00:00:07 Lead times’ significance in inventory management.
00:00:50 Unpacking technical aspects of supply processes.
00:02:56 Measuring lead times for supply chain optimization.
00:04:12 Dissecting lead time steps for optimization.
00:06:59 Breakdown of lead times for accurate measures.
00:08:02 Uncertainty in lead time and distribution algebra.
00:10:03 Real-life example: improving lead times.
00:12:37 Explaining unexpected business process delays.
00:13:02 Warehouse inefficiencies causing delays: an example.
00:15:00 Resolving issues in warehouse example.
00:16:16 Box pyramid formation by the conveyor.
00:17:26 Final thoughts.
In this interview, Kieran Chandler and Joannes Vermorel discuss the importance of lead times in supply chain optimization. Vermorel highlights the need for accurate lead time measurement at various stages, from manufacturing to customer-ready items. By decomposing lead times into distinct components, companies can better estimate overall lead times and identify bottlenecks. Vermorel also emphasizes the benefits of shorter lead times, such as improved planning and reduced forecasting windows. They examine a real-world example where inefficient warehouse processes, like LIFO order processing and a short conveyor belt, led to delays, emphasizing the impact of small implementation details on supply chain efficiency.
In this interview, Kieran Chandler and Joannes Vermorel, the founder of Lokad, discuss the importance of lead times in supply chain optimization and inventory management. Lead times are often overlooked by companies, who instead focus on inventory forecasting. Vermorel explains that understanding lead times is crucial for business success and that ignoring them could lead to inaccurate forecasting.
Lead times represent a more technical aspect of the supply chain process, requiring attention to detail and knowledge of production and transportation processes. They may not be as glamorous as focusing on future demand, but their impact on inventory management is significant. Longer lead times generally require larger stock levels.
The first step to improving lead time management is to measure lead times accurately. Companies often have poor measurements of lead times since it’s not considered mission-critical for supply chain operations. However, for optimization purposes, it’s essential to have good measurements for every step of the supply chain process.
Vermorel suggests that companies should measure lead times at various stages, such as manufacturing, port-to-port transportation, port-to-warehouse transportation, and the time it takes to have items ready on the shelves for customers. These measurements don’t need to be extremely granular, but having a clear differentiation between different stages of the supply chain can provide a solid basis for lead time forecasting and optimization.
By measuring lead times accurately, companies can identify potential bottlenecks or delays that may be accidental or unnecessary. For example, in a 13-week lead time to import goods from Asia, there might be a two-week delay due to the decision-making process for reordering items. By identifying and addressing such issues, companies can improve their supply chain efficiency and better manage their inventory levels.
Vermorel explains that variability in lead times is due to several factors. For instance, transportation can be affected by weather conditions, causing a cargo from Asia to take a couple more or fewer days than expected. Additionally, customs at ports can add an extra week of delay randomly. Furthermore, production lead times are subject to change, as suppliers may be busy serving other clients or not solely dedicated to one customer. These factors contribute to random fluctuations in lead times.
The suggestion to manage these fluctuations is to decompose lead times into distinct parts, such as production, transportation, customs clearance, and decision-making, among others. While this may add complexity, Vermorel argues that it is the only way to reasonably measure lead times. By breaking down lead times into their fundamental components, businesses can better understand each segment and gain a more accurate estimation of the overall lead time.
However, each of these individual factors comes with its own range of uncertainties and probabilities. To have confidence in predicting lead times, Vermorel suggests using an algebra of distributions to combine these probabilistic variables. Although adding variables may increase uncertainty, the overall lead time uncertainty does not skyrocket, as each distinct process is estimated more accurately.
In the context of a real-world example, Vermorel describes how, when a company starts accurately measuring its lead times, it often discovers accidental delays that can be eliminated through automation. This can lead to significant improvements in the company’s supply chain processes and overall efficiency.
