Playing To Win
Strategy on Rugged Landscapes
The Importance of Human Tractability
I have written and spoken numerous times on why there is so little strategy out there — that it feels increasingly like a lost art. There are multiple reasons, but the subject for this Playing to Win/Practitioner Insights (PTW/PI) piece is the role of intractability. It is called: Strategy on Rugged Landscapes: The Importance of Human Tractability. All previous PTW/PI can be found here.
Smooth vs. Rugged Landscapes
There are highly contrasting terrains on which one can be put in a position to make decisions. Let’s imagine two such terrains.
In the first, you are plopped down randomly in the middle of Tanzania and given the task of finding your way to the tallest peak in the country. There is actually a pretty simple decision-making algorithm that you can use: walk uphill. Of course, that is a bit of exaggeration, but eventually you will get to Mount Kilimanjaro, which rises spectacularly and obviously from the vast plains of Tanzania. On a clear day, you can see it from over 100 miles away. Yes, you have to avoid getting confused with nearby Mount Meru, but it too rises distinctly from the plain and is obviously shorter.
In the second, you are plopped down randomly in the middle of Nepal and given the task of finding your way to the tallest peak. Nepal is only one sixth the size of Tanzania — so that should be easier, right? But if you employ the same rule, you will have almost zero chance of seeing, let alone getting to the top of Mount Everest. Somehow, you would have to have to find a path from one of the many deep, jagged valleys running between nine of the highest ten mountains in the world, all within 5000 feet of the same height. Chances are, you would exhaust all of your energy climbing one of the other eight and only when you get to the top, see that there is a giant valley between you and Everest in the distance.
That is the difference between a smooth and a rugged landscape. In a smooth landscape, there is a singular peak and a simple rule for getting there. The more rugged, the more peaks — and no simple rule.
The central determinant of the ruggedness of a landscape is interdependence of variables in the system that produces the landscape. A landscape gets complex very quickly as variables interact. A landscape can have one million possible peaks (i.e. distinct outcomes) when there are only twenty interdependent variables. Such highly interdependent landscapes are what complexity theorists and evolutionary biologists call complex adaptive systems — for example, the Amazon jungle. Heat, moisture, ground cover, existing tree stock, human involvement, and many more variables interact to create a single unpredictable outcome at a given point in time — but next week it can be different because the variables interact with each other continuously to create varying outcomes. The ruggedness is even more vexing because of adaptation. Unlike a mountain range, the landscape of the Amazon jungle keeps shifting.
Complexity theorists and computer scientists geek out about dealing with these kinds of rugged landscapes. A problem is classified as P (polynomial time) if an algorithm is able to generate the solution (i.e., the highest peak) to it in a reasonable amount of time. If the validity of the solution can be verified (again in a reasonable amount of time) it is NP (nondeterministic polynomial time). (And they have argued for years whether the proposition P=NP is correct — or not, with a $1 million prize waiting for anyone who solves that question conclusively.)
If instead, it is an NP-hard problem, there is currently no known algorithmic solution to it. That is, the problem is algorithmically intractable. Theorists may figure out how to make some current NP-hard problems tractable — i.e. resolvable algorithmically.
There are many arguments about NP-hard problems. For example, a favorite subject of argumentation is the problem of optimizing the route of a traveling salesman. Depending how it is framed, it is or is not NP-hard. However, there is general agreement the ‘halting problem’ is irresolvably NP-hard.
There is even a more elusive form of problem — an undecidable one. These are problems that algorithms can never solve because they feature characteristics such as subjective variables, circumstances that vary faster than any algorithm can run, and/or the necessary data is simply unavailable.
Even though theorists place intractable and undecidable in different categories, I am simplifying them by referring to both as intractable, by which I mean that humans do not have access to an algorithmic approach — a formula — for reaching a solution in which they can be confident is optimal in the circumstance.
The stakes can be very high for humans with respect to such problems, as emphasized by evolutionary biologists who study landscapes and the fitness of species for the nature of the landscape on which they exist. To survive, a species needs to find the proverbial high ground of a safe peak on the ‘fitness landscape.’ It doesn’t have to be the tallest peak — i.e. a set of characteristics and behaviors that makes the species perfect for the landscape. But the fitness has to be pretty strong. You want to be a giraffe — which has been around for 12 million years, not a dinosaur — all of which are long gone. Evolution is a slow boat to adaptation to your landscape. In human decision-making circumstances, you can change choices faster. But you can only change your choices for the better (i.e. reach higher ground) if the problem is tractable to you.
