00:00:07 Intro to supply chain fill rates, service levels.
00:00:37 Distinguishing fill rates from service levels.
00:02:09 Bookstore case: fill rates and service levels.
00:03:59 Discussion on optimal service, fill rate percentages.
00:05:27 ERP system implementation of fill rates, service levels.
00:07:23 Complexities in measuring service levels, fill rates.
00:09:42 Business optimization via service level, fill rate metrics.
00:10:43 IKEA case: service level limitations.
00:12:52 Fill rate measurement and its challenges.
00:13:46 Economic drivers for supply chain measurement intro.
00:15:25 Importance of economic drivers for good fill rate.
00:16:40 IKEA case: economic drivers in inventory management.
00:17:45 Establishing customer costs and loyalty analysis.
00:19:18 Challenging economic drivers’ use in inventory management.
00:20:56 Case for economic drivers in inventory management.
00:23:35 Supply chain managers’ loyalty, ERP limitations.
00:25:02 Service level entry vs. output, reverse engineering.
00:25:29 Changing service level settings within organizations.
00:26:22 Managing service levels, difficulties in process change.
00:27:28 Business mission, economic value in quantifying results.
00:28:51 Service level optimization alternatives assurance.
Summary
In an interview, Lokad founder Joannes Vermorel differentiated between service levels and fill rates in supply chain management. He stressed that high service levels don’t always mean high fill rates, and vice versa. Vermorel suggested that optimal service levels balance inventory costs and client service costs. The accurate measurement of demand, he said, is challenging. He argued that service levels inadequately reflect business impacts and proposed considering supply chain economic drivers instead. Vermorel also encouraged a shift from traditional metrics to an economic driver approach, considering supply chain’s economic value for optimization. He warned against organizational inertia, which often results in adherence to outdated techniques.
Extended Summary
Kieran Chandler begins the interview by introducing the topic of the day: fill rates and service levels, two commonly used tools in supply chain management. Joannes Vermorel, the founder of Lokad, is tasked with clarifying these terms and their key differences. He explains that these terms have varying definitions depending on the organization, but academically, service levels represent the probability of fulfilling an incoming request, while fill rate is the percentage of the overall demand that can be served. Vermorel emphasizes the difference by noting that a high service level doesn’t necessarily translate to a high fill rate, and vice versa.
To illustrate this point, Vermorel uses a bookstore example. Suppose a bookstore has one book of interest to its patrons, and it attracts two types of customers - students who ask for one copy of the book, and a professor who asks for twenty copies at once. In this example, if the bookstore has 20 units in stock and services 20 students (one book each) and then fails to serve the professor’s request for 20 more, the bookstore has a service level of over 95% (20 out of 21 requests served) but a fill rate of only 50% (20 out of 40 units of demand served).
When asked about the ideal service level or fill rate, Vermorel explains that there’s no simple answer. Aiming for 100% in either isn’t necessarily the best strategy because higher service levels require more stock, which increases inventory carrying costs and the risk of inventory write-offs. Theoretically, achieving a 100% service level would mean infinite stock, which isn’t realistic. He posits that the ideal service level is a balance between the cost of inventory and the cost of not serving clients.
Vermorel explains that one of the main challenges is measuring demand accurately. For example, if a client requests 1,000 units but the supplier is unable to fulfill the order, the next day, the same client may return with a similar request. The question then becomes whether these requests should be counted as two separate 1,000 unit requests or as the same one. This problem becomes even more complex when the client’s requests vary slightly, for instance, requesting 1,000 units one day and 1,001 units the next, or 800 units because they sourced 200 from another supplier. This makes the measurement of demand a fuzzy and complicated process.
Service level, which measures the percentage of customer demand that is met through immediate stock availability, is also discussed. While this measure can provide useful insights, Vermorel argues that it falls short in truly reflecting the impact on customers and the business. For instance, it doesn’t account for the different impacts of being out of stock on high-demand items versus less significant items. Using IKEA as an example, Vermorel explains that the impact of a bed being out of stock is far greater than a lamp being out of stock because the bed is likely the primary reason for a customer’s visit. Service level also doesn’t reflect the business cost, as maintaining a high service level may result in a surplus of inventory, which is costly for the company.