The interview highlights the importance of understanding the variability in lead times and the benefits of decomposing them into distinct components. By doing so, businesses can more accurately estimate and manage their lead times, leading to improved supply chain optimization and overall efficiency.
Vermorel emphasizes that reducing lead times can have a significant impact on supply chain efficiency. By continuously moving goods rather than keeping them stationary for extended periods, companies can shrink lead times and make better use of their resources. Shorter lead times not only help reduce the amount of stock that needs to be kept on hand, but they also make the entire supply chain more efficient in a variety of ways.
One key benefit of shorter lead times is that they allow for better planning. When companies can compress their lead times, they don’t need to forecast as far into the future. For example, if a company can reduce its lead time from twelve weeks to eight weeks, it only needs to forecast demand for the next eight weeks, rather than the next twelve. This reduction in forecasting window leads to more accurate forecasts, as predicting demand in the short term is inherently more accurate than attempting to predict it further into the future.
In addition to improving planning, shorter lead times can help companies avoid unexpected delays. While businesses should ideally have a comprehensive understanding of every step in their supply chain process, Vermorel points out that unexpected delays can still occur, often due to details being overlooked. He shares an example from a warehouse he visited a few years ago, where electronic orders for goods were printed on paper for employees to process manually. The way the printer was set up caused the orders to fall into a box, creating an unnecessary delay in processing the orders.
Vermorel describes a company’s warehouse process where a stack of orders was built up. The issue was that the orders were being processed in a last-in, first-out (LIFO) fashion instead of first-in, first-out (FIFO). The employee would always take the most recent order from the top of the stack, meaning the order at the bottom of the stack would remain there until all other orders were cleared. During busy periods, this could result in some orders staying at the bottom for an indefinite amount of time, possibly days.
The LIFO issue was caused by the orders being printed and placed in a stack. Interestingly, a similar problem occurred further down the supply chain due to a conveyor belt being too short. Employees would pick items and place the boxes on the conveyor belt. However, the conveyor belt was manually operated and only about 10 meters long, which was insufficient for the volume of boxes. Consequently, the conveyor belt would become full, and employees would start placing the boxes on the ground in front of it.
As the conveyor belt remained full, a pyramid of boxes would build up in front of it. When the conveyor belt eventually started moving again, employees would take the boxes from the top of the pyramid, resulting in a LIFO situation once again. The oldest box would only be brought back to the conveyor belt after the entire pyramid was unstacked.
Kieran Chandler: Today on Lokad TV, we’re going to understand exactly why lead times are so important to business success and also discuss why if you don’t take them seriously, you might as well use a random number generator when it comes to forecasting. So, Joannes, I think this is probably one of the more simplistic topics we’ve talked about on Lokad TV. Why is it something we’re discussing today?
Joannes Vermorel: I would say lead times may not be the elephant in the room, but something slightly smaller than an elephant, let’s say a horse in the room that gets ignored. The elephant in the room that is ignored is just the probabilistic nature of the demand forecast itself. There is uncertainty on demand, and you have to deal with it. But the horse in the room, just next to the elephant, is the fact that lead time also has some uncertainty and needs to be properly forecast.
Lead times are more technical; it’s not very complicated, but you need to be able to be aware of the fine print of your supply process. You need to know exactly what is happening on the production side, what is happening in terms of transportation, and if you have several transportation options available, such as transporting your goods by sea or by air, you need to be aware of that. It’s a lot of small details, and you need to pay attention to those small details to get a correct understanding of the situation. It may not be as glamorous as focusing on future demand, but the reality is that because lead times have a linear effect on inventories, if you have lead times that are twice as long, you need stocks that are twice as large. It is frequently underestimated because it’s more technical, I guess.
Kieran Chandler: You mentioned the complexity of the approach. What sort of first step should someone take if they wanted to improve the way they were approaching lead times?
