Agility and adaptability are the answers
Donald Rumsfeld was a US Defense Secretary in the 1970s and again in the early 2000s. He became known for his entertaining and difficult-to-understand answers at press conferences, so much so that the BBC in London would regularly broadcast the most noteworthy of his latest obfuscations. But there is one particular press conference phrase for which it seems he will forever be associated. As we know by now, this is the famous “known-knowns, known-unknowns, unknown-unknowns” exchange. In this exchange, Rumsfeld answered a reporter’s question, as follows:
“There are known knowns. There are things we know we know. We also know there are known unknowns. That is to say, we know there are some things we do not know. But there are also unknown unknowns, the ones we don't know we don't know."
Huh? Understanding this the first time through is like trying to understand the meaning of a poem the first time through. In other words, it probably had the effect Rumsfeld desired.
Afterwards, this became not just an entertaining riff, but it also created a lot of discussion regarding the underlying meaning. It’s also a testament to the power of words and ideas that Rumsfeld became almost as well-known from this exchange as he had from his fifty-plus years of public service. In fact, the attachment was such that he titled his memoir “Known and Unknown: A Memoir.”
It’s important to point out that Rumsfeld was not the first to use these words or the ideas behind them. He simply repurposed them at the right time in the right context. He also seems to have popularized it as a framework; it has been widely used in the past 15 years across a wide variety of domains.
What Rumsfeld was explaining is a framework for cataloging risks. Last month, I published an article titled “Supply Chain Volatility,” in which I discuss volatility in the context of the supply chain stress caused by the coronavirus pandemic. A number of readers astutely pointed out that you cannot know, a priori, everything that the future holds. Given that, how do you deal with unknowns?
Here, I will use the Rumsfeld Known-Unknown framework as a tool to address these questions and add to the discussion. I also venture into a fairly wide-ranging discussion on supply chain dynamics, all based on the earlier variability-volatility discussion. My apologies for perhaps not staying on point, but, like a lot of things related to supply chains, these topics are highly intertwined.
Note: in the earlier article I discussed agility as the antidote for volatility. In this article I draw a distinction between agility and adaptability. While I draw this distinction, I would like to point out upfront that agility is the enabler of adaptability.
Supply Chain Volatility Matrix
First, let’s clear up what Rumsfeld meant in the known-unknown dialog, putting it in a supply chain context. To do that, we will adapt the known-unknown construct to a 2x2 framework (this has been widely done over the course of the past 15 years for various risk contexts). In this situation the 2x2 is used to catalog events – either internal or external to companies – that create supply chain volatility. The general matrix is shown below.
Along the X axis are the things that we know or do not know; therefore, the left half of the chart represents things we know; the right half of the chart represents things we do not know. In a supply chain context, this means there are certain events that we have seen before, or even if we have not personally seen them, we know about them. This is a binary axis – we either know about them, or we don’t. The Y axis represents what we know about the impact of events (this is slightly different from what Rumsfeld meant, but is useful for this discussion). As shown in the above, we will use an of X-Y approach (x-axis first, y-axis second) when referring to the four quadrants of the 2x2.
The Rumsfeld framework originally included just three of the four 2x2 quadrants; there was no unknown-known quadrant (lower right corner). This quadrant was suggested afterward to represent things we know but choose not to know; in other words, these are things we choose to avoid or fool ourselves into disbelieving. This is the Ostrich quadrant. It might also be called the “you can’t handle the truth” quadrant. I will refrain from commentary as to what might go into this quadrant. We can all think of many things that might go into this category. Pain avoidance, short-term thinking, convenience, inertia, hubris, and just plain stupidity all contribute to things that might go into this quadrant.
Now let’s see if we can make this a little clearer by cataloging volatility-causing events into the quadrants.
A lot of these event categorizations are of course, in the eye of the beholder. What is known to some may be unknown to others (therein lies opportunity. See the book “The Big Short,” which I recommend people read about every five years).
For example, certain events – pandemics, volcanoes, financial derivatives – are in fact known, but only by a small sliver of the population who cannot convince the rest of us to get educated and prepared for the consequences. This means that for a lot of us, these are unknown-unknowns, when in fact when we look back in retrospect, they were known-unknowns. Bill Gate’s prescient warnings about the pandemic are an example of this. In fact, he now regrets he was not more forceful in his arguments.
For example, some have suggested that while the pandemic may have been a known entity, its economic impact was a big unknown. This makes it a known-unknown in the 2x2 framework. As I have previously said, this is probably true, but once you have a pandemic, your position on the chess board changes considerably, therefore changing significantly the conditional probability of a whole host of other events.
It’s interesting to consider climate change in this 2x2 framework. What makes it interesting is that it could fall into every one of the four 2x2 quadrants, depending on whom you ask. Some people say it’s a known-known, some say it’s a known-unknown, some say it’s an unknown-unknown, and then there is a certain segment that would just rather not talk about it, so it could also be considered an unknown-known.