Vermorel suggests that a more effective approach would be to consider the economic drivers of the supply chain, such as the cost of inventory, the margin achieved when a unit is successfully served, and the cost of not servicing a product (the ’non-service penalty’). This penalty, Vermorel argues, is essentially a cost to the company, as customers may seek alternative suppliers if they consistently fail to receive their requested products, leading to a loss of loyalty.
Vermorel begins by clarifying that a high fill rate, which indicates a lower probability of stockouts, doesn’t necessarily result in a more optimized supply chain. He cautions that determining a “good” fill rate isn’t straightforward as it depends on various factors, notably, the economic drivers that influence the supply chain. He illustrates this with the example of selling strawberries, where a lower fill rate is acceptable due to the perishability of the product, which necessitates daily stockouts to avoid loss.
When asked about how economic drivers would work in a practical example, Vermorel discusses the concepts of carrying cost and obsolescence cost. He explains that these factors are essential when determining if a product loses value over time, such as a product tied to a specific event like the World Cup. The difficulty arises when assessing the cost to the client if a product is unavailable, particularly in business-to-consumer (B2C) situations where client feedback isn’t readily available. In such cases, correlation analysis and common sense are employed to determine the impact of a stockout.
Chandler poses a counter-argument, suggesting that economic drivers might also be influenced by personal opinion or “gut feeling”. In response, Vermorel concedes that the process can be arbitrary but argues that it’s a more strategic approach. By focusing on economic drivers, supply chain managers are better equipped to approximate an economic model of their supply chain, rather than sticking to arbitrary service levels. Vermorel emphasizes that the economic drivers’ approach seeks to approximate “something that is approximately correct rather than exactly wrong”. He adds that the ultimate goal of economic drivers is to translate everything into monetary terms, providing a constrained range for loss estimates and relative balance between products.
Moving on, Vermorel discusses the loyalty of supply chain managers to outdated techniques such as fill rates and service levels, attributing this to the simplicity of implementation and organizational inertia. He cites that many enterprise resource planning (ERP) systems have built-in settings for service levels, which makes them easy to use but not necessarily accurate or effective. The gap between the targeted and actual service level often results in a reverse engineering process, leading to a culture of mitigating discrepancies between the two. Consequently, companies end up entangled in complex processes that make changing these outdated techniques challenging.
In the final segment, Vermorel offers guidance on how to transition to an economic driver approach. He advises businesses to first understand their primary mission and their supply chain’s economic value. This step is crucial to establish the starting point for supply chain optimization. Vermorel emphasizes the need to think in terms of economic value – dollars or euros – since this is the basis for any effective supply chain optimization.
Full Transcript
Kieran Chandler: Today on Lokad TV, we’re going to clarify exactly what they are and also discuss what you can do to reduce stockouts and ultimately keep your customers happy. So, Joannes, these two tools are often fairly confused in the marketplace. Perhaps a nice place to start is if you could just clarify what they are and also what the key differences between the two are.
Joannes Vermorel: Yes, I mean, you will find almost as many definitions for those two ideas, service levels and fill rate, as you have companies. But let’s stick to the academic definitions of those two concepts. Service levels represent the probability of being able to service an incoming request. So when you say, “I have a 90 percent service level,” what you’re saying is that nine times out of ten, when a customer, which can be an internal customer within the company, asks you for a good to be delivered, you can fulfill the request. That’s the service level.
The fill rate is different. It’s the percentage of the overall demand that you can serve. So the question is, when you say that you have a 90 percent fill rate, it means that in total you had, let’s say, a hundred units that were demanded and you’ve been able to serve 90 of them. You might be wondering whether there is any difference but actually, there can be a significant difference between the two depending on the situation.
Kieran Chandler: I see, the thing about an academic approach is it’s not always so clear. So perhaps do you have an example we could use to illustrate this?
Joannes Vermorel: Let’s say a bookstore is selling a book of interest to its patrons, and we have two types of clients. We have students who walk into the bookstore and ask for one copy of the book, and once in a while, we have a professor that walks into the bookstore and asks for 20 copies at once. Let’s say that on average, we have twenty times more students than we have professors. In terms of service level, let’s imagine that the bookstore has 20 units on a shelf.
