00:00:07 Introduction and Pinak Dutta’s background in supply chain.
00:01:33 Evolution of network design and the role of technology.
00:04:07 Digital maps and their impact on network design.
00:05:32 Data types and aggregation levels used for strategic network design.
00:07:45 Differences between network design and inventory optimization.
00:09:27 Long-term vs. short-term decision making and the role of transactional data.
00:11:01 The growing spectrum of options in between long and short-term decisions.
00:12:20 Using third-party logistics (3PLs) for flexibility in warehousing and fulfillment.
00:14:02 The ease of working with short-term tactical decisions and hidden costs.
00:16:01 Third-party flexibility in supply chain.
00:16:58 Analyzing competitor locations to design networks.
00:18:14 Future of supply chain networks and complexity.
00:20:58 Inventory optimization and its evolution over time.
Kieran Chandler interviews Joannes Vermorel and Pinak Dutta on supply chain network design, inventory decisions, and the role of technology. They discuss the importance of optimizing factors such as warehouse locations and transport routes to improve inventory allocation and efficiency. The interview covers the historical context of network design, the use of digital maps, and the distinction between network design and inventory optimization. The conversation also addresses the role of third-party logistics providers, long-term planning, and data in designing supply chain networks. They emphasize the need for advanced optimization algorithms, inventory placement strategies, and risk assessments for a sound supply chain strategy.
In this interview, host Kieran Chandler discusses the importance of supply chain network design and its implications on inventory decisions with Joannes Vermorel, founder of Lokad, and Pinak Dutta, Head of Network Optimization at Spreetail. Pinak Dutta has 10 years of experience in various aspects of supply chain management, including engineering, transportation planning, production and supply planning, operations research, network design, and energy optimization. He believes that prescriptive analytics can deliver significant value for organizations and uncover opportunities not observed using traditional methods.
Joannes Vermorel provides an initial overview of the network design domain, explaining that it has been heavily influenced by technology and has gone through several stages. In the 1950s, it emerged as a field of operation research, with mathematical models being developed but not widely used in the industry due to the lack of computing power. Around the year 2000, the advent of the CD-ROM allowed for easier access to detailed maps, which enabled companies to create map-driven software for enterprises. This marked the beginning of supply chain design, which focused on simple tasks such as dropping a pin on an electronic map to represent a warehouse and calculating the regions that could be reached within a certain time limit by trucks.
The interview then delves into the importance of good network design in supply chains. Both Joannes Vermorel and Pinak Dutta agree that it is crucial to consider a variety of factors when designing a supply chain network, such as the locations of warehouses, transportation routes, and regional constraints. By optimizing these factors, organizations can improve their inventory allocation, reduce costs, and increase efficiency.
Pinak Dutta emphasizes the value of prescriptive analytics in uncovering opportunities that may not be apparent through traditional methods. He argues that advanced techniques can help organizations reap financial benefits by providing insights into optimal supply chain design and inventory allocation. Joannes Vermorel adds that the development of technology, such as electronic maps and data-driven software, has played a significant role in enabling companies to better design their supply chains.
The interview also discusses the historical context of network design in supply chains, with Joannes Vermorel highlighting the evolution from paper maps to electronic maps and the corresponding impact on supply chain design. He notes that in the 1990s, companies like Blockbuster relied on paper maps, making network design and optimization a time-consuming and tedious process. The advent of the internet and the development of software tools for supply chain optimization have since made it easier for organizations to design and manage their networks more efficiently.
The conversation begins with Vermorel describing the challenges of designing large-scale supply networks, particularly when dealing with overlapping locations. He highlights the importance of digital maps in overcoming these difficulties.
Dutta explains that the first step in optimizing networks is to determine the main purpose of the network design. This could involve developing strategic initiatives, considering capital investments, or addressing inefficient networks that have grown through mergers, acquisitions, or natural growth. He emphasizes that data for strategic-level decisions should be aggregated, as opposed to the granular level of data used in inventory optimization. Dutta suggests using techniques like integer linear programming models and building in stochastic elements for risk management.
Vermorel agrees with Dutta’s distinction between network design and inventory optimization. He notes that network design requires looking much further ahead due to the long-term nature of investments in facilities. Vermorel adds that most transactional data loses relevance beyond a one-year horizon, and very few industries can rely on such data for longer-term projections. However, he acknowledges that demographics can provide valuable insights for long-term planning.