As the arrow in the above chart indicates, unknowns become knowns as time unfolds. This is the learning process. With each new unknown making itself known, we learn, understand, and develop playbooks. This process is critical to developing resilience. Alas, there are obviously some people and organizations that are better at this process. What do they do better than others? This is explored further below.
In the above chart, I have included in the unknown-unknown examples a couple of interesting items:
Poor current decisions
Poor past decisions
Disruptive new competition
Multiple interacting knowns
These are unknown-unknowns because it is almost impossible to say that a series of decisions resulted in one outcome or another. As Daniel Kahneman and others have pointed out, it’s also difficult to control for the factor of luck.
Furthermore, in complex systems (described below), it may be impossible to know beforehand the impact of a sequence of events, each of which individually seems benign.
What’s Needed in Each of the Quadrants?
Agility and adaptability are the key capabilities needed for dealing with different volatilities. Agility is necessary for dealing with known volatilities, and adaptability is necessary for dealing with unknown volatilities. I’ve added a third capability – enlightenment – that’s necessary for dealing with unknown-knowns – those things where we deceive ourselves or where we fit data to our biases.
What are agility and adaptability?
Agility is about responding rapidly and intelligently using existing supply chain structure and policies; adaptability is about rapidly evolving the structure and policies to suit changed business conditions (e.g. selection stress, discussed below). Agility exercises existing muscle; adaptability creates new muscle.
The things that make you agile also make you adaptable, so in many ways adaptability is really a subset of agility. Here, we distinguish agility from adaptability for discussion purposes.
Hau Lee wrote an excellent article in HBR in 2004 titled “The Triple-A Supply Chain.” His research and experience showed that supply chains that focus largely on efficiency and cost underperform over the long-term. The article identifies three key attributes of long-term successful supply chains – agility, adaptability, and alignment. The discussion here is consistent with the definitions in that article.
But what forms do agility and adaptability take? To answer that, let’s start with a short discussion about supply chains as complex adaptive systems.
Supply Chains as Complex Adaptive Systems
It is often said that supply chains are complex. But let’s try to put some formality around what is actually meant by “complex” by putting supply chains in the context of the definition of Complex Systems (capital “C”).
A Complex System is a system composed of many parts which interact with each other. There are a couple of key characteristics of complex systems:
The behavior of the overall system cannot be predicted by looking at any one component. In other words, the behavior of the overall system emerges from the interaction of its components.
The behavior of the overall system is dependent on its current state and can be highly sensitive to changes in input. For example, a small change in input can result in large change in output. This same change at another time will result in an entirely different output. Systems with such behavior are said to be nonlinear.
From this definition, it is clear that a supply chain is an example of a Complex System.
The following additional material from the University of Waterloo Institute of Complexity and Innovation is pertinent to supply chains:
“Complex adaptive systems — predominantly living systems, including human social systems — exhibit all these features; but, in addition, they survive and reproduce within dynamic selection environments. To do so, they have sets of embedded rules that guide their action in response to their external environments. These rules evolve under selection pressure.”
In a business context, disruption creates selection pressure. The pandemic environment is a business example of high selection pressure. Those companies that evolve their rules the best will survive and even thrive in the future. What are rules in the context of supply chains?
Rules are what I call policies. Supply chain policies are measures and stipulations about how a supply chain is structured and run in order to execute business strategy. Policies are considered at different levels:
Strategic policies – policies typically dealing with deployment of physical assets
Operational policies – policies typically dealing with how physical assets should behave once they are deployed.
Policies are driven by business strategy, which is a company’s game plan for winning in the market.
Strategic, or structural policies, are typically considered at time of design of the supply chain and then periodically reviewed quarterly or yearly. In recent years, this review process has occurred with increasing frequency. Times of intense selection pressure, or stress, force companies to review their strategic policies. The number, location, capacity, and links of retail locations, warehouses, plants, and suppliers are all part of this area. So are single versus multi-sourcing, outsourcing, and off-shoring. Intertwined with physical considerations are segmentation, postponement, and inventory and asset business strategies.
Operational policies are also considered at time of design and periodically reviewed and adjusted to changing business conditions and business strategies. What are operational policies? Consider the following different business strategies:
Drive market share for product A in the Nordics region in the brick and mortar retail channel
Drive margin for product B in the Nordics region in the brick and mortar retail channel
The underlying supply chain policies for products A and B have to be significantly different. For example, for product A, I want to make sure there is always inventory available and I will probably ship single units from anywhere and to anywhere to save a sale. For product B, I may accept some stockouts if they prevent me from having to incur markdowns, and I may ship in bulk on a regular basis to take advantage of economies of scale.
