From demand forecasting to actual demand planning: In times of Corona this is now for real

Over the decade, the profession of demand forecasting has been renamed many times to reflect the broader designation of the role. Demand management has been a term that has often been deployed. The term suggests that this is not just about forecasting, but also includes the shaping of demand, for instance by pricing and management of promotions. Als the term Demand planning has been frequently deployed as well. This term suggests that the role also involves the allocation of supplies to the demand, or maybe even actively planning or shaping demand to meet supply.

In reality, however, many companies that have deployed alternative names for the forecasting role have not done more than actual demand forecasting, i.e. the estimation of future demand – often based on some algorithm that uses past sales information augmented with some human judgment. The current Coronacrisis however presents both a need and an opportunity to really develop this function.

With drastic demand changes, algorithms need to be shut off

Demand forecasting algorithms that are deployed are based on time series of previous sales. Most of these are using relatively simple statistical methods that extend and smoothen previous time series. In the last few years, there has been much promise of machine-learning (sometimes dubbed as “Artificial Intelligence”) that takes additional external information, such as the weather or the pricing of competitors, into account when determining the forecast. With any of these algorithms, the basic presumption is that the underlying system does not change: relationships between independent variables (previous sales, assortment, weather, etc) and dependent variables (future demand) remain unchanged. Obviously, this is not the case in the current circumstances. Hence, it makes perfect sense to switch off your automated forecasting support tools, especially if they are “hands-of-the-wheel” linked to orders being placed by your replenishment system without any human interference.

When will demand go back to normal?

This is of course the million-dollar question that everyone is facing. In all cases, it is important to first estimate when the situation may go back to more normal patterns of demand. This time estimate is difficult, but there are some basic elements that you can work with:

  1. Remaining time of the governmental measures that cause you drastic demand change. From the China situation, we know that this may likely be at least 3 months; maybe longer if the hospital capacity is really stressed.
  2. Lead time between you and the consumer market: the is the cumulative lead time that a molecule, part, or product leaves your plant and is being consumed by a final consumer. For a retailer, this is a just a few days, while for a chemical producer this can easily be half a year.

Demand changes at the consumer level will roughly take the sum of these two times to reach you. However, they are affected by inventory adjustments:

3. Cumulative surplus or shortage of inventory in the supply chain: if your customers or consumers have a shortage of inventory now compared to how much they would stock normally, this will need to be brought back to “normal”. This could imply additional orders in case of shortages, or reduced orders in case of surpluses. For instance, consumers will likely not buy toilet paper for a while in many countries, while the shortage of electronic parts in the automotive supply chain will need to be replenished. Such inventory replenishment requests will reach you more or less instantaneously and will not face the delay above. We have learnt this from the recovery after the credit crisis.

Finally, all of this will be exacerbated by the infamous bullwhip effect. Again, from the credit crisis we may reasonably estimate that the order inflation may easily be in the tens of percentage points, with this inflation being larger if you are further upstream from the consumer.

Real demand planning requires strategic choices and subsequent analytics

Current demand planning requires both strategic choices and subsequent operational analytics. The strategic choices depend on whether your market is currently facing (huge) drops in demand or (huge) increases in demand, or whether there are significant demand shifts between products or channels.

Currently facing drops in demand

Of course you are currently looking to serve alternative markets or trying to make alternative products for which there may be demand. But in many cases this may not be feasible. Then what remain is trying to estimate when demand will go back up, which I have discussed above. It makes sense then to decide whether you want to build up inventory in advance. If you have the financial means to do so and your products are not perishable, this may be a very sensible strategy. Demand will go up, and more than you expect, as discussed above, but the exact timing is hard to tell. What is critical is to involve your sales force in this plan. They need to be aware of the constraints of supply.

Currently facing demand increases

You are currently scrambling to make supply meet demand. Several companies have reduced assortment size in order to keep capacity up (by saving on changeover times). At some point, demand will go back down. In order to avoid a bullwhip, it is really important to keep a very clear picture of the actual consumer demand and the accumulated inventories between you and the consumer. Thinking cumulatively rather than incrementally makes a lot of sense. From a planning perspective, you will need to ask from your sales force to do something they don’t like to do: sensing in the market how demand will go down eventually. Your sales force needs to understand that if they are too late with their sensing, they will be causing potentially large amounts of unsold inventory somewhere in the supply chain.

Currently facing demand shifts

In particular channel shift have been happening: from out-of-home to supermarkets and from in-store to online. The first question of course is how much of that remains after the crisis eases. I am reading many reports that this is the definite breakthrough of online grocery. I seriously doubt this. I am definitely not a marketing expert but the current online experience is poor with extensive delays in delivery and many out of stocks. Also, after having been locked down for months, I can imagine that going out to a store will be a great experience for many. Hence, the argument could just as well be made that online sales will drastically decrease after the crisis eases. Hence, I think it makes sense for any supplier to hedge their bets: a bit of additional inventory makes sense, and building the ability to shift demand between channels or products is a worthwhile investment.

I seriously doubt the many reports that this is the final breakthrough of online grocery shopping

In conclusion

All of the above implies that companies will need to set up true interdisciplinary demand planning teams that actually have the ability to plan. Such teams should be able to make (or prepare) strategic choices and be able to conduct analytics of the consequences of such choices. This requires different information than just prior demand; it requires knowledge of the full state of the supply chain. And much of this requires humans to do the job.

Disclaimer: This is not the direct result of any specific academic study. The above is my current analysis and interpretation based on prior research conducted by me and others in the area of supply chain management, inventory management, and demand forecasting during crises. It is not an advise for anyone specifically. 