Vermorel also discusses the growing spectrum of options available to companies, such as Fulfillment by Amazon, which offers tactical advantages. He points out that these options can have significant influence on long-term investment decisions, such as whether a company wants to offload capacity during peak periods or strategically position itself to have extra capacity for the rest of the market.
Both Vermorel and Dutta touch upon the role of third-party logistics (3PL) providers in offering flexibility for warehousing, fulfillment, and handling peak periods. They mention that 3PLs can help companies access additional resources, such as fulfillment centers, material handling, and labor. However, Dutta notes that relying on 3PLs can potentially impact service levels due to reduced control over operations.
The conversation covers the use of third-party logistics providers (3PLs), short-term flexibility, long-term planning, and data in designing supply chain networks.
The use of 3PLs offers flexibility and cost savings for organizations, especially when their own resources are highly utilized. However, the COVID-19 crisis has revealed hidden costs to relying on third parties. Companies like Amazon, which have consistently invested in their infrastructure, emerged stronger from the crisis, as they could prioritize their own needs during times of increased demand.
Short-term flexibility can be beneficial for companies, but long-term planning should not be overlooked. By examining competitor and industry data, companies can gain insight into optimal locations for their fulfillment centers. This can be crucial when starting out in a new market or industry.
The future of supply chain networks will require increased versatility, as complexity, regulations, and economic uncertainty continue to rise. Companies that are prepared for a wide range of possibilities, like Amazon, will be better positioned to succeed. Automation, such as the use of robots in warehouses, can help insulate companies from disruptions like pandemics.
In terms of inventory optimization, the field has evolved from simple safety stock calculations to more sophisticated models and algorithms. The work of Dr. Steven Graves and Dr. Deshawn Williams on guaranteed service time models has been particularly influential. As supply chains continue to adapt to changing attitudes and technologies, companies must be prepared to navigate the challenges and opportunities that lie ahead.
They emphasize the growing complexity of supply chains and the need for advanced optimization algorithms to address customer-first approaches and service level expectations. The conversation also highlights inventory placement strategies and risk assessments, including the impacts of disruptions like natural disasters on raw material supplies. The speakers emphasize the criticality of a sound supply chain strategy and the role of advanced optimization tools in achieving this.
Kieran Chandler: This week, we’re delighted to be joined by Pinak Dutta, who’s going to explain to us what constitutes good design and how organizations can exploit this to improve their inventory allocation. So Pinak, thanks very much for joining us today. Perhaps we can just start by telling us a little bit more about your background.
Pinak Dutta: I have roughly 10 years of experience in supply chain across different functions like engineering, transportation planning, production and supply planning, operations research, network design, and energy optimization. I have been in the chemical manufacturing industry, inventory optimization, and the e-commerce space. Currently, I am the Head of Network Optimization at Spreetail. I firmly believe that prescriptive analytics can deliver significant value for any organization and uncover opportunities not observed using trivial methods. My goal is to sow the seeds of these advanced techniques and help organizations reap financial benefits from it.
Kieran Chandler: Brilliant. And Joannes, today our topic is all about the importance of network design in particular on our supply chains. So, what’s your initial overview?
Joannes Vermorel: My initial overview is that it’s a domain that has been heavily driven by technology, and it went through several stages. It’s very intriguing. In the 50s, it emerged with tons of mathematical models for operations research, but at the time, it was highly theoretical. Computers were nowhere near a place where they could be actually used, so a lot of theoretical research was done super early on, but it was almost never really used in any industry for at least a couple of decades.
Around the year 2000, in the very late 90s, there was a first breakthrough, which was the CD-ROM. You may wonder what the connection is between the CD-ROM and network design. What was a big breakthrough is that suddenly, you could have a very detailed electronic map of a large region like the US or Europe on a compact disc. Internet at the time was too slow for things like Google Maps. So, companies started to ship electronic maps to a very large audience.
On the business side, it’s very interesting because that was the point in time when a short series of players started to deliver map-driven software oriented for enterprise. As part of that, the idea of supply chain design emerged. You could take an electronic map, drop a pin on it, and assume that it’s a warehouse. Then, you could define all the regions covered by trucks assuming a 30 minutes time limit in terms of reach. This might sound simple, but at the time, when the only thing you had was paper maps, doing this exercise was incredibly tedious and time-consuming.