Policies reside in software configurations. When software is implemented, policies are configured to reflect the business strategy of a company. For example, a stocking location for a product will have an inventory policy that reflects how often and in what quantity the inventory will be replenished. A fulfillment policy reflects the priority of customers and what to do in times of shortages or excesses. A production policy reflects how often to run a particular product before changing over. A supply policy reflects minimum inventory levels and frequency of replenishment. Many supply chain policies are a reflection of trade-offs between efficiency and responsiveness.
It’s important to ensure that policies across different functional domains are synchronized back to business strategy. For example, a highly responsive fulfillment strategy that is linked to upstream strategies based on bulk production and distribution will result in costly mismatches.
Businesses may choose to have different business strategies at the intersection of Product-Geography-Channel-Customer. If businesses operate with this level of granularity, it’s critical that policies embedded in software have this same level of granularity.
Agility is the ability to react within existing policies; adaptability is the evolution of policies to stay ahead of market conditions. However, agility is the enabler of adaptability. For example, in order for a company to be agile, it may have to periodically break policies in order to do the right thing for a customer. If it has to regularly break policies in order to adequately serve customers, it must evaluate the underlying policies to see if they need to be adapted to a new reality. This is an example of a company’s agility enabling its adaptability (further discussed below).
Pandemic Exposes Adaptability Winners and Losers
The current pandemic has created significant selection pressure. This is showing itself prominently in the retail space with Walmart, Costco, Target, and Amazon accelerating in the market, and already-weakened others filing for bankruptcy.
As has been shown in the retail space for the past decade, high debt loads significantly reduce a company’s adaptability. Inability to invest, or investment in the wrong things, is a business example of selection stress. This has played itself out in e-commerce for more than a decade. Companies that invest heavily in the right things are selected for survival. Companies that don’t – while they may continue to live for some time – will eventually be selected out.
This dynamic is particularly strong in technology spaces such as software where high levels of R&D investment are critical. Underinvestment over time leads to what I call the Hemingway factor, which comes from the book “The Sun Also Rises.” In the book, one of the characters, Mike Campbell, was at one time a wealthy man. He has the following exchange with another character, Bill Gorton:
“How did you go bankrupt?” Bill asked.
“Two ways,” Mike said. “Gradually and then suddenly.”
The business landscape is littered with once-wealthy companies that were undone by the Hemingway factor. In many cases, the business environment simply changes faster than the companies themselves. This is a near-identical parallel to what happens in biological systems. Shocks to the system, such as the current pandemic, accelerate this process.
Dealing with Unknowns
So, how do we deal with unknowns?
Do businesses need to employ a person like Herman Kahn to Think the Unthinkable and devise strategies to deal with such unthinkables? This is an attempt to uncover, a priori, all of the unknowns and turn them into knowns, and to then determine their impact and then develop playbooks for dealing with them. This is an impossible task, not just because the unknowns are constantly increasing, but because it’s actually sequences of events and their consequences that have also proven very difficult to catalog.
I should point out that companies do have chief strategists, who have the job of looking around corners, figuring out where things are going, and evaluating and devising strategies for new competitive threats.
The critical ingredient necessary for dealing with unknowns is not just a plan, but the process of planning. Robust planning is a critical ingredient to agility, and by extension, adaptability. Strategists can create a plan, but it’s more important that they create a robust process of planning, enabled by robust technology.
This is one of the reasons that Dwight Eisenhower, he who led the largest amphibious military invasion in history, said “Plans are worthless, but planning is everything.”
Eisenhower goes on to add the following:
“There is a very great distinction because when you are planning for an emergency you must start with this one thing: the very definition of “emergency” is that it is unexpected, therefore it is not going to happen the way you are planning.”
What he was saying is that the result of planning is not just the plan, but more importantly a set of capabilities that provides knowledge, alignment, agility, and resilience. You see, a plan is a snapshot, but planning is a living, breathing process of discovery, learning, collaboration, and decision making. It uncovers risks, opportunities, contingencies, and choices. In other words, Eisenhower was saying “plan to re-plan.”
Perfect Plan or Perfect Planning?
Over the years, various sources have predicted the death of planning. The reasoning went along these lines: the world is changing at an ever-faster pace making planning less important; what’s more important is the ability to react. There is an element of this that rings true. However, the general statement is incorrect. In fast-changing environments, planning becomes even more important. What’s become critical is linking planning with execution in a very dynamic way. The truth part is that the tremendous focus on the plan itself has become less important.
But what exactly is a plan? A plan is a statement of intent and means at a given point in time. It is an organizational contract, or agreement, to achieve something. “We intend to deliver $100M in revenue, $10M in net profit, and 15% return on capital through the following procurement, production, and distribution.” At the next point in time, the intention may still be the same, but the detailed way by which it will be achieved has changed. For example, the plan originally called for one shift at the plant, but unforeseen events now call for two shifts.