This article was published on LinkedIn on April 3, 2020

Making your (retail) distribution center Corona-resilient

Now that the Coronavirus has reached most of Europe and North America in sizeable numbers, preparations at many companies are in full swing to do whatever they can to try and be resilient to infections in their distributions centers. Here, I provide an overview what can be done, even at this late stage, to try and contains the effects on supplying your customers.

1. Take care of your associates

Your employees’ health is of critical importance to them and their loved ones. It is also critical to your operations, since infected associated may force you to close your facility. Much has been written about hygiene matters and access control checks, and I will not repeat them here. However, specifically for distribution centers, there is more that you can do. In many cases, distribution centers are staffed with temporary contract workers. In Western Europe, they are often workers from other parts of Europe that often change jobs and live in shared housing. Work with your temp agency to increase hygiene across these sharing housing facilities. Your responsibility in this case extends beyond your fully employed associated, also in your own interest. If you do home delivery to consumers, your associates will be in touch with many people, further increasing their and your risk. You will need to provide hand hygiene materials in your delivery vans.

2. Make sure you have sufficient supply

As trivial as it sounds, this is less than obvious. Many retailers have followed policies to try and limit inventories to reduce inventory cost. Instead, they require suppliers to hold inventory and only replenish when there is an immediate need. Given the likelihood of some disruption occurring due to facilities being temporarily closed, you want to pull in inventory to your locations rather than leaving them at your suppliers. Having the inventory in hand gives you more control in case of shortages and disruptions. Obviously, you should avoid creating a bullwhip further upstream but providing transparency to your suppliers that this is not demand-driven, but driven by an analysis of the supply chain risk. In all cases, it is important to be transparent and communicate intensively with your suppliers about such actions.

3. Distribute your inventory across multiple locations

Given the likelihood of disruptions, it is now more critical then ever to keep inventories at at multiple locations. Most companies tend to have more locations downstream rather than upstream, so moving inventory downstream makes all the sense in the world. With a virus like Corona around, traditional inventory risk-pooling suddenly becomes more risky rather than less risky. If you pool your inventory in a single location, your are unable to deliver if that location needs to be isolated. Spreading inventory across multiple locations increases the likelihood of regular out-of-stocks in each of the locations, so you need to equip your distribution network with options for lateral transshipments between locations. In the same line of thought, you need to be able to serve markets from alternate distribution centers. This may require making arrangements with logistics service providers. In case of regulated products, such as pharmaceuticals, this may require permission from the relevant authorities. It could be a good idea to obtain such permissions in advance.

4. Improve robustness in your warehousing operation

Warehouse staffing is typically fully flexible. Moving order pickers across the warehouse provides such flexibility and lowest cost. However, is also provides for maximum contact and infection risk. An option could be to virtually compartmentalize your warehouse, and assign pickers to a certain compartment only. A friend of mine relayed this to me as an option, which I think is highly interesting and where others could benefit. Slightly more cost, but much less risk.

This article was published on LinkedIn on March 6, 2020



Extreme scenario analysis for the Coronacrisis

With the Coronacrisis now seemingly getting firm ground in Europe, we are in dire need of extreme scenarios for the supply chain. We have been seeing very unconventional supply chain effects unfolding over the past weeks:

  • Apple faces great difficulty with ramping up the Foxconn manufacturing plant in Zhengzhou due to an insufficient number of dormitories to quarantine incoming migrant workers
  • Liner shipping companies have been canceling huge numbers of sailings from China, both due to demand drops but also due to access limitations on crew
  • Road transport in China is scarce due to quarantine and travel restrictions across the country

The art of scenario thinking has been mastered by Royal Dutch Shell over the past 60 years. Unlike the thinking of many, scenarios are typically not possible realizations of the future for which to prepare. The purpose of scenarios is to help managers stretch their thinking. Stretching your thinking helps you prepare for less extreme developments than pictured in the scenarios.

Can we think of some components of scenarios that could stretch our thinking of this Coronacrisis? It would be good to share some of your thoughts on this, as scenario thinking benefits from co-creation, i.e. letting multiple ideas flower at the same time.

Some initial ideas from my side to spark the mind:

1. European border closures

Over the past two years we have seen many scenarios pass for the hard Brexit. We would see large queues at borders between the UK and the continent, products could be stopped from crossing the borders and people movement could come to a halt. The only advantage we had at that time is that companies had time to prepare. They did so by massively building up inventories. Suppose we would see Brexit-like border restrictions at every internal European border? It may hence many sense to build up some stock ahead of time.

2. Much slower rampup in China

While on the epidemic side the Chinese seem to gradually get the situation under control, at the same time it may very well be that the rampup is much, much slower than anticipated. Companies that I talk to have all been estimating that by April supply volumes should be up to normal. Now that Korea, as an important supplier of critical high value parts to many Chinese plants, is facing rapid growth in the number of infections, this could slow down the rampup drastically. Suppose it will not take six weeks, but rather six months until supply starts to ramp up? This implies inventories will really fall dry and sales will plummet due to lack of products. Do you have sufficient cash to overcome this? The Netherlands government has already enabled affected companies to allow employees to receive partial unemployment benefits. Not sure if other countries have followed suit.

3. Absence of containers for exporting your products to Asia

Even if you are not sourcing from China at all, and even your demand in China is not affected, a lack of containers my stop you from exporting. Container shortages have already been reported, in particular reefers. Suppose a general shortage of export containers would appear? Do you have sufficient storage space to store your products? Is there sufficient cash to finance this?

These are just a few scenarios. I am sure anyone could think of more. It would be good for any company to sit together with leadership and supply chain expert to think of such extreme scenarios as an aid to improve preparedness. If we cannot hope for the best, at least we have to prepare for the worst.