Joannes Vermorel: The first step is measurement. Very frequently, there is relatively poor measurement of lead time. It’s not mission-critical to measure lead time to operate the supply chain, but if you want to optimize, you need to have good measurements. Ideally, you want to measure things for every single step. In theory, you could go into insane granularity, measuring every minute over several weeks of time, where the exact position of a product is to have complete visibility on exactly where those products get stuck. The reality is that most of the time, products are not moving.
Without getting into that extreme perspective, if you can already clearly differentiate the manufacturing time, the time it takes to transport port to port, the time it takes to transport from port to warehouse, and the time it takes once things have been received in your warehouse to have them on the shelves ready to be picked and served to clients, that would be great. To have all those measurements done properly, we are not talking about millions of measurements; it’s like half a dozen that give you a great entry point to do your lead time forecast and your lead time optimization. The first thing you will realize is that maybe some delays are accidental, and out of your 13 weeks of lead times to import from Asia, you have two weeks that are just the delay it takes you to reach a decision on how much to reorder.
Kieran Chandler: Let’s talk about that variability. What are the key factors that are contributing to this sort of variability in lead times?
Joannes Vermorel: Transportation itself has a bit of variability. For example, depending on weather conditions, a cargo coming from Asia might take a couple of days more or less. It’s not super precise. Then, if you have to go through customs in a port, the customs themselves can add an extra week of delay randomly. Before transporting the goods, you have to produce them, and your supplier might have received other orders from other clients, which you have no control over. So, the workload of your supplier at a precise point in time is something that is a bit outside your control. Even if, in theory, producing 1,000 units with a production unit takes exactly one week, if this production unit is already busy serving another client, it might take longer. So, the production lead time varies because your suppliers are not solely dedicated to you. There are many causes that can create random fluctuations in lead times.
Kieran Chandler: So, what you’re saying is, instead of taking the whole lead time as one thing, we should be splitting it into distinct parts. Doesn’t this add a lot of extra complexity?
Joannes Vermorel: It does add complexity, but it’s also the only way to have a reasonable measurement of your lead time. If you want to understand your lead times, you need to decompose them into their fundamental pieces. Yes, you end up having maybe half a dozen time segments, like the delays to produce, the delays to transport through cargo to port, the delay to transport by truck from the port to your warehouse, the delay to ship from your warehouse to the final clients, the delay for you to take a decision on how much you need to reorder, and maybe the delay between reorders. So, yes, you have to decompose, but the reality is that if you want to have a reasonable measurement of your lead times, there is no real alternative. It’s not because you decide to ignore the problems that the problem goes away.
Kieran Chandler: It sounds like with each of these individual factors, there’s going to be a lot of uncertainty, with a real range of probabilities that could possibly happen. How can you have any confidence in what you’re predicting a lead time to be?
Joannes Vermorel: That’s where you typically need an algebra of distributions, where you can combine your variables and probabilistic variables. Lokad has one, but there are also other options that exist. Bottom line is that, yes, you will have probabilistic measurements for every step of the process that represents a variable amount of days, and you will need to combine them, which is just making an addition like this delay plus this delay, except that you have probabilistic variables that you’re adding. You have to understand that the statistical noise does not really compound. So yes, you have more uncertainty as you’re adding variables in terms of total spread of possibilities, but it doesn’t mean that the lead time uncertainty skyrockets just because you’re adding steps. What you’re doing in the end is decomposing a large delay into smaller pieces that you can estimate more accurately because they represent distinct processes.
Kieran Chandler: To wrap things up, let’s have a real-world example. Can you give an example of a time when a customer of yours improved the way they’re approaching their lead times, and how did that affect their business processes?
Joannes Vermorel: Yes, as you start measuring your lead times, the most classical finding is that you just realize that you have delays that are completely accidental.
Kieran Chandler: You just realize that goods arrive in a location, stay there for five days with nothing happening, just to be moved again. And those five days can be shrunk to nothing if you just decide, okay, instead of putting them on shelves, wait for five days without doing anything and then keep them moving. It’s just keeping them moving all the time. So, and then again, better forecasts will help you to reduce the amount of stock that you need to keep over the place. But shorter lead times are doing that in a way that is much more efficient.
