As if the human effects of coronavirus were not tragic enough, it is now clear there will be a huge economic and societal impact, too. The short-term forecast for hotels, airlines, conference providers, restaurants, retailers, and movie theaters are bleak. It will take some time to measure the true cost and calculate the damage, in lives cost, failed companies and lost jobs + income. There will be a handful of beneficiaries, perhaps dried pasta and toilet roll manufacturers achieving record levels of business – assuming they can continue getting their goods to the stores – and that’s a big assumption.
Supply chains worldwide are reeling. The Wall Street Journal recently reported that congestion and interruption at Chinese ports had severely disrupted the normal patterns of trade. This has resulted in a lack of empty containers on the west coast of America, especially the refrigerated units (“reefers”) which are needed right now to export agricultural products such as vegetables, meat and citrus fruits. They noted that more than 370 container ships are idled in the Pacific, and container terminals at the Port of Los Angeles and Long Beach are operating at 1/3 of their usual gate capacity. The reefers available are commanding a $2000 surcharge over standard rates, essentially doubling the cost of sending a crate of oranges to Asia. The end result may be a lack of food for humans and animals, product waste as good food is forced to rot because it is in the wrong location, and economic damage and suffering to many agricultural workers and organizations. Companies are battening down the hatches, minimizing all expenditure, and doing their best to survive these uncertain times. The best organizations and leaders have also adopted policies to support their staff, suppliers, and customers, with some great steps taken by firms such as Apple, McDonalds, and many others.
What if there were a better way of coping with disruption?
As Nicholas Taleb pointed out in Black Swan, you can’t predict what type of unforeseen event will occur in the future, but you can predict that there will be such events. The system that has demonstrated greatest success in dealing with such changes is nature. A large animal can accidentally knock down a termite nest, but the insects will rebuild it, and survive. A wildfire can ravage a forest, but it will recover. A giant meteorite can strike the planet and blanket the skies in ash for years, but many species (including our predecessors) will find a way to survive and prosper. The difference often comes down to a few key factors:
- Size
- Agility (often linked to size)
- Collaboration
Some of the most successful species on the planet are smaller creatures that work together in a nimble fashion to solve problems; ants, bees, humans, even bacteria. Modern research is showing that forests are also connected via huge fungal networks (the mushrooms we see are simply the fruits of huge underground systems) and collaborate far more than we imagined. Many of these groups, and even the human brain with its neurons, are networks of independent elements that combine to produce emergent properties such as intelligence, or even consciousness. There is a wonderful story in The Lucifer Principle by Howard Bloom, which looks at the behaviour of a colony of bees (actually from original research by Thomas D Seeley in his 1985 book Honeybee Ecology). The story is so good, that it stuck in my mind for years:
“In one experiment, scientists began by placing a dish of sugar water at the edge of a hive. Over the course of time, they moved the water, first a few inches from the hive, then a few feet, then a few feet more - always increasing the distance by a precise increment. The researchers expected that the bees would follow the dish and cluster around it. To their surprise, after a few days, the insects were doing far more than merely tagging along after the moving sugar water. The bees would fly from a hive and cluster on a spot where the dish had not been placed - the site where the insects anticipated the dish would be put next - and their calculations were right on target.”
Astonishing behavior by a swarm, but unfortunately, in too many of our human managed processes – such as those in a supply chain – we still rely on a centralized command and control system. In dynamic times, when a black swan event causes disruption, these traditional systems are rapidly overwhelmed.
A supply chain example.
Sam is a supply chain manager at a food producer, who decides which loads to send to each customer every day, taking into account the number of pallets, weight, etc. and also what stock should move from the production site to the distribution centers. Sam has to figure out how to combine loads to fill the third-party logistic (3PL) trucks, and which firm to use for each trip based on the cost, region and type of loads (some require refrigerated trucks). There are also inventory levels, typical customer orders, and the demand forecast to take into account. Sam has several spreadsheets with summary costs and data, plus years of experience to help in deciding the best options. If something changes, like a 3PL firm going out of business, or a new customer unexpectedly requesting a large order, Sam can handle it.
What happens when coronavirus hits? Suddenly, the demand forecast is changing erratically, several of the preferred 3PL’s are not available, and the others have ramped up prices. Some of Sam’s customers want a halt on deliveries, or else they want more products, but not necessarily those held in the distribution centers’ inventory. What’s more, the company's goods are perishable, so there is a risk of substantial waste – Sam needs to understand the sell-by dates for thousands of items stored at the distribution sites and figure this into the planning cycle. Normally, this is not an issue as the typical cycle operates within the shelf-life of the products. The CEO has mandated that all staff work from home, cutting Sam off from some resources he might use to help make these critical decisions. On the third day of crisis management, Sam develops a fever and cough, and is forced to pass responsibility to a junior planner, who looks at the mess with absolute horror. The department tries to cope as best they can, but the end result is a company loss totaling millions of dollars. If margins were tight before the crisis, this may be the end of their business.