This article was published on LinkedIn in February 2020

Coronacrisis affects global supply chains in multiple dimensions

Since the unfolding of the Coronacrisis, gradually more and more attention is being given to the consequences for the global economy. It has been interesting to witness initial reports affected the Chinese travel industry (24 January), international air travel (29 January), jet fuel in China (January 30), Chinese retail (5 February). Only later the effects of Chinese manufacturing closings reached the mass media, affecting the supply of retailers abroad with products and global supply chains with parts and raw materials. The first announcements affected car manufacturers in Asia (having little inventory due to just-in-time single sourcing policies and being in close proximity to the parts manufacturers in China), with the world’s largest car plant of Hyundai as a prime example. Then followed the high value retail products abroad that use air freight to supply their retail stores worldwide, with Apple becoming testimony that single sourcing may look good on the balance sheet but leads to dramatic effects in case a crisis hits.

The news unfolds as more parts of the supply chain were hit. One of the first to recognize that this virus outbreak could lead to massive disruption of global supply chains was my good friend Professor Yossi Sheffi of MIT. In late January, he sent out a first survey to supply chain leaders on which he reported on January 31 on LinkedIn. Interestingly, at that time most companies were waiting and watching, rather than investigating and acting. In an interview on CNBC, Yossi made an interesting observation that this is a crisis where both supply of the world’s most critical manufacturing base is hit at the same time as the world’s second-largest consumer market.

It is remarkable that still much of the responses are slow. One explanation could be that companies are silent about anticipated supply, as they are currently looking to find alternative sources of supply. In that case, it might indeed be better to be ahead of the competition, and keep quiet about shortages. However, I fear that many companies are not aware and have not yet made this a C-level priority. Shortages may not be visible yet, as many companies had extra inventories prior to the Chinese New Year, and shipments take 4-6 weeks to reach Europe or North America. Hence, for products that are normally transported via ocean, shortages may only start to become apparent next month. If there is a lot of safety stock, maybe only in April. And then the virus might be contained, so why worry?

What will happen in the next weeks and months?

Over the next weeks we will see more and more shortages appearing. Dutch retail chain Action announced this week that it is seeking alternative sources for its products. Action sells extremely cheap products that are mostly sourced in China. They were one of the first to recognize their problems at the retail side. Other retailers with extensive sourcing in China announce in the media that they do not expect any problems. It is hard to imagine that a company extensively sourcing in China will not be facing some empty shelves next month.

Moreover, even if you are not sourcing from China, it is likely that some of your suppliers or your suppliers’ suppliers are sourcing from China. It may only need to be a small single-sourced part somewhere that can take down an entire supply chain. As CNN said: “You can’t build a car with 99% of the parts”.

It is also reported to me (have not found any confirmation of this news yet from additional independent sources) that apparently in China road transport capacity is very difficult to get, since there are many transport restrictions in place.

The good news is that some car manufacturers are partially restarting their operations in China as of next week, as Toyota announced yesterday. However, others argue that the main problems are with second and third tier suppliers, for instance in the electronics industry. There is little visibility on these, often much smaller, companies.

Actually, even if the virus were contained some time in April, I predict that the effects on global supply chains will be visible for a long time, possibly far into 2021. The reason is that also the coronacrisis will lead to a bullwhip effect, just like other recent shocks such as Brexit, the US/China tariff war, and (with a much larger effect) the collapse of Lehman Brothers. The corona bullwhip is a complicated one to estimate. I would estimate the following effects based on current information and based on common knowledge of the bullwhip effect:

  1. Increases in prices for fast capacity-constrained transport modes, such as airline prices. Since many will try to obtain whatever stock is available and many flights have been canceled (belly capacity reported represents more than 50% of airfreight capacity out of China), this will lead to an increase in freight prices.
  2. Inflation of orders at Chinese suppliers and their alternatives in other countries. If there is a shortage of products, companies will place more and inflated orders at alternative suppliers. If alternatives are available (which in many cases will not be at short notice), they are likely to become capacity constrained soon. This will give them an opportunity to inflate their prices, but also they may induce their (potential customers) to inflate their orders in order to secure whatever is available.
  3. Propagation of this effect upstream the supply chain. From bullwhip theory we know that such inflated orders in case of shortage are increasing further upstream. Hence companies that are typically upstream, such as the chemical industry, could be affected only months later but in much larger quantities.

How can this be moderated?

Bullwhip theory that the effect can be moderated by information sharing. I understand from many of my corporate relationships that it has been very difficult to get updates on the actual status of their supplies from China. Many workers in China are still off, or do home office, and it is very difficult to obtain information. If there is anything the Chinese government can do, is to encourage the companies in China to provide transparency.

If this were not possible, companies in Europe and North America should still try and share as much as possible. If this were not possible for competitive reasons, it may make sense to report this anonymously. To any of my LinkedIn contacts, I would really appreciate a LinkedIn direct message if you want to share your current status, which will help me get a better picture. Also, please make use of the invitation of Professor Sheffi to take part in his online survey on the Coronacrisis.

In conclusion

From an academic perspective, we never want to waste a good crisis for future learning. We can only hope that companies would do the same. Really embedding on dual-sourcing strategies is the only way to move forward in a world that is increasingly connected and can be disturbed more and more easily by such crises.

This Article was published on LinkedIn in February 2020

Dedicated unloading bays are good for deliveries – and for cars too!

The fight for curb space in cities is getting firmer. Recently, Uber commissioned a study to define curb productivity. There is surprisingly little research on the use of curb space for urban delivery. Cities deploy unloading bays (in many case erroneously called loading bays) in apparently random manner, with vehicles often parked outside of such bays due to lack of space when unloading space is needed. Our newest research suggests that savings of over 40% in route duration can be reached, with less claims of public space. We showed this in a live controlled experiment in the historical downtown of Queretaro in Mexico – an environment that could be considered one of the most difficult to experiment in. And the learnings for last-mile distribution all across the world are impressive!