Joannes Vermorel: If you could bring all your lead times to zero, you would not need any inventory whatsoever. If you could just say, whenever you have one unit that is being requested by a client, instantaneously you could produce this unit and have this unit transported to the client. That would be like just-in-time everything, and you would not need any stock, any planning, anything. So, compressing the lead times has positive effects through the entire supply chain. Also, it’s not just that it reduces stock, but if you have shorter lead times, you do not need to plan as far ahead as you used to. If you can compress your 12 weeks lead times into 8 weeks lead times, it means that you only need to forecast the demand 8 weeks ahead instead of 12 weeks ahead. And the reality about forecasts is that the further in the future, the more inaccurate the forecasts are. So, if you can forecast only the short term and not the long term, your forecasts by design are more accurate, which simplifies everything and makes your overall supply chain more efficient just because you do not have to look that far into the future.
Kieran Chandler: And you mentioned unexpected delays. I mean, that’s probably quite surprising to a lot of our viewers. I mean, surely a business should have a good understanding of every step of that process. So how can these unexpected delays actually occur?
Joannes Vermorel: Sometimes, the devil is in the details. Just to give you an example of a warehouse that I visited a few years back: they were receiving electronic orders for goods to be delivered, to be shipped. The system they had inside the warehouse was actually printing the electronic orders that were received. So, the idea was that they would print a request, and an employee would come, take the request that was printed on paper, and go do the picking and ship the goods. It was expensive goods, so it was fairly reasonable to have a relatively manual process for picking inside the warehouse. But the thing was that the way the printer was set up, the shipment orders that were received would be printed, and then the sheet of paper would fall in a box, and then…
Kieran Chandler: Joannes, can you explain a specific problem that occurred in the supply chain process?
Joannes Vermorel: Yes, of course. So, the issue was related to the order stacking system. When an order was received and printed, it would be placed on top of a stack. However, the problem was that the employees would always pick up the last order that was received electronically. This meant that if there was a stack of orders, the one at the bottom would stay there until the entire stack was cleared. So, during busy periods, orders would pile up, and the last item in the stack would be delayed indefinitely.
Kieran Chandler: Ah, I see. So, instead of following a first-come, first-served approach (FIFO), they unintentionally implemented a last-in, first-out system (LIFO). Is that correct?
Joannes Vermorel: Exactly. They had inadvertently implemented LIFO because the employees were picking up the last order received, which was printed on top of the stack. This happened because the order processing was happening at the printer side. But the same problem occurred further down the stream due to a short conveyor belt.
Kieran Chandler: Oh, I see. So, what happened with the conveyor belt?
Joannes Vermorel: Well, the conveyor belt was too short for the volume of orders. As a result, when it became full, employees couldn’t place more boxes on it. So, they ended up putting the boxes on the ground in front of the conveyor belt. But when they returned, they found that the belt hadn’t moved, so they put more boxes on the ground. This led to a pyramid-shaped stack of boxes building up. When the conveyor belt started moving again, they would pick up the boxes from the top of the pyramid, starting the process all over again.
Kieran Chandler: That sounds like a challenging situation. It seems like they unintentionally had a LIFO system again due to the physical setup and the short conveyor belt.
Joannes Vermorel: Yes, exactly. It’s interesting how small implementation details can create unexpected situations in the supply chain. Sometimes, even slight imperfections in the physical setup can have significant consequences. So, my suggestion would be to invest in a longer conveyor belt and perhaps make adjustments like putting a hole in the bottom of the boxes to avoid stacking issues.
Kieran Chandler: That’s a great suggestion, Joannes. It’s fascinating how important these seemingly small details can be. Well, that’s all for this week. Thank you for joining us, and we’ll be back next week with another episode. Until then, thanks for watching!