Inventory Management and Warehouse Operations.


Component Safety Stock

 By Dave Piasecki

Despite what you may have been taught, there are valid reasons for carrying safety stock at the component / raw material level. In my book Inventory Management Explained, I purposely stayed away from this topic (other than a brief mention of it) because it can get very complicated and with the amount of complexity already in the book, I figured it was better to not open up this can-o-worms. To adequately cover dependent demand safety stock probably requires a whole separate book, and I'm not exactly eager to get started on that one. But there are a couple of points I figured I could touch on in an article, so here it is.

Using safety lead time?

Before you attempt to determine safety stock for components, you need to determine why you need it. If you have safety stock at the finished goods level, that is typically calculated to cover variability in demand that occurs during the lead time (typically the manufacturing lead time). And while it isn't specifically calculated to cover component lead times, it does actually provide some coverage there. Unfortunately, since almost all inventory/MRP systems treat safety stock as inventory that you should never plan on actually using, it doesn't really "buffer" the impact of demand variability as far as the way it plans component requirements. For example, if in week #3 you expect to have 150 units in stock, but 50 of those units are listed as safety stock, it basically treats it as 100 units in stock. So, if you get customer orders for 120 units in week #3, it's going to tell you to make more even though you clearly have enough. This is one of those cases where you kind of want it to work that way, but also kind of don't want it to work that way. But for now, we'll just accept that it does work that way, so we now need to be able to cover that demand at the component level.

Just to clarify, I don't mean to imply that everything we are doing here is the result of a flaw in MRP. There absolutely are times when your finished goods safety stock does not provide adequate coverage at the component level.

So let's take this to the next step. We already have safety stock at the finished goods level, and we want to make sure we aren't duplicating that at the component level. But we do need to cover for the impact finished goods variability has on our component demand. In order to cover for this "impact", we need to better understand it. In a typical make-to-stock manufacturing operation, you produce items in batches. These batches typically have fixed lot sizes. So basically I would have my system set up so that whenever MRP determines I need more of Item X, it will create an order for the lot size I have set up (we'll say 500 units for this example). It will create these "planned orders" as far out into the future as my forecast extends.

If my manufacturing lead time for this item is one week, and I use one each of Component Y to make Item X, I will show 500 units of demand for Component Y one week prior to each planned order for Item X. This schedule is what I will base my orders (deliveries) for component Y. Now here comes the important part. What happens when I experience demand for Item X greater than the forecasted demand? Does it increase the quantity of the planned orders, and therefore increase the quantity of the planned demand for Component Y? In most cases, the answer is NO. What typically happens is it simply moves the planned orders up in the schedule. So instead of it showing I need 500 units of Component Y in week #3, and another 500 in week #10, it may now show I need 500 units in week #2 and week #9.

 Week 1Week 2Week 3Week 4Week 5Week 6Week 7Week 8Week 9Week 10Week 11
Item X planned orders       500             500
Component Y planned requirements     500             500  

Here are my planned orders and requirements initially.

 Week 1Week 2Week 3Week 4Week 5Week 6Week 7Week 8Week 9Week 10Week 11
Item X planned orders     500             500  
Component Y planned requirements   500`             500    

Here is the likely outcome after experiencing demand greater than the forecast. Because we are dealing with fixed lot sizes, our order quantities do not change, therefore our planned requirement quantities do not change. Instead, they just move forward in the schedule.

And that's where safety lead time comes in. If you set your safety lead time to the typical amount of time you see your planned orders move as a result of an increase in demand, you are covered. Obviously you need to conduct some analysis to determine what this amount of time is, but you will find this works very well. If you're thinking that you could just carry the inventory equivalent of that time period in safety stock rather than safety lead time, you need to think a little harder about how safety lead time works in comparison to safety stock. Your inventory system always plans to have safety stock in stock. But safety lead time only impacts inventory levels when you have expected demand. So if I just did a production run and used all my inventory, but did not expect to do another production run for another 10 weeks, and had a safety lead time set up to 1 week, I would order more inventory to be delivered in 9 weeks. But if I had safety stock set up to the equivalent of one week's demand, my system would tell me to immediately get more in stock. There is a big difference here in how it impacts your inventory levels.

What if that's not enough?

The ongoing trend in sourcing components overseas has resulted in a scenario where you have many manufacturing companies that have short manufacturing lead times, but very long component lead times. Since the finished goods safety stock is generally calculated to cover finished goods lead times (manufacturing lead time), a short finished goods lead time results in minimal safety stock requirements. This is typically fine for the main purpose of short lead-time finished goods safety stock (random variability known as "noise"), but does not cover the variability that most greatly impacts long lead-time components. The impact of noise actually diminishes over longer time periods (like long lead times), but the impact of trend can be significant.

One way to deal with this it to intentionally bias your long-term finished goods forecast. In other words, you're going to take a very optimistic approach with your long-term forecast. So rather than forecasting your most likely sales scenario, you're forecasting your best possible sales scenario (or something in between). This will make sure you have plenty of components in the pipeline, but will not necessarily have a significant impact on your finished goods inventory because of your short manufacturing lead times. Obviously this puts you at risk of getting stuck with excess component inventory for those long lead-time components, but that's just part of the risks associated with overseas sourcing. If you have a history of dramatic trend variation or significant noise (even over long time periods), you can optionally run a statistical safety stock calculation against your component history. The trick here is that since you don't really have a component forecast, you need to use your planned order history instead. And since your MRP system probably doesn't have a "planned order history" file, you'll need to start making one.

What if that's still not enough?

Hey, I told you this gets complicated. We've only covered demand variability--and only some specific scenarios of that. If you have supply variability, or other issues (quality problems, scrap, etc.), or just some unique demand characteristics or business policies, you need to analyze those particular issues and come up with an appropriate action.

More Articles by Dave Piasecki.

Dave Piasecki, is owner/operator of Inventory Operations Consulting LLC, a consulting firm providing services related to inventory management, material handling, and warehouse operations. He has over 25 years experience in operations management and can be reached through his website (, where he maintains additional relevant information.

BookBook Banner02