Together with my colleague Gaston Cedillo of the Mexican Institute of Transportation, I had a dream of experimenting in a live urban environment. While there are many initiatives for pilots and other models in smart city environments, I wanted to run a controlled experiment, i.e., to measure the exact effects of introducing a certain change in the public space on the logistics efficiency and the use of public space. Building on my extensive work over the past years studying distribution to nanostores, we focused on distribution to traditional retail. With 60-90 stops on a route, such distribution activities have lots of similarities to parcel delivery and other fragmented deliveries in the developed world.

The downtown area of Queretaro is beautiful. It is a UNESCO World Heritage area. It is also a parking nightmare. With 900 nanostores in just one square kilometer the area is frequented by hundreds of vehicles per day for very small deliveries. Finding unloading space if difficult. Bays are present, but enforcing that these space is only used for unloading freight is next to impossible. With just a small number of traffic police, there is widespread illegal parking. The ideal area to run an experiment for which is almost too hard to believe this could be done.

No alt text provided for this image

Our experiment was quite simple in concept: during one week, we would ensure that unloading bays would be used for unloading only. However simple in concept, it was a big challenge in execution. It required an extensive deployment of traffic police: during the experiment, 20 police officers were present in our experimental area – compared to only 3 in any regular period. Effectively this meant that unloading bays were now actually available.

No alt text provided for this image

The results surprised me. I had expected at best 5% improvement in efficiency. We reached 44%. Delivery vehicles parked much more within loading bays than before (from 18 to 37%), reduced their total time parked by 38%, and reduced the total time driving in the area by 54%. Making available parking space for unloading hence reduces the time parked and reduces the driving time. Drivers reported much less stress, and in some cases were back at the depot an hour earlier.

Urban logistics is not so much a traffic problem; it is a parking problem

Obviously, this is just a one-week experiment, with only 10 vehicles, and we need to be careful with generalizations. However, even if the gains are only half of what we measured, the gains are substantial, and justifies additional research and experimental work on the topic of parking. As I have said at various occasions: I am more and more convinced that the urban logistics problem is not so much a traffic problem; it is actually a parking problem.

Note: The research reported above is still under peer review, which implies that we are still expecting validation by other academics on the quality and rigor of our work. However, we feel confident that the results and insights presented in this brief overview are robust.

This article was published on LinkedIn in July 2019

Acknowledgements: The unique experimental study that we were able to do in Querétaro has been made possible with financial and governance support of the Royal Netherlands Embassy in Mexico City, and with our partners: IMPLAN Querétaro, Heineken, Bonafont, Bimbo, Jumex, TomTom Telematics, and PTV Group.

Smart mitigation investments can help the European petrochemical industry manage supply chain disruptions

The 2014 incidents and subsequent outages at the Shell Moerdijk plant received a lot of media and industry attention. Incidents like this significantly affect the petrochemical market, both in terms of price and product flows. In general, 2015 showed unprecedented production issues in the European chemical supply chain, over 50 force-majeures had been declared across all polymers. Given the age of the European petrochemical production sites, operational reliability remains recognized as a major challenge to the industry today.

Given this operational environment, there is a strong desire to obtain quantitative insights in the long-term effect of supply disruptions on the value, forecastability and volatility of EBITDA and the effect of potential mitigation options. However, chemical supply chains are highly integrated networks, implying that a disruption in one plant affects operations throughout the entire supply chain. Therefore, mitigation measures need to be evaluated coherently. Furthermore, there is high uncertainty about the location, timing, and impact of the next major disruption, adding additional challenges to the investment decisions. After all, ‘nobody gets credit for solving problems that did not happen’.

To improve the insight in supply disruptions and potential risk mitigation options, our latest published research – jointly conducted with my former student André Snoeck and my former colleague Maximiliano Udenio – provides a methodology to identify, categorize and quantify risks and the impact of disruptions. A specific risk mitigation option impacts several risks, whereas a specific risk can be mitigated by several mitigation options. The kernel of the project is the development and use of a two-stage stochastic optimization model to analyze these complex interdependencies and dynamics in an integrated way.

We draw three main conclusions based on our research:

1.      Supply chain risk mitigation investment trade-offs in the uncertain and interconnected chemical industry are not trivial and require advanced quantitative methods, such as stochastic optimization.

2.     Smartly placed small investments may outperform a much more expensive poorly placed investment. For instance, a combination of small buffer inventories with reduced response times for external supply might outperform multiple investments in redundant processing capacity.

3.     Investments that mitigate multiple disruptions are undervalued when considering disruptions and mitigation options in isolation. For instance, individual disruptions might not justify investing in decreasing lead times of external supply. Our advanced model that captures the entire supply chain shows the aggregate value that the investment provides to the network.

4.     There exists a trade-off between long-term expected costs minimization and short-term risk minimization, where the latter leads to a more aggressive investment policy. This implies that an increased focus on a stable and forecastable quarterly EBITDA actually justifies larger investments in supply chain mitigation options compared to a long-term focus on expected future EBITDA. Ironically, this implies companies under private equity aiming short-term returns may need to invest more in mitigation measures than companies aiming long-term returns.

There is a clear need for the European petrochemical industry to take supply chain disruptions seriously. Chemical supply chains are integrated, interdependent, and complex systems and evaluating risk mitigation investments should be complemented by advanced quantitative models. Such methods are not new to the chemical industry, advanced models are being used to optimize cracker operations for years. Leveraging this capability when dealing with supply chain disruptions will positively impact the value, forecastability and volatility of EBITDA.

