88 -41.88
-2 300 183.75 +116.25
-3 150 217.5 -67.5
-4 220 215 +5
-5 180
-6 250

So, round up to 221, and this is the new basic value and also the forecast that will be used in the next 6 periods.

Error Total

In this case the error total is the sum of all the non-initialization periods’ error totals. Since we used periods –5 and –6 for initialization, there is no forecast and therefore no error total to use. So if we sum up the error totals for periods –1 through –4, we get an error total of 11.87, which is rounded to 12.


MAD

Ha ha, now it gets fun. Let’s start with the table we left off with.

Period Usage Forecast Error Total Abs Error
1 220.94
-1 200 241.88 -41.88 41.88
-2 300 183.75 +116.25 116.25
-3 150 217.5 -67.5 67.5
-4 220 215 +5 5
-5 180 215 -35 35
-6 250 215 +35 35

But let’s add in the “forecastâ€

[Rick]

Interesting. Thanks LTJ, you sure are a nice guy. It must have taken you a while to come up with that!

- Rick

[yesasvi]

wow..thnks a lot for explaination.....

[stevo]

Here’s a refresher of our consumption data:

Period Usage Forecast
1
-1 200 ??
-2 300 ??
-3 150 ??
-4 220 ??
-5 180
-6 250


We have already chosen 6 historical periods and 2 initialization periods. The 2 initialization periods tell us that we will use the 2 earliest periods

Hi LTJ,

I have a question relating to Initaization Periods. I just want to confirm how this affects the forecast when used with/ without initilization.

We are using model G: Moving Average. If we are forecasting for the next 12 months using 12 historical periods and only set the initialization periods to 3.

Are you saying that this will only always take the 3 oldest periods (-10, -11, -12) into consideration when determining the forecast requirements? How would you expect the system to react if the Initaization Periods were blank? Finally, if the initialisation flag was fixed "X" in the forecast profile, would this impact on the forecast results using this model?

[Rick]

The example given by LTJ is pertaining only to the constant model.

Using initialization periods with the flag should not have any impact your model G.

If you weren't using moving average or weighted moving average, and were using initialization periods of 3, then the system would take into account the oldest 3 periods during the initialization "phase" of the forecast calculation, and then it will take into account the remaining future periods when doing the rest of the forecast calculation.

[stevo]

The example given by LTJ is pertaining only to the constant model.

Using initialization periods with the flag should not have any impact your model G.

If you weren't using moving average or weighted moving average, and were using initialization periods of 3, then the system would take into account the oldest 3 periods during the initialization "phase" of the forecast calculation, and then it will take into account the remaining future periods when doing the rest of the forecast calculation.

Thanks Rick,

Are you then saying that forecast initialisation is NOT required for model G: Moving Average?

[higunj]

Hi LJR,

Could you include the initialization detail on MAD for the example which you have shown in this thread. It will go a long way in making our understanding clear

Regards
Gunjan Kumar

[angelares]

to LTJ
Can you please explain netx part - MAD in constant model with initialization and others like safery stock, reorder point etc.

regards, angel ares

[Danger]

Hi

Could someone explain me where does the value 170 come from ? In the (.5 * 200) + (1 - .5) * 170

Greetings

Danger

[Baz]

locked! necroposting = bad!