Strategy as Hypothesis
Treat strategy as a testable hypothesis and its execution as an experiment
Rumelt argues that good strategy rests on functional knowledge about what works, what does not, and why. While generally available knowledge is essential, it is available to all competitors and rarely decisive. The most precious knowledge is proprietary, available only to your organization. A new strategy is, in the language of science, a hypothesis, and its implementation is an experiment that generates proprietary knowledge.
This framework draws on the scientific method: anomaly detection (noticing things that do not fit existing theories), hypothesis formation (creating explanations for what is observed), experimentation (testing the hypothesis through action), and adaptation (adjusting the strategy based on results). Just as scientists build knowledge by testing hypotheses against reality, organizations build proprietary strategic knowledge by treating each strategic initiative as an experiment.
The key insight is that strategy and science share the same fundamental epistemology: both involve forming educated guesses about complex situations and testing them against reality. The strategist who understands this can design strategies that generate valuable learning regardless of whether they succeed or fail in their immediate objectives.
- A new strategy is a hypothesis; its implementation is an experiment that generates knowledge
- Generally available knowledge is necessary but not sufficient for competitive advantage; proprietary knowledge is decisive
- Anomalies, things that do not fit the expected pattern, are the most valuable sources of strategic insight
- Good strategy generates proprietary knowledge regardless of whether its immediate objectives succeed or fail
- Scientific empiricism, actively exploring your chosen arena through experimentation, builds the knowledge base for future strategies
- Seek AnomaliesActively look for things that do not fit the prevailing theory or expectations about your industry. Anomalies are the raw material of insight. Why is this customer buying our product for a reason we did not expect? Why did this competitor's move produce unexpected results? Why is this market segment behaving differently from what the model predicts?Pro tipHoward Schultz noticed an anomaly when visiting Milan: Italians treated espresso bars as community gathering places, not just caffeine delivery. This anomaly, that coffee could be an experience rather than just a commodity, was invisible to everyone in the American coffee industry because it contradicted the prevailing theory.WarningMost people dismiss anomalies