This article was published on LinkedIn in January 2019


The text above has been written for ease of public access, and may contain texts that have been simplified for this purpose at the expense of scientific rigor. In-depth and verified information can be found in our peer-reviewed journal article published in the European Journal of Operational Research. The article is based on a Master Thesis completed at Eindhoven University of Technology. 

Could a Brexit bullwhip cause turmoil in European industrial production?

Last week, the UK newspaper Guardian reported that in the UK companies have started massive stockpiling. While this is seen as a measure to counter uncertainties around a potential hard Brexit, little has been said about the additional production needed to pile up this inventory and the effects of a substantial decrease in production next year once the inventories are sold off. Linking such active inventory decisions to supply chain understanding can only tell us one thing: even a small inventory adjustment may lead to years of instability across European supply chains.

In the Guardian article, “industry representatives” are quoted to have said: “Frozen and chilled food warehouses, storing everything from garden peas to half-cooked supermarket bread and cold-store potatoes are fully booked for the next six months, with customers being turned away”. Not just fresh and frozen food is being stockpiled. British manufacturers are also storing ingredients. UK food manufacturer Premier Food announced two weeks ago that it is building up about 10 million pounds’ worth of ingredients inventory, and British Tobacco company Imperial Brands said it would add about 30 million pounds worth of inventory.

Of course, this is good business for those leasing out warehouse property. Apparently, in many places in Southern England it is very difficult to even find qualified food-grade warehouse space. Seems like a good opportunity to bring over some reefer containers to the island as my bet is that these may be leased at a premium in the upcoming months.

However, an aspect of stockpiling that has not received any attention in all of this is the bullwhip effect that it may cause in European industrial production. Exactly ten years ago, we conducted research that showed that the sudden decline in inventories caused by the collapse of Lehman Brothers led to massive global fluctuations in industrial output for many years. This so-called “bullwhip effect” was well-known in the literature and to anyone operating a supply chain, but common knowledge was that this was exclusively caused by demand fluctuations. Our work demonstrated that sudden, coordinated inventory adjustments would cause such fluctuations as well. I believe that this is what we will also see as a consequence of Brexit: the Brexit bullwhip.

What is a Brexit bullwhip?

I believe a Brexit-bullwhip could be facing us. The reasoning would be as follows:

  1. Inventories of raw materials, intermediate products, and consumer products are all built up in the UK over a relatively short period of time. (Note: this is not just happening in the UK, as UK exporters will build up inventories on the continent, but I have left this out of scope in my analysis for now).
  2. These inventories need to be produced; this leads to additional industrial production for those products sourced from the EU.
  3. Manufacturers observe an increase in demand for their products, and hence also order more supplies from their suppliers. This surge in demand propagates upstream in the supply chain. Since supply chains are long and not very transparent, these suppliers are unlikely to relate their increase in orders to stockpiling in the UK.
  4. At some point, the stockpiled inventory will reach their targeted level. Orders will go back to their “normal” level. It takes some time, however, before the supply chain adjusts. Cumulatively, across manufacturers, their suppliers, and again their suppliers, inventories could easily amount to a year’s worth of sales. As a consequence, manufacturers further upstream in the supply chain may only feel the consequences of the original stockpiling decisions many months later. Just like if you turn up the heating in your house; it takes some time for the system to respond and you may overheat your house.
  5. Some time next year or in 2020, hopefully supply chains will have been adjusted to post-Brexit border controls and inventories will start to be reduced again. This will further amplify the Brexit bullwhip, leading to a huge decline in industrial production.

How large could the Brexit bullwhip be?

This is not easy to estimate. Inventory records in national statistics are not very good. Also, the complex supply chain relationships are not captured in detailed statistics. But we should be able to make a first-order estimate of the effect.

First we estimate UK imports of (physical) goods from the EU27 to be about 300 billion euros annually (1) with corresponding inventory levels to be at about 6 weeks across the board (2). This would value inventory related to UK imports from the EU at around 34 billion euros.

Second, we estimate how much additional inventory is likely to be built up as part of the stockpiling process. For this, we can only rely on anecdotal evidence from the various newspaper reports, suggesting that this is an additional month of inventory, so about 25 billion euros of additional imports from the EU, most of them likely to be purchased in this quarter (Q4 2018).

Next we estimate total European industrial production, including the processing of agricultural products. Based on (3), we will work with a figure of about 2,500 billion euros.

Assuming these numbers are more or less correct, this implies that the inventory build-up in the UK would be 1% of European industrial production. Since virtually all of this would have been produced in the last quarter of 2018, the last quarter would then see an additional production of about 4%. That would be a massive number.

Note that none of this includes an additional bullwhip of overreaction. Our studies of the 2008 financial crisis suggest significant overreaction since it is not clear to companies further upstream in the supply chain what causes the increase in demand.

A huge drop in 2018?

If this were all more or less true, the rebound on the bullwhip in 2019 could be huge, with drops in import figures that would go well beyond the growth in the current quarter.

Obviously, this is an effect analyzed in isolation, and without any modeling at this stage. There are many other effects surrounding the Brexit which I am sure are receiving extensive analysis by economists in the UK, EU and elsewhere. Just today, the Bank of England released such an analysis. The Brexit bullwhip however deserves to be part of this analysis. If anyone has better numbers, I am happy to adjust my calculations accordingly.

Updates and comments after publication

(29 Nov 2018) Gaston Cedillo of the Mexican Institute of Transportation shared an interesting study where they studied the consequences of the variability at the US-Mexican Border. They show that additional safety stock will be needed to deal with the varaibility of the border process itself.

(29 Nov 2018) I received information that the demand for ADR (dangerous chemicals) containers has been increasing significantly. This would suggest that also further upstream in the supply chain (chemicals are typically upstream), companies start preparing for adding inventory.

