Tier 2 Material planning – using MRP data

This page will deal with:

1. Understanding the algorithm of MRP planning. Gaining deeper understanding of how it works.
2. Understanding how to do past data forecast the right way.
3. How to combine both methods and investigate the differences.

To understand the basic algorithm of MRP go to the

Basic page.

 

Understanding the MRP planning algorithm

Again we will deal here only with the demand created from the sales forecast without dealing with inventory. Inventory is a huge subject and it will be discussed on other pages. There are plenty of inventory strategies that you can use with your MRP system. But if the demand is not right, inventory strategy is having a lot of inventory….            

Our goal on this page is to understand:

1. How the MRP algorithm breaks down the sales forecast to RM (raw material) forecast
2. How to correctly use the past demand data to create another forecast
3. How to compare the 2 forecast methods and analyze the differences.

 

MRP breaks down the sales forecast to RM forecast

The MRP algorithm will take every sales forecast record and one by one “explode” it to RM (raw material) forecast records using the BOM (bill of materials).

Then it will sum up all the quantites needed for each RM at a single time period – and that will be the RM forecast for a single RM ID number at one time period.

The process will look like this:

Take every record from the sales forecast:

Use the following BOM (Bill Of Materials) to transfer it into RM records:

The table above tells the MRP algorithm that every flying car Model A will have 4 wheels and 2 folding wings .Model B will have 4 wheels and 2 straight wings, and so on.

Then by adding each item (ID number) for each time period (in this case months) we will get the following table:

The table above is the projected demand of each RM for any given time period. This demand comes from the sales projection.

(After getting this demand the MRP will use the inventory and the lead times to come out with a purchase plan, but that will be covered on a different page).

So how can we know that the sales forecast is really valid?

(We will cover how to do a better sales forecast on future pages too.)

We need to validate the demand forecast we got from the MRP by checking it against past data and try to understand the differences. We are looking to understand when it is a real change in the demand and when it is just noise that happens due to one or more reasons. Please remember that sales forecast has a different goal than material planning. The sales forecast is the revenue projection, it affects what the company wants the investors to think, it affects the salespeople’s bonuses and so on. So most of the time the sales forecast is not always what it seems. It might not be accurate.

So we need to look at a few problems that the sales forecast will have, such as:

1. Overforecast – someone in sales might overestimate a client, a region etc.
2. Underestimating – someone underestimates a client, a region etc.

3. No real knowledge – no one can really estimate the sales of a model, of a client, or anything else. Although no one can estimate, the sales forecast had to be filled in, so someone put a number in there.

All 3 of the above problems might result in overstock or understock of inventory. It is impossible to get an accurate forecast and that is OK. We just need to figure out which estimation might have a strong effect on the demand. For example, if the salesperson doesn’t know how many flying cars he will sell to Zambia next year and he picks the number 4, that is OK if he chooses a flying car model that sells 400 a year in the US. But if he happens to pick a model that has specific parts (like antifrost wheels) that only sells 20 cars a year, it will make a big difference. It is very difficult to find this kind of mistakes since we will usually have thousands or more ID numbers.

That is why we need to validate the data and this is how we do it:

 

Use past demand data to create another forecast to validate the sales forecast

Again the main idea is to compare the forecast based on sales forecast with the past demand data and look for major changes.

Can we take the actual consumption of raw materials and assume this is the past demand?

 

Just in case the above picture was not understood: NO! Don’t take the past consumption as your validation since the actual might contain extraordinary things that happened. Might contain old material parts or one or more of the following reasons:

1. Production constrains:

 –  Last month was the holiday season and the factory was closed  for many days, so there was a decrease in consumption.– Last month the production plant produce only a batch of flying car type B and the month before only a batch of flying car type A. That will have a dramatic effect on what the consumption will look like (one month it will have twice as much material and one month it will have none).

-Tool malfunction – suppose one or more of the machines stopped working for a few days/weeks. Then the actual consumption will also see a big impact.

– There are many more production reasons that can be applied here.

2. R&D changes:

– If the R&D changed the BOM, then all the past data might be gibberish. For example, if the R&D decided to change the aluminium wings to plastic wings, there is no way to compare the actual consumption to the forecast consumption. The past consumption will show a usual consumption of aluminium wings and zero for plastic wings. The sales forecast demand will show the opposite (only plastic wings and no aluminium). This might be OK if we deal with a major change like wings. The planner will know about it, everyone will know about it. But what if we only changed the cover of the car seats? What if we changed it only in car type C? How can we compare the actual consumption to the forecast consumption?

– What if the formula changed? Suppose we have an ice cream factory. We used to add 2 drops of vanilla for every 1kg of ice cream. Now the formula changed to 3 drops per 1kg. That is major change in consumption and we will not see it in past consumption.

3. Material shortages– suppose there is a substitute raw material for a specific RM. For example, different quality grade, or different concentration grade. During a shortage in the original material we will see much more consumption in the substitute raw material than usually, and the original RM will drop to zero. If we only look at the consumption, we will restock the substitute raw material, because it has an increase in consumption and we will not order the original RM because it has zero consumption. The result will be an ongoing increase of the substitute RM.

For example, suppose we manufacture lemonade juice. The basic formula is to take a 50% concentrated lemon juice, add 45% water and 5% sugar. Now let’s suppose that there are 3 time periods of shortage of the 50% concentrated lemon juice and the supplier only has 70% concentrated lemon juice. The 70% product is more expansive and not easy to use. Usually it is used for different products, but it can still make the same lemonade (by applying a different formula). If we only look at past consumption, we will see an increase in the 70% RM and a decrease in the 50% RM. The normal material planner will order more 70% RM and less 50% RM. That will cause another shortage of the 50% product, so the production will use more of the 70% product and so on.

There are many more examples of why not to use actual consumptions. So let’s get to the solution:

What past data should be used?

The data we need to take are the actual demand data. The actual demand = the customer orders. The real demand is reflected in customer orders. Since customer orders are only for finished goods, we need to use the BOM (the last and updated BOM) to translate it into RM demand. We then sum it all up for each RM ID for each time period.

We would get the real RM demand if we didn’t have shortages/production problems or BOM changes.

This method reduces noise and unexpected events that we have not planned for. We will still plan with safety stock to deal with other unexpected events, but the basic demand data would be much better than the actual consumption data.

So what next?

How to use past demand to create a forecast ?

How to compare the forecast data and past base data ?

Go to Next page