The sheer number of elements in play in this supply chain example has produced a combinatorial explosion of options. A seemingly small number of factors can combine to make decisions nearly impossible. For example, a problem with six factors that can be combined has 720 possible solutions. With twelve factors the solution space rises to 479 million choices. At 24 factors you are looking at 620 x 10^21 possibilities. No person can be expected to make a reasonable decision in those circumstances, and even the fastest analytic-based system would take weeks or months to reach a conclusion.
How could this be different?
If we look to nature for our solution and see where the latest AI research is being brought to market, there is a hopeful picture. Let’s imagine that Sam uses a next generation multi-agent system to recommend the loads each day. This is a network of software agents that operate collaboratively like a swarm of bees, or a nest of ants, using nature as a model. The system has a simple goal, which is to minimize the total costs of:
- 3PL fees (Per journey / route, Minimize number of journeys by combining loads, Ensure truck is full not LTL (less than full load))
- Product waste (Check perishable dates, Ensure best blend of products moved to distribution centers)
It might be fed data from a 3PL portal with the latest prices available, plus the demand forecast, and the customer order processing system. It may also have been instructed to take into account constraints such as:
- Max weight per truck
- Max pallets per truck
- Customer order delivery ‘slot’ (e.g. 12-2pm on Thu 13 Mar)
- Products of class ‘Cool’ must be on a refrigerated truck
There are still a huge number of options here, because of the combinatorial explosion, but a multi-agent system takes a different approach to solving it. Rather than trying to calculate every possible option, the agents try a range of solutions and compare the results. If one iteration is better or worse than the last, the agents can decide whether to persist in the new approach, go back to the old, or try a new starting point. There are many different algorithmic approaches the agents can take, with radical new approaches coming out of academia and software vendor labs on a regular basis, and some systems use multiple approaches and choose between them. The goal is not to necessarily find the very best solution (known as the global maxima), but to find the best option (the local maxima) within a given timeframe. Given the proliferation of options, and the speed and accuracy of computers, the choice will almost certainly be superior to anything Sam could have picked. The system will also work just as effectively for the junior planner, or any other member of the supply chain team.
Am I not losing all control?
The key here is that the goal and the constraints are still controlled by the human. The multi-agent system is simply helping the operator find the best solution possible. What’s more, the results are transparent – the human can see the planned list of loads by 3PL carrier, and decide if it makes sense, because the operational plan is an output of the system, and not a black box decision. Which leaves the human operator fully in control of the final decision. They can adjust parameters and constraints to alter the result if they so desire, or if new information emerges that is not yet in the system (like a specific warehouse being closed).
The other thing the multi-agent system can do, is link aspects of the process together. It may be that the demand forecast is unreliable, because it suffers from many of the same issues as load planning. A different multi-agent system could be deployed to solve this demand problem, and then linked to load planning, creating one large optimization. This can work even when there are conflicting objectives – in fact, this is where the benefits really stack up, but perhaps that is a story for another post 🙂
Why didn’t you tell me this two months ago…?
You’re swamped, I know. The supply chain team can barely come up for air, let alone consider a new approach. I’m not suggesting you adopt this today. Right now, there are too many other critical tasks that will determine the survival of your business, and the health and welfare of your employees and customers. Remember, though, that the coronavirus is just this month’s black swan event. It may mutate and return next year in a more virulent form, or there may be a different crisis – a trade war, a real war, an asteroid strike. The point of black swans is that we don’t know what they will be in advance, but we know another will inevitably appear.
This particular black swan has been a tragedy in human lives – those lost and affected, and I can only hope that the measures put in place will reduce its impact in the coming months. Coronavirus arrived a few months too soon for the latest technology to really help global and local supply chains, but I believe we will be in a better position to tackle the next one. Once the dust settles, it may be time to build the capability to automatically manage disruption into your supply chain. If you would like to know more, please feel free to reach out. Tackling inefficiency in supply chains isn’t all about money – it prevents food waste, keeps people gainfully employed, reduces emissions by improving logistics, and ultimately feeds the world. That’s something we are proud to be involved in. Stay safe.