(6 December 2018) The London Evening Standard today published an article based on this blog

(2 January 2019) The Netherland PMI, after slowly sliding in the past months, suddenly revamped in December. It is attributed to higher exports to the UK, indicating stockpiling.

(2 January 2019) The Guardian today published an article that also British manufacturing is working at full capacity to help build inventories ahead of the Brexit.

Sources and disclaimer

For the order-of-magnitude analysis, I have received assistance from my valued colleague Alan McKinnon. We rely on publicly available sources and realize our estimates may not be very accurate. However, even if we are 50% off, the Brexit bullwhip will be large. If you have any further or better data that we can easily incorporate, please get in touch. I thank Alan for his help; of course, all errors remain mine.

(1)   The source my colleague Alan McKinnon helped me use for this data is aParliamentary briefing document on UK – EU trade. This gives a figure of £341 bn for UK imports of goods and services from the EU in 2017 – at current exchange rate around €390 bn.  The briefing report does not give a breakdown of goods and services. According to a BBC report the value of service imports from the EU was £81.2 bn in 2016 or €92 bn. Assuming this value did not change much between 2016 and 2017 €300 bn might be a reasonable estimate for the value of goods imports.

(2)   Total value of product sales by UK manufacturers was £385bn in 2017 ( while the value of physical inventory in manufacturing industries (analysis by Alan McKinnon based on British statistics) (in current prices) at the end of that calendar year was £62.4bn – i.e. 16.2% of sales – average inventory rotation of 6.2 annually. This is about 2 months worth of inventory. Retail inventory may be a bit less, but is unlikely to be less than one month of inventory currently. Hence, we assume 1.5 months of inventory on average.

(3)   Eurostat suggest that EU GDP was €15373 bn in 2017 (in current prices) of which the UK contributed €2332 billion (15.2%). Removing the UK, reduces EU GDP to €13040 bn. Manufacturing represents around 16% of EU GDP, which using the Eurostat GDP figure for 2017, would yield a figure of €2460 bn. I am not sure at this stage what exactly is included in “industrial production” and whether this covers all of the producing sectors to export to the UK. At the same time, it definitely won’t include all of the wholesalers and traders in between, that may further aggregate the bullwhip. For now, we’ll work with the 2500 bn.

Disclaimer: the current analysis is not based on peer-reviewed research specifically focused on Brexit or Brexit-based numbers. We rely on simple order-of-magnitude calculations as a contribution to the debate.

Note: This blog appeared on LinkedIn on 28 November 2018

Ever thought about your supply chain’s water risk? It could harm your business unexpectedly.

Industrial production in the Indian State of Maharashtra, where Mumbai is located, got a major hit in 2016 when water supplies were short. In the same year, about half of the manufacturing companies in the Bolivian city of Cochabamba were affected by water shortages, leading to a decrease of 15% in industrial output. Last year, India’s apparel industry was heavily affected by water shortages, and this year even Cadbury and Jaguar in the UK had to temporarily close down their factories due to freshwater supply problems.

Water shortages are commonly seen to affect agricultural production and household water supply in underdeveloped parts of this world. Due to the globalization of sourcing, also manufacturing companies are increasingly faced with problems due to water shortages somewhere upstream in their supply chain. While supply disruptions due to natural disasters or terrorism have received substantial attention and more and more companies have been mapping their risk, few companies have realized that they may have a little time-bomb ticking at one of their (maybe far-away) upstream suppliers. A new research article in the Journal of Cleaner Production (“Water Risk Assessment in Supply Chains” – free download until 9 December 2018 at this link; after that, please contact us to request a copy) by my co-authors Torben SchaeferMaximiliano UdenioShannon Quinn and myself sheds light on this risk and provides a methodology for companies to take action in mapping their supply chain’s water risk.

What is water stress?

The availability of clean water is one of the most important sustainability challenges we are facing today. It is a challenge that is expected to increase in the future; and yet, its visibility by supply chain managers community is limited. An area is said to be experiencing water stress when the amount of clean water available is smaller than the amount of clean water required. Clean water is a precious resource. Civilian populations need clean water to live, agriculture and farming need water to produce our food, and virtually every industrial process needs clean water to function. However, even though clean water is a universal requirement, in the face of its unavailability the response is clear: the priority for securing access to clean water will always be for civilian populations and food production. There are numerous examples in recent years where governments divert limited water resources such that water is allocated to civilians. In such cases, industries must continue without water, or shut down. In a globalized world, production in any particular location often depends on raw materials or components sourced from across the globe. Supply chains literally span the earth.

What should you do as a supply chain director?

For firms interested in reducing their exposure to water risk, this means that they must monitor the situation of their entire supply chains. If your suppliers, the suppliers of your suppliers, or the suppliers of the suppliers of your suppliers are located in an area that runs the risk of access to clean water, you run the risk of supply or production disruptions due to access problems with clean water.

In our new article we term this risk, relevant for firms across industries and across the world, water risk.

It is not easy for a company to understand your exposure to water risk: data is difficult to collect, and even when data is available, it is difficult to compare the different dimensions of the problem to come up with a metric that summarizes their risk exposure. Therefore, we developed a new index intended for companies to measure and understand the water-risk of an entire supply chain. Our risk index is composed of 6 base indicators that each measure a different dimension of either “physical” water risks (baseline water stress, seasonal variability, and drought severity) or “amplifying” water risks (external dependency ratio, governance and regulation, and infrastructure). 

Given a geographical location, our risk index quantifies the water risk using a single number, by taking into account the aforementioned components, and experts’ assessments to weigh each of these components into the single water risk indicator. Looking at the problem from a strategic perspective, our water-risk index allows firms to assess geographical areas in terms of water-risk. This allows you to, for example, compare geographical locations of potential new suppliers, or forecast the water risk of your current supplier base for the next decades.

