The Emergent Approach to the OKC Thunder's Strategy

I read Peter Compo’s The Emergent Approach to Strategy (referred to as EAS from here on) because I was frustrated by the lack of structure in Richard Rumelt’s Strategy Kernel from Good Strategy Bad Strategy (GS/BS going forward). Don’t get me wrong… I really liked GS/BS. But the methodology is so flexible that I struggled to organize its concepts when I applied it in the real world. Particularly:

1. The lines between Diagnosis, Policy and Action often felt blurry.
2. Rumelt’s most compelling cases compose multiple policies, but his Kernel is very clear about having a single Guiding Policy. It wasn’t clear to me how to reconcile that contradiction.

To put it simply, the structure from EAS addresses my frustrations quite nicely. To prove that, I’ll rework my attempt to describe the Oklahoma City Thunder’s strategy across the last decade.

EAS vs. Rumelt

First, let’s look at a few ways that EAS differs from Rumelt’s approach. ¹

• Triads: Rumelt’s Strategy Kernel is a triad of diagnosis / guiding policy / coherent actions. EAS has its own triad of aspiration / bottleneck / strategy.
• Strategy: Rumelt says a strategy must have all three pieces of his Kernel, while Compo says that the strategy is nothing more than “the central rule of a framework, designed to unify all decisions and actions around busting the bottleneck to achieving aspirations.”
• Kernel vs. Framework. Compo’s “strategy framework” is analogous to Rumelt’s Kernel, but it has more categories: values, aspiration, diagnosis, rules, plans, metrics. There are sub-categories in each, but I’ll omit them for brevity.
• Nested Frameworks: Rumelt’s theory says that a strategy must have a single Guiding Policy, but, as stated above, this conflicts with many of his cases which have multiple overlapping or reinforcing policies. EAS organizes these clearly: subsystems should have their own strategy frameworks that have the parent or peer frameworks as external constraints within their own diagnosis.
• Option generation and fitness. The “emergent” part of EAS is that you should generate lots of options. Part of the diagnosis should include modeling multiple possible future scenarios, and EAS’s template requires you to model multiple strategy options. From there, you or your team should list criteria which would validate or invalidate each option, then analyze and debate until one option emerges as the winner which you then move to implement. Rumelt’s cases mention similar concepts, but there’s no specific methodology or structure. This is a point of EAS that is unique compared to GS/BS.
• Adaptation. Rumelt discusses how Starbucks iterated until they found a winning strategy, but there’s not much specific advice on how to replicate that. This is common in GS/BS: much of the knowledge is embedded in cases to be used as heuristics, rather than embedded within core theory or its methodology. EAS is clearer here: you should change the strategy as soon as you see that it’s not fit. This might happen in 1 week or 1 year. Timing doesn’t matter. In fact, EAS states that execution is “doing the work needed to adhere to your framework and fixing the framework it when it is no longer fit.”

EAS Applied

Applying EAS by-the-book would involve creating a strategy alternative matrix (SAM) with multiple potential strategies plus criteria for evaluating them. Only one strategy would actually be implemented.

I’m not going build a SAM because that level of detail isn’t necessary to illustrate the utility of the framework for our OKC case. Luckily for me, EAS has an escape hatch early on in the book saying you should implement as much or as little of the methodology as necessary for your purpose, and I shall take that to heart.

The SAM is supposed to be a living, evolving document. To approximate what would, in practice, be an evolving series of documents, I’ll split the strategy framework into static and dynamic parts.

(Mostly) Static Parts of the Framework

The Thunder’s behavior is consistent across years in many ways. This is a good sign! It indicates that they have a clear perspective on winning and the discipline to stick with it.

• Values:
  • We’re not in the business of predictions, we’re in the business of creating advantages through careful observation and evaluation²
• Aspiration: to build a team core capable of winning an NBA championship across a multi-year window.
• Diagnosis
  • Propositions:
    • Championship caliber teams have at least 1 top 10 player, 2-3 all-NBA players, complementary role players, and at least 2-4 high value contracts³
    • Investor mentality, analytics and scouting departments, willingness to experiment, humility⁴
  • External Constraints: NBA salary cap, player career trajectories that may not line up, small market budget
  • Scenarios (omitted for brevity)
  • Bottlenecks (dynamic, listed below)
• Rules (dynamic, listed below)
• Plans & Projections (omitted for brevity)
• Metrics (omitted for brevity)

Dynamic Parts of the Framework

This is where EAS gets interesting. Its adaptive Aspiration / Bottleneck / Strategy triad neatly describes the Thunder’s behavior. The strategy changes as their evaluation of the core bottleneck changes.