Kieran Chandler: Let’s say, with Blockbusters, at their peak around the year 2000, they had 5,000 locations in the US. So, if you wanted to decide whether it was smart to add an extra location somewhere, it was very difficult because you had overlaps all over the place. Designing large scale supply networks, when you end up with thousands of locations, turns out to be a very messy exercise if all you have is paper maps.
Joannes Vermorel: Yes, and as Pinak mentioned, the use of digital maps has become essential for supply chain optimization.
Kieran Chandler: Today, what sort of data should we be looking at if we’re optimizing our networks?
Pinak Dutta: The first thing that we have to ask ourselves is the main purpose of the network design. Developing key strategic initiatives, whether due to cost or other factors, sometimes involves considering capital investments. Also, addressing inefficient networks that might have grown through mergers and acquisitions or natural growth can help eliminate clutter. The data you’d be looking for at the strategic level of decisions would be highly aggregated, which is different from inventory optimization where you look at data at a more granular level.
For network design, you’d look at time periods in annual or quarterly buckets, not at weekly or daily levels. Similarly, for products, you’ll examine data at a product category level or higher levels of aggregation, rather than individual SKUs. And instead of looking at individual storage locations, you’ll be aggregating the data at facility levels.
Regarding mathematical techniques, you’re likely going to use integer linear programming models. Sometimes, you’d want to build some stochasticity into the models to account for risk management or other potential factors. This is different from inventory optimization, where you’d be looking at data in weekly or daily buckets, at SKU levels, and at storage locations or facility levels. For these situations, you’d probably use more stochastic approaches, taking care of time variations, forecast errors, and other sources of uncertainty.
Kieran Chandler: Pinak mentioned a few approaches to network design. How do inventory optimization and network design work together and complement each other?
Joannes Vermorel: That’s interesting because, on this question, I completely agree with Pinak. When you’re thinking in terms of network design, you’re looking much further ahead since you’re considering investments like building new factories or warehouses, which are likely to be in operation for one or two decades. This requires looking far into the future. As a rule of thumb, most transactional data starts to lose its relevance if you look further than one year ahead. There are very few industries where a transaction has any meaning for the future beyond one year.
It doesn’t mean that you won’t be selling any more products, but the products will have changed. You will have changed your assortment, branding, packaging, and many small details. You might still be selling shampoo, but it’s not exactly packaged in the same way. You might have reorganized brands across these evolutions, which are slow.
Kieran Chandler: And gradual, but if you look far into the future, suddenly the very disaggregated data really stops making sense, so I completely agree in this regard.
Joannes Vermorel: I would even say that statistics overall tend to lose most of their relevance if you start looking anywhere like five years ahead. There are very few industries, maybe aerospace, where statistics have any relevance, at least based on transactions. In contrast, if you look at statistics based on demographics, it’s pretty safe. For example, the New York area was a densely populated region 50 years ago, and I’m not taking too much risk saying that unless something catastrophic happens, the New York area will still be a very wealthy, densely populated area 50 years from now.
But when we change the focus and look at, for example, the sales of blockbusters in the late 90s, could they tell us about what would happen for the next two decades? Obviously not, as now they have dropped to zero. You cannot predict these sorts of things just by looking at transactions.
Now, what is interesting is that between this long-term horizon and ultra-short-term horizon, typically inventory decisions are decisions that you take every single day, and building or placing a warehouse is a decision that you take every five years. But in between, I believe there is a growing spectrum of options that is kind of new. For example, you have players like Fulfillment by Amazon, which gives you options that are tactical and fast, but the simple fact that those options exist has a big influence on your long-term investments.
Do you want to position yourself as offloading capacity in peak periods before Christmas to Amazon, and by the way, beware, Amazon has some very special pricing that is very high during this period? Or, on the contrary, do you want to strategically position yourself as a company that precisely has extra capacity for the rest of the market to leverage during the peak period?
Kieran Chandler: Let’s talk about one of those options then, like the common one of those 3PLs, and the ability to actually rent space within warehouses and shipping. Does that change the approach that companies are taking and the philosophy they have?
Pinak Dutta: Absolutely. So, of course, 3PLs give you more flexibility, like Joannes mentioned, in terms of warehousing and fulfillment, and specifically for peak periods. For example, during peak periods, you can leverage various 3PLs that are available, whether it be their fulfillment center, material handling, or labor. Anything that you can take off, you can potentially use them. The flip side is there could be an impact on your service levels because you do not have full control over it, so you should be prepared for some of that. But of course, it does provide great flexibility, especially if your own organization’s fulfillment centers or other resources are at high utilization.