Moreover, our risk index allows for tactical analysis of the water risk at the level of individual processes, products, or manufacturing locations. In this way, you can identify the products or processes with the highest water footprint and consequently analyze its supply chain to detect water risks.

Application at Procter & Gamble

We worked together with Procter & Gamble and applied our methodology with them. From a tactical perspective, we were able to immediately identify some suppliers of a critical raw material that are located in areas with high water risk. Moreover, the water risk in these areas is expected to increase in the coming years. From a strategic perspective, we mapped the water risk of their more than 1000 suppliers to identify the suppliers and areas with highest risk today and in the future.

Procter & Gamble have not only realized the potential of water risk assessment for the reliability of their supply chains, but also for the communities where their supply chain’s plants are located. Consequently, in its recent new sustainability strategy, specific targets have been included for reducing the company’s water footprint, and with that, likely also the company’s water risk.

Including water risk in your supply chain risk analysis is critical. The great news is that acting on this will not only reduce the vulnerability of your supply chain, but also make many communities that face water shortages a better place to live.


  1. This blog has been written with the purpose of making our research accessible, sometimes at the expense of nuance and methodological limitations. A full evaluation of our work should only be based on the peer-reviewed article itself.
  2. This blog has appeared first on LinkedIn on November 26, 2018.

The Death of Supply Chain Management – Really?

A rebirth of supply chain management

This blog was published on LinkedIn on June 19, 2018. The LinkedIn article contains all relevant links.

Yesterday the grand title “The death of supply chain management”[1] triggered my thinking, as it must have worried many that are currently working in some supply chain role. Re-reading the controversial blog on HBR a few more times, I first concluded that the title did not match the content (which is much more about a new era for the supply chain as increased algorithms and visibility come into play) but later realized that the perspective that is being sketched is different from what I think is happening. And current and past research on supply chain operations planning can help us better judge this.

Supply chain planners have decisional, informational and interpersonal roles

First, it is important to be aware how most supply chain planners are spending their time. Since many decision support tools (including Microsoft Excel as the most used tool for planning) have entered into the market over the past twenty years, the way a planner spends her time has more and more moved away from actually making decisions (their decisional role) to other key aspects of the planning role. These other aspects can be classified into two main categories [2]. The first one is informational. The informational role is about collecting relevant information that typically is not present in an ERP system. This could pertain to soft or ambiguous information (a promotion being in the pipeline but not yet decided upon, an estimate of the chance of winning a large tender, or a carnival week coming at a major supplier likely causing disruptions in supply), or simply data that are currently not linked or not visible. The second role is an interpersonal role. For instance, if a product is short on supply, a good planner will be able to call his “friends” elsewhere in the operation to speed matters up. Technically this implies that the lead-time parameter in the system may – sometimes but not too often – be expedited. Of course, this favor is kept in a mental account of the supplier’s planner. The supplier’s planner will expect this favor at some point to be returned. The interpersonal and informational roles start to overlap when others in the organization consult the planner on getting to know information about something that is not (yet) in the system. The planner usually knows best among his organization’s peers. And earlier work by my former colleague Ton de Kok has shown that even between the most advanced algorithms and the actual performance reached by planners, there is a gap – usually demonstrating the added value of the human planner in modifying the real supply chain beyond what a model can do. We have also done some work in a retail environment showing this added value of the human decision maker compared to state-of-the-art models [3]. For sure, there are many situations in which an algorithm may outperform a human, and our understanding of this is developing gradually.

Supply chain planners spend most of their time collecting, verifying, and disseminating information

As our research [4] and those of others have shown, planners spend (or “waste”- depending on the perspective) more than half of their time on the informational role. Typically they spend less than 25% (and a much lower number in manufacturing supply chains) on the decisional role: actually deciding on the plan, schedule, or replenishment. The AI impact is often presumed to have an impact on this decisional role. I would contest this for a variety of reasons not elaborated on here. The potential of big data and AI is actually in impacting the efficiency of the planner in the time she wastes on the informational role.

Interestingly, in the HBR blog, many examples refer to control towers and real-time information. In most supply chains real-time information is too late, as the information is in almost all cases only useful ahead of time when it is still possible to do something with this information. However, having more information readily available, well-searchable, and visualized in an intuitive way, brings great opportunities to increase the efficiency of the planning process. Not by eliminating the planners’ decisions, but by reducing the time they waste on collecting information.

This line of thinking is much less sexy than having all decision makers replaced by robots as is suggested by many. But currently replacing people by robots just leads to what I heard from a major global CPG manufacturer supplying to a European retailer that took its hands off the wheel. The CPG manufacturer now needs to monitor all orders to take out the “crazy” ones, call (!) their client to double check these orders, and then manually reset everything in the supply chain. This really would be the death of supply chain management.

[1] Allan Lyall, Pierre Mercier, and Stefan Gstettner (2018) The Death of Supply Chain Management, HBR blog, June 15.

[2] The actual classification of roles is more extensive, as planners typically also do maintenance tasks of master data and ensure the plan or schedule gets executed, but for the sake of simplicity I have grouped these under informational and interpersonal roles. If you are interested, read the early work by Sarah Jackson: Jackson, S., Wilson, J. R., & MacCarthy, B. L. (2004). A new model of scheduling in manufacturing: Tasks, roles, and monitoring. Human factors, 46(3), 533-550. This early work has been done in manufacturing settings, but much of these also apply to inventory planners in retail and transport planners.