Implications for Strategy
The context for strategy is a rugged landscape. There are clearly many interdependent variables in strategy. The actors are numerous — the company, competitors, customers, suppliers, channels, regulators. There is plenty of heterogeneity within each of those categories. Each player exhibits free will. The actions of each influence the outcomes for all. They don’t set their actions and then stop. Instead, they keep acting. The context is unequivocally rugged — i.e. there are many, many competing peaks.
The fundamental strategy question is what is the strongest competitive position — metaphorically, the tallest peak — you can occupy? As with most highly rugged landscapes, that is an intractable question — if you insist on an algorithmic solution (that is, you can reason algorithmically to a correct and verifiable solution). Managers in modern business are overwhelmingly from the educational fields of business, engineering and/or economics, and there they are taught that to be a noble, effective manager, you must calculate your way to any decision you make. Anything else is unbecoming.
This puts modern managers in a bind with respect to strategy. They have been taught a methodology — even a way of being — that is only good for smooth terrains, but they need to make the most important decisions for their companies on rugged terrains. The bind is pretty much as straightforward as that.
Because of the algorithmic intractability of strategy, managers tend to default to one (or more) of four modes — none of which is conducive to high-quality strategy:
1) Don’t think about strategy and hope that things will work out well anyway. Unfortunately, as my friend AG Lafley always says: Hope is not a strategy. This mode tends to produce a random walk of decisions that rarely add up to anything good.
2) Follow the pack. That is, do whatever the rest of your identifiable peers are doing. Sadly, this mode tends to overpopulate one domain of the fitness landscape and wreck that domain for everybody — like all the wildebeest choosing the same drinking pond. This mode is what commoditizes otherwise perfectly fine industries.
3) Split the challenge into tractable pieces and do micro-sensible things. This is planning masquerading as strategy. Each micro-decision (e.g. ‘initiative’) makes sense on its own but doesn’t add up to a productive integrated strategy.
4) Analysis Paralysis. Managers hold to the hope that with enough dedication to analysis, they will come up with an algorithmically-driven strategy solution — and they just keep analyzing hoping that the answer is just over the horizon. But they never get there.
Human Strategy Tractability
The key to avoiding these unproductive modes is to focus on human tractability in strategy. That is, one must take the algorithmically intractable task of setting strategy on a rugged landscape and make the task tractable for human managers. That means providing them with a heuristic for taking on the tough task of strategy that builds their confidence in tackling it. Otherwise, they will default into one (or more) of the four unproductive modes.
In doing so, I focus on the following three heuristic features/tools:
1) Imagining a happy end state. Determining strategy on a rugged landscape is a very difficult task. When people are attempting a difficult task, to maintain their confidence and enthusiasm they need to know what they are fighting for. So, I get teams to focus on happy end states — problems that go away, for example — rather than on the enormity of the challenge. In fact, I minimize the enormity of the challenge.
2) Hold a competition of possibilities. It is too hard of a question to ask: what is the single best strategy choice we could make? I.e., what is the single highest peak? It is much easier to have a competition of possibilities where each is simply plausibly the highest peak but subject to competition from the others. Taking the bar down from the single highest peak to a plausible one tends to give managers the confidence to suggest possibilities — and on this front, more is better.
3) Focus on the logic of the possibilities, not the data. Data is a tractability trap. Since there are no algorithmic solutions to strategy decisions on rugged landscapes, focusing on data is a vain attempt to populate non-functional algorithms with data. Not to mention, there is no data about the future, while strategy choice is all about the future. I focus teams instead on the logic of strategy possibilities — what would have to be true (WWHTBT) for a possibility to be the most attractive one?
In my experience, these three heuristic features help human teams feel the challenge of strategy on a rugged landscape is tractable to them — and gives them a path forward.
Practitioner Insights
A good rule in life is it depends. Aristotle said that about the utility of data analytics 2500 years ago. Sometimes data analytics is useful; sometimes it is definitely not. Sometimes landscapes are relatively smooth — and algorithmic answers are great. Sometimes they are very rugged — and algorithmic answers are fundamentally unhelpful, if not dangerous.
When it comes to business strategy decisions, the context is almost always a rugged landscape — for which there is no algorithmic solution, despite the world being massively hyped about artificial intelligence (AI) taking over strategy.
Don’t let anyone try to convince you that you must use data analysis and algorithmic approaches to determine your strategy. That isn’t an absolute rule — despite what you were probably taught.
To be a useful strategist, you need to develop a heuristic for tackling strategy. If you do develop your own unique heuristic, you will never need to fear being replaced by AI. If you don’t, AI will replace you in strategy — though I would argue it will produce a steady diet of blandness and hallucinations.