First, you have 20 students walk in, each asking for one book. The bookstore has 20 units in stock, so it can serve all of those students. Then a professor walks in and asks for 20 books. Unfortunately, the bookstore cannot service the professor’s request. So in this case, we have something that is above a 90 percent service level. Out of 21 people, 20 were served.
In terms of fill rate, we have only a 50 percent fill rate. Why? Because the total demand was 40 units - one book per student plus 20 books for the professor. So, the total demand was 40 units, and the bookstore only served 20 books because they had only 20 in stock. So in terms of fill rate, we had 50 percent. Service level is above 95 percent, and fill rate is 50 percent. So that’s the difference between how frequently you can serve your client versus how much of the overall demand you can serve.
Kieran Chandler: So how do you sort of know what is a good fill rate or what is a good service level that you should actually work to? I mean, what is the percentage you should be choosing?
Joannes Vermorel: There is no simple answer to this question, and higher is not necessarily better.
Kieran Chandler: There is no simple answer to this question and higher is not necessarily better. Can you elaborate on that?
Joannes Vermorel: Certainly, there is one common misunderstanding that the best service level would be 100 percent. But that’s not the case. The reason being, in order to have a higher service level, you need more stock, which increases your inventory carrying costs and your risk of inventory write-offs. Mathematically speaking, a hundred percent service level means infinite stock because it signifies that no matter how unlikely the demand, you can always meet it. So, if you want to be perfectly sure that you will always have enough stock, you need something that is akin to infinite inventory, which is not a realistic position. Essentially, your service level is a trade-off between the cost of inventory and the cost of not serving your clients. That’s how you can converge toward a good service level.
Kieran Chandler: It’s something that’s implemented a lot of the time into ERP systems. So how’s that actually realized in practice? How does it work?
Joannes Vermorel: The interesting thing is that in practice, there are several angles to this. One is just having a measurement and here it gets relatively tricky. In theory, the service level counts how many times you’ve been able to service a request. But in many situations, you do not necessarily record every single request. For example, if you run a hypermarket, you won’t record that somebody was looking for a bottle of milk and couldn’t find it on the shelf. You will just record that you had a stock-out because your electronic stock record was zero at the end of the day. You will not know exactly how many clients you missed. In situations where you do not record the actual client requests, which is frequently the case in B2C businesses, service level is typically approximated as the percentage of products that are out of stock compared to the total number of products that you have in stock. It’s a bit arbitrary because you can have a lot of diversity and some products might be much more important than others. There are complications in establishing a measurement.
Even in the case of B2B setups, where you’re serving businesses and where you might actually record the requests, you can end up with a lot of bizarre artifacts. For instance, a client requests 1,000 units, you can’t serve this client, but you can record that you missed 1,000 units. However, the next day, the same client, where you failed to service their 1,000 units, comes back and asks for 1,000 units again. The question is, should you count those two requests as twice one thousand units or is it actually the same request where the client just asked for one thousand units, you said no, they tried other suppliers who also said no, and so the same client is coming back with the same request?
And it gets more challenging in the real world. The client, on the first day, is going to ask for 1,000 units and then for the second day is going to ask for 1,001. Why? Because they need something a bit different, maybe because they had more consumption from the stock, so now they need more than what they requested. Or maybe the next day they will come back to you and ask for let’s say 800 units, and the reality is that they have been able to source 200 units from another supplier but they are still short of 800 units. So the situation can be quite a fuzzy.
Kieran Chandler: It seems that the measurements can be fairly simplistic, and because of that, there’s a lot that can slip through the gaps. Is there a better way of illustrating these sort of problems?
Joannes Vermorel: I would say having good measurements is the first step of having a good optimization. The first step is to really think about what you’re measuring exactly and if it is the most desirable measurement for your business. Service level and fill rate are interesting, but they have clear limits on what you can do with them. The main problem with service level is that it reflects very poorly the pain that you inflict on your clients. Let’s take an IKEA store as an example. You have two floors. One where you find the nice furniture you’re looking for and a second floor, a little bazaar, where there’s plenty of cheap stuff. If people were looking to buy a new bed and the bed is out of stock, it’s painful for the client and for IKEA too because it was an expensive product. On the other hand, if a cheap lamp is missing from the second floor, the customer might not even notice because they weren’t coming for this product in the first place. Service level doesn’t really reflect the satisfaction of the client or your own costs because you could have a high service level but a lot of inventory.