[3] Van Donselaar, K. H., Gaur, V., Van Woensel, T., Broekmeulen, R. A., & Fransoo, J. C. (2010). Ordering behavior in retail stores and implications for automated replenishment. Management Science, 56(5), 766-784.

[4] Larco, J. A., Fransoo, J. C., & Wiers, V. C. S. (2018). Scheduling the scheduling task: a time-management perspective on scheduling. Cognition, Technology & Work, 20(1), 1-10.


This blog was published on LinkedIn on June 19, 2018. The LinkedIn article contains all relevant links.

3D printing will impact global trade, but much less than previously thought

By Bram Westerweel and Rob Basten (Eindhoven University of Technmology) and Jan C. Fransoo (Kuehne Logistics University)

Last fall, ING released a report on the growth of 3D printing as a manufacturing technology. The report includes a scenario in which the rapid growth in of 3D printing equipment would lead to a total share in the global manufacturing equipment of about 50% in 2040. This in turn would lead to a dramatic drop in cross border trade in goods of 38% in 2040.

3D printing will have a significant effect on the manufacturing industry and on global supply chains. However, our analysis shows its effect to be much smaller than ING’s scenario predicts, as our analysis concludes that the decrease in cross border goods trade will likely be less than 7% rather than the 38% announced by ING. Our conclusion if different because we believe the ING report to contain a number of flaws. We outline our reasoning here below, and created some graphs to make that clear.

The key assumptions in the ING report are (on page 8):

  • The annual growth rate for investment in 3D printing has been 29% over the past five years, compared to an average of 9.7% for global investment growth in traditional machines. ING assumes that this difference continues to hold in future years.
  • In the ING scenario, investments double (to 58%) for 3D printing after five years, while it will fall by a third (to 6.5%) for traditional machines after ten years.

Our analysis of the ING report and the questioning of the above assumptions leads to our following findings:

  1. The ING report confuses annual investments with the installed base.In the ING report, the current rapid growth in annual investment in 3D printing equipment is reflected proportionally in the share of the manufacturing equipment base. However, manufacturing equipment has a long life cycle. In our analysis, we have assumed, for the sake of argument, a life cycle of 20 years. This implies that many of the assets that are acquired in the upcoming decade will still be there in 2040. Given that the far majority of these assets is still in traditional manufacturing technologies (despite the smaller growth rates), in 2040 a much higher share will be still in traditional manufacturing. Inserting this effect into ING’s calculations reduces the projected share of 3D printing equipment in the total manufacturing equipment to 35% in 2040 (instead of 50%), as shown in the TU/e line in the figure above. Note that while this number is clearly lower than ING’s, this is still very impressive.
  2. The ING report assumes an unsubstantiated annual growth rate of investments in 3D printing equipment of 58% from 2022 onwards. The growth rate in 3D printing equipment is 29% in 2017, according to ING’s source. In the scenario under consideration, ING assumes that this growth rate doubles in 2022 and then continues at this level until 2040 (and beyond). We believe this to be highly unlikely. Growth rates are likely to level off at some point, and unlikely to double on such short notice. The source of this doubling in 2022 and the subsequent persistence of the annual growth rate of 58% for decades is unclear and not mentioned in the report. To illustrate the dramatic effect of this assumption on ING’s calculation, we have used a growth rate of 40% (which we still believe is extremely high) from 2022 onwards. We then find that 3D printing equipment makes up only 8% of the total manufacturing equipment in 2040, as illustrated in the bottom line of our figure.
  3. The ING report assumes annual growth rates in total investments that are unrealistically high. We are unaware of the source for the 9.7% growth rate in conventional manufacturing equipment that ING assumes. This number may be realistic for a limited period of time, during the current crisis recovery period. However, having year-on-year, until 2040 (and beyond), an annual growth of 9.7% in the investment rate seems unrealistic to us. Still, let’s assume that this is realistic. If we then go back to ING’s scenario, we see that in 2040 the annual growth rate in total manufacturing equipment, i.e., conventional plus 3D printing, is 21%. This effect is due to the extremely high investment growth rates in 3D printing, which by then makes up an increasingly large share of the total manufacturing equipment. With (highly optimistic) growth rates of 40% from 2022 onwards, annual growth in total manufacturing equipment is still 9% by 2040. This type of growth seems unfounded, with a substantial impact on ING’s conclusions.

We are a strong believer in the future of 3D printing technologies. Application in personalized devices, spare parts and – ironically – making flexible tools for traditional manufacturing technologies, will lead to massive changes in many factories. However, the timing and impact on global manufacturing will be very far from the numbers presented by ING. It is unclear to us how ING has translated the impact on global manufacturing exactly to an impact on global trade, but if we simply scale the reduction proportionally, our results at point 2 above would mean that instead of a 38% reduction in trade by 2040, the reduction in trade would be less than 7%.

Besides the much smaller impact on global trade, our analysis also illustrates that models with exponentially increasing growth over long periods of time are unsuitable for conducting such analyses. Even under our adapted assumptions, annual manufacturing equipment growth rates will approach unrealistic values as time progresses as 3D printing would ultimately completely dominate traditional manufacturing. A more credible analysis should, therefore, be based on a more realistic model that allows the ratio of 3D printing and traditional manufacturing to reach some equilibrium, if one wishes to make any claims on the timing and impact of 3D printing on global trade.

Note: We have received a response to a draft of this text from ING’s Raoul Leering. This has clarified a few points, resulting in some changes in the text. His key comment is that the “report does not predict anything! It is only a scenario analysis that shows what happens with world trade IF the current growth rate doubles.” The ING report contains a second scenario in which investments in 3D manufacturing equipment equal traditional manufacturing equipment in 2060. In this response, we focus on the 2040 scenario.Furthermore, ING has not responded to our request to share with us the source of their investment data.