Instead of going out and renting a new building, it sometimes just makes sense to utilize some of the 3PLs, especially for certain programs such as cross-docking. Some organizations use 3PLs because it gives them more flexibility, and also the fact that they don’t have to store the inventory themselves. During cross-docking, you do not expect the inventory to be stored for more than a week anyway, so in those cases, 3PLs make good sense.
Kieran Chandler: Joannes mentioned a very key word, the idea of flexibility and the idea that maybe by taking more short-term decisions, you can be much more flexible in response to what might occur in the future. How easy is it to work with short-term tactical decisions compared to more long-term ones?
Joannes Vermorel: It’s easier to operate if you have a lot of short-term flexibility. However, there are hidden costs, as the COVID-19 crisis has shown. Companies like Amazon, which have been investing in infrastructure for decades, emerged from the crisis even stronger. When everything collapsed during the crisis, companies who controlled the infrastructure prioritized their own needs and interests. In countries like the US and France, during lockdown, e-commerce became even more relevant. Amazon made strategic choices in adjusting their assortment to gain efficiency but also privileged their own products. Having third-party flexibility is easy, tempting, and usually financially better, but in more extreme situations like the COVID-19 crisis, companies that patiently grow and nurture their infrastructures can have a huge payoff and emerge even stronger.
Kieran Chandler: Are there any types of data that might not be classically considered that we can use in designing our networks?
Pinak Dutta: Absolutely. One thing you can look at is where your competitors have their locations. If you’re starting out in the business, you can do all the analysis and use software like supply chain guru, but to make the final decision, it helps to look at where your competitors or other players in the market have their locations. This is why you often find retail or e-commerce companies having their fulfillment centers close by. It’s an easy way to decide on locations even without extensive analysis.
Kieran Chandler: It’s interesting because if Blockbuster had been looking at Netflix, which is located everywhere, they would have been in all sorts of trouble. Joannes, looking forward to the future, what could be of interest for designing the networks of tomorrow?
Joannes Vermorel: In the 20th-century supply chain, the idea was to get super big so that your average costs are super low, and you have economies of scale. That works very nicely when you have low complexity environments with a few products and clients, ideally in B2B markets.
Kieran Chandler: The future is incredibly versatile with more products and complexity. Unfortunately, I think it’s very complicated with more regulations and changing regulations. You end up facing a lot of unknowns, and it’s not going to get better anytime soon. Companies that tend to be prepared for everything, like Amazon, can deal with challenges like the pandemic. If you have warehouses served pretty much only by robots, you’re not too impacted by human viruses. There are many options, like additive manufacturing, which is costly but creates a lot of potential. What do you think about the future of supply chains and inventory optimization?
Joannes Vermorel: I believe that uncertainty and economic changes are on the rise. As a result, companies need to be prepared for everything. Automation can help mitigate risks, such as human viruses, and provide additional options for dealing with complexity. Additive manufacturing is one example of a technology that offers potential for the future.
Kieran Chandler: Pinak, how do you see the future of supply chains being shaped by changing attitudes towards inventory optimization?
Pinak Dutta: For optimization, we have come a long way. It started years back with simple safety stock calculations using service levels. Later, the concept of inventory optimization evolved with fantastic work done by Kimberly Clark Simpson. Dr. Steven Graves and Deshawn Williams brought in concepts of guaranteed service time models, which have been implemented in some software in practice. We have come a long way from where we were years ago.
As supply chains become more complex, we see the importance of moving towards more advanced inventory optimization. However, these are computationally complex, so there have been some simpler solutions like using normal distribution, which is not the case in reality. A lot of work has been done in this field. The current environment, with increased competitiveness and a customer-first approach, emphasizes the importance of better service levels.
Companies like Amazon with their Prime Now service put further pressure on inventory optimization. Inventory placement strategies, service levels, and tying everything together using advanced optimization algorithms have become critical nowadays. Risk assessment is also important for situations like the pandemic or natural disasters. All of these factors contribute to having a solid supply chain strategy based on inventory.
Kieran Chandler: Thank you both for your time. That’s everything for this week. We’ll see you again in the next episode. Bye for now.