Kieran Chandler: So, it sounds like fill rate is slightly better than the service level. Is there anything out there that is even better than that, that we should perhaps be measuring?
Joannes Vermorel: Yes, what is better is to start introducing the notion of economic drivers. You really want to keep track of the cost of inventory, the margin that you make, and what are the economic drivers that drive your supply chain. Fill rate gives you an estimate of the total demand you could potentially serve if you had infinite inventory, which has some business interest because it’s like the maximal market you could serve if your supply chain execution was perfect. The downside is that fill rate is quite difficult to measure, and you can’t really measure it without making some kind of forecast.
Kieran Chandler: It sounds like the fill rate is slightly better than the service level. Is there anything out there that’s even better than that, which we should perhaps be measuring?
Joannes Vermorel: Yes, introducing the notion of economic drivers is better. You want to keep track of the cost of inventory. The economic drivers that drive your supply chain include the cost of inventory, the margin you get whenever you serve successfully one unit, which can be tricky if it’s an internal customer in a web unless it’s a production unit. But it still exists. There’s also the stock-out penalty or the non-service penalty, which is the economical damage you do to your customer by not servicing the product. This ultimately becomes your cost because at some point, if a client loses money with you, they will find an alternative supplier, leading to a loss in loyalty. By focusing on these drivers, you can measure things in dollars or euros, which will give you something more tractable for the optimization of your supply chain.
The problem is that even if you have a very accurate measurement of the fill rate, it doesn’t necessarily translate into anything that you would do better. This loops back to your question of what is a good fill rate. Usually, the answer is we just don’t know. Fill rate is a percentage; you can increase it or decrease it. But until you’ve plugged in these economic drivers, you can’t decide for sure if it should be improved.
For instance, even if you have an 80% fill rate, in some situations that’s completely acceptable. Trying to go above that could lead to enormous risk in terms of inventory write-off. If you’re selling strawberries in a hypermarket, a 60% fill rate might be fine. You’d want to go for stock-out pretty much every single day because if you don’t sell your strawberries the same day, the next day your merchandise will have gone bad and the value of the merchandise declines rapidly.
Kieran Chandler: Going back to your IKEA example, how would these economic drivers work in that example? How would they work with the related products?
Joannes Vermorel: In terms of carrying cost, it’s about establishing whether your inventory loses value over time. It’s very specific. Do you have obsolescence costs? Is there some kind of fashion factor built in? Is it a long-lived product? For example, car brakes can last for a couple of years on the shelf without losing too much of their value. On the contrary, a T-shirt for the next World Cup is going to lose its value very quickly as we reach or even go after the World Cup.
That part of the cost is typically measurable, though not easy. You can have reasonable hypotheses with some degree of expertise in your domain. What is more difficult is to establish the cost for the clients. Here, the solution is frequently to ask, if you have the opportunity. In B2B businesses, you ask if it’s a problem if a product is not available. For B2C, it’s much trickier because you don’t get the opportunity to ask. Then you have to do some kind of loyalty analysis and figure out with correlation whether being out of stock for a given product really impacts your clients or not. But also, common sense does apply. One of the products that is most critical in the markets is diapers. If you’re out of stock for diapers, it’s a mission-critical product for young parents. So, usually, most companies have some kind of gut feeling about what products are really critical. The challenge is to organize all those elements so that they can be translated quantitatively.
Kieran Chandler: I’m going to play devil’s advocate a bit here. With your economic drivers, there’s still going to be a bit of a person’s opinion involved. There’s still going to be that sort of gut feel, that sort of understanding. Who’s to say that taking the economic driver approach, because it’s still someone’s opinion, is any better than just using service levels or fill rates? Why is it better?
Joannes Vermorel: Yes, it’s arbitrary, but only to some extent. Let’s take the other example. When you say, “Let’s shoot for 95 percent service level,” why is it even a good target? Maybe you used to shoot for 95 percent because it was traditional. But why is that? The question is a bit reversed. It’s very arbitrary if you say that you drive your supply chain by shooting for a given service level versus shooting for specific business drivers. Yes, it’s arbitrary on both sides.