Creating a precise, perfect plan has been a focus of optimization experts for many years. Creating a perfect snapshot is important, but more important is your ability to rapidly and intelligently edit your snapshot as events unfold in real-time.
Furthermore, a lot of the emphasis for plan quality is placed on optimization (and/or heuristics) techniques. These techniques are run against models of the real world. These models – fashionably called digital twins today – are only as good as their ability to represent the real world at a given point in time, and their ability to evolve that representation as the underlying real-world changes.
Companies like Kinaxis – where I am a board member – have figured this out by applying machine learning to continuously adapt models as the real-world represented by the model changes. They call this process “self-healing,” meaning if the model drifts out of synch with the real-world it represents, algorithms will instruct the model to self-correct.
In the past, this required human intervention and was a historical reason why companies lost faith in earlier generation SCM systems. In other words, model drift led to plan quality drift which led to confidence drift which led to organizational drift around the system. AI systems, which are also model-based, are already encountering similar problems. Builders of AI systems can learn a lot from the thirty years of experience of those who built optimization systems.
Furthermore, historically, plans were deterministic, meaning they were based on a single possibility, versus a range of possibilities (stochastic). For example, a forecast, even though it is only 70% accurate and subject to wide fluctuations, was a single data point. A big reason for this is that information systems have been historically designed to only accept a single data point, and at the end of the day, you have to order a discrete number of something, not a range.
While this is true, today’s modern software solutions, such as that provided by Kinaxis, provide for an endless number of scenarios, effectively providing for a stochastic approach to planning. This provides the ability to consider an endless range of demand and supply scenarios, thus supporting not just creation of a robust plan, but enabling a robust process of planning, as advocated by Eisenhower.
In light of this, let me clarify the earlier reference to “plan to re-plan.” In today’s world, this doesn’t mean that you are throwing everything up in the air and creating a new plan every time you encounter a problem. What it means is that you are surgically repairing the plan as you react to conditions in the field of competitive battle.
Agility Enabling Adaptability
Kevin O’Leary (Mr. Wonderful of Shark Tank) asks his portfolio company CEOs “what would happen to Apple’s business if it had to close all its retail locations?” His answer: “absolutely nothing. People would still find a way to get their iPhones.” This is a stress test question meant to get his CEOs to figure out a way for their customers to still get their products, even as all their retail locations are closed.
David Simchi-Levy recommends that companies and governments perform regular stress tests on their supply chains, in an approach similar to that put in place by governments to evaluate the risk of large banks. This is basically what O’Leary is asking and seems like a very good idea, particularly since it takes away the need to identify and catalog events. (Yossi Sheffi suggests the same in his book The Resilient Enterprise from 2007). In other words, such a stress test would look at supply chains and say “what if we took away this source of supply or that source of supply,” without having to figure out what caused the loss of the source of supply a priori. Of course, it would also come in handy to know that an event has the potential to cause such a problem so that when that event surfaces you can get out ahead of the game.
To do this, companies can employ the same agility techniques they use in their tactical supply chain planning processes. They can simply add a risk evaluation process with a different cadence. The idea is not to try to imagine all possible risks, but to make the process of planning more robust.
Simchi-Levy’s idea can be applied to both strategic demand and supply risk. O’Leary’s question to his CEOs is an example of stress testing for demand risk. Similar questions that might be asked and formulated into scenarios, include:
What if you were to lose your key customer? What about your top 5?
What if your B2B channel completely shut down?
What if your retail channel completely shut down?
What if one of your key products was tampered with?
What if there was product contamination somewhere upstream?
What if a plant were to shut down for day, week, or month?
What if there were a port strike that lasted six months?
What if an earthquake were to shut down a west coast port (or 2) for six months?
And, in a question directly opposite to that asked by Kevin O’Leary, “what if your B2C channel were to shut down?” For a day? For a week? For a month? This is not so far-fetched given concerns regarding cybersecurity.
Therefore, the same capabilities that make your S&OP and other planning processes agile today, enabled by the likes of Kinaxis, can be used to consider these larger risks. These processes, if not already part of the organization’s planning center of excellence, can easily be added by extending agility capabilities. As pointed out earlier, agility then becomes the enabler of adaptability.
For leading companies (and hopefully governments), the current crisis is acting as a catalyst to make their planning processes more robust. Leaders will rapidly learn and determine which policy changes will be temporary and which will be permanent. But it is not just the actions they take that will be important for the long-term; what will be more important is improving the processes by which they take action.
This is called agility and adaptability. It’s not necessary to figure out all the unknowns; it’s important to have a robust, agile, planning process that considers consequences. Creating, updating, and exercising a catalog of potential consequences through scenario planning on a regular cadence will go a long way to this end. When a crisis does arise, you may not have previously thought of or exercised its exact consequences, but exercising the process of planning itself will have prepared you for it.