Year	Bottleneck	Strategy
2017	Lack of talent around MVP candidate Russell Westbrook	Trade future assets to acquire All-NBA level talent
2019	Team assets (current players + expendable future capital) insufficient to reach title contention	Reset the team core by acquiring draft capital by every means available
2021	No players have yet emerged as All-NBA level talents	Patience with experimentation to surface/advance top-end talent
2024	Too many high-upside, soon-to-be-expensive players who need the ball and have overlapping skills and weaknesses	Shift to acquiring role players on contracts aligned to the young core’s paydays, using excess cap space and trading the least-valuable young players
2025	Young core is very expensive	Cash in optionality for stability by signing core players to extensions as soon as possible (bonus tactic5: secure ownership approval to pay luxury tax based on projected playoff revenue6)

Subsystems

I really like EAS’s approach to nesting subsystems. I won’t go into the same level of detail with these, but it’s worth briefly considering them.

• Analytics: One can imagine the analytics department’s strategy being dependent on the overarching strategy. And not in a way that just restates the parent strategy (this is a heuristic from EAS). Perhaps that strategy of the analytics department is to start evaluating the next potential strategies. I.e. perhaps this department was analyzing role players during the “patience” phase, in anticipation of at least a few stars coming out of the Thunder’s development pipeline.
• Scouting / Drafting: The Thunder seem to reliably draft either: 1) high risk / high upside players or 2) solid wing players capable of fitting an athletic “3 and D” archetype. In other words, the shifting high level strategy, while always being an external constraint, does not have to change the lower system’s strategy.
• Player Development: While I don’t think I have enough information to reliably evaluate the Thunder’s internal strategy for this department, I can make some assumptions that they adapt their approach based on what the overarching strategy is. For example, perhaps they would have a strategy like “throw our highest upside rookie players into the NBA fire” during the “reset” phase of the overall strategy vs. a “leave our highest upside rookie players in the G League until [some criteria] is met” during the later phases.

The point is that EAS provides a real methodology to compose these pieces of their organization together in a coherent way. The result of this for OKC is the mutually reinforcing fit of activities, such as Michael Porter describes. As the analytics department gets better, so do the scouting and player development departments. The reverse is true as well because these departments are generating more data to analyze.

That said, nothing in EAS actually specifies how to ensure these things mutually reinforce. In practice, it can’t be so simple as filling in the external constraints field in a strategy document. Perhaps EAS would say this result is completely emergent; however, I suspect that there’s additional technique required. The minutiae of coordinating these departments day-to-day is likely part of the executive skill that sets Sam Presti apart from other GMs.

Comparison to the Rumelt Kernel

So, looking back, how does this compare to my first attempt to analyze the Thunder’s strategy? First of all, there’s quite a bit of overlap. That’s to be expected. After all, I’m really just trying to organize the same basic soup of publicly available information.

The major difference is that the Guiding Policy is quite generic in the Rumelt kernel. It’s closer to a “Value” in EAS, and doesn’t inform clear tradeoffs. Modeling a single short Strategy that evolves across time is much more descriptive, with clearer tradeoffs.⁷

I also really like the clarity that the “Bottleneck” provides the Diagnosis.

Finally, the “Coherent Actions” are quite different. In the Kernel, I struggled with whether the Coherent Action was “trade away core players” (which reads more like a policy) vs. “trade Paul George and Russel Westbrook” (which is unambiguously an action). With EAS, the policy-like-actions are clearly Strategies. EAS also discusses tactics, which are still rules, just with a smaller and less unifying scope. Things like “use our cap space to take on bad contracts in exchange for extra draft picks” fit there.

Overall, I feel EAS is more useful in describing this case vs. the Rumelt Kernel.

Heuristics All the Way Down

So does this mean that EAS “wins”? Is it the One True Strategy Framework? I don’t think so at all. The application of strategy is highly context dependent. Rumelt’s approach is to create a very flexible framework with a library of cases. Compo adds a lot more structure to his framework. There is a surely a tradeoff there. Fully building out EAS’s SAM is a lot more work. The concepts are quite overwhelming. And while I don’t have a clear example yet, I suspect there are situations for which the structure doesn’t work at all⁸.

Because of how complex strategy is, both frameworks are largely built on heuristics, rather than theoretical proofs. My working theory is that the proper way to approach such a domain is read widely, both theory and cases, so that you can pick and choose from different toolkits as situations present themselves.

Compo and Rumelt don’t disagree much in their theories; it’s the approaches to application that are different. I suspect that the more that I read on the discipline of strategy, the closer I’ll approach the asymptote of what we humans know about it. As such, I’ll be curious to see how much my thinking evolves when I read Playing to Win from Roger Martin. If I find it refining edges in my strategic thinking, rather than wholesale re-categorization, then I’ll start assuming that strategy practice will become a more valuable teacher than learning strategy theory.