So how can you differentiate between those two? I argue that economic drivers are better because at least you attempt to do a calculation that is aligned with the strategic vision of the purpose of your supply chain. Why is it better? It’s because at least you’re trying to approximate something that is an economic modeling of your supply chain.
Maybe your approximation is going to be super crude, but I would say that it’s better to be approximately correct rather than exactly wrong. And the point of service level is that saying that you shoot for an arbitrary service level is completely arbitrary. There is nothing to back it except traditions. At least when you go for economic drivers, in the end, you’re still trying to translate everything into dollars. You can challenge whether those estimations of dollars are accurate and revise them extensively. Still, they cannot be arbitrarily insane.
For example, what is the stock out cost for diapers in an open market? Let’s say a pack of diapers is something like 30 euro and on average, you sell 20 a day. So you would say that the loss, if you look at the margin which is at 10 percent, is 60 euro for the day. Let’s say we can say that maybe because of the loyalty loss, we have customers that are not going to go back to this open market. Maybe the loss is 10 times the margin, so that would be 600 euro.
This estimate makes sense. If you say the loss is a thousand times the margin, that doesn’t make any sense. Can the loss be less than the margin? It doesn’t make any sense because if the product had been on the shelf, we would have made that. So, it gives you a range where 60 euro, which is directly the margin, is kind of the minimum loss, and 10 times the margin is not the maximum loss but something like that is starting to be quite heavy and reflects a good estimate of what could the loss be. One of the good things with economic drivers is that it doesn’t really matter whether you get it absolutely correct. What matters is that the proportions are relative to each other, they are kind of balanced. So that is something that is easier to achieve, just balance between the products.
Kieran Chandler: Okay, let’s discuss the economic drivers then. It always amazes me how many systems and supply chain managers remain loyal to techniques that may be somewhat out of date. Why do you think they remain so loyal to metrics like fill rates? Is it because they’re easier to implement? Why are they still using them?
Joannes Vermorel: Yes, in many ERP systems, you’ll find built-in settings to manage your inventory with service levels. This means you can set your SKUs, or stock keeping units, at a 95 percent service level, and the system will aim to meet that level in terms of reorders. However, the first catch is that while the system allows you to set a target of 95 percent, it does not guarantee that you will hit this target.
This can be tricky because, in many systems we’ve audited, you often end up with nonsensical service levels, like 99.5 percent, when the company is actually achieving 97 percent. There is a significant disconnect between the set service level and the actual result. This prompts a reverse engineering process within the company to create settings that deliver a service level they want, even if it’s not what they initially set.
Kieran Chandler: Why is that?
Joannes Vermorel: Underneath the system, you have a demand forecast and an assessment of the risk or uncertainty associated with this forecast. This usually involves safety stock, but without going into too much detail, the loyalty to this system starts with having a service level setting, and then building your own culture to mitigate the fact that what you set is not what you get.
This requires significant effort and organization, and eventually, the entire organization becomes entangled in a legacy of processes. A good portion of these processes are geared around micromanaging these pseudo service levels that you’ve entered into the ERP system, in order to hopefully achieve better service levels.
Over time, this requires a lot of time, effort, and organization, and you end up having a lot of processes around it. It’s not so much about loyalty as it is about the fact that if you want to change this system, it becomes a large initiative within the company, and it challenges a lot of things. Changing the status quo is complicated.
Kieran Chandler: That’s a nice place to conclude. If a company is entangled in these processes, how easy is it to change? What would be the first steps in taking an economic driver approach?
Joannes Vermorel: The first step is to step back and get a big picture view of what you’re trying to optimize. Understand the primary mission of the business and how you can estimate whether you’re doing a good job or not. I would also suggest starting to think in terms of dollars or euros.
Our vision is that if you cannot measure what you’re doing in terms of economic value, you cannot do any kind of supply chain optimization at all. That’s a requirement. There is no alternative to that. So step back, try to have this economic perspective, and then see if your supply chain modernization efforts align with a service level KPI that is the right tool for you to execute. Most likely, it’s not, and I believe that these high-level insights are a good starting point to figure out a better alternative.
Kieran Chandler: We’re going to have to leave it there, but thanks for your time today, Joannes.
Joannes Vermorel: Thank you.
Kieran Chandler: That’s everything for this week. Thanks very much for tuning in, and we’ll see you again next time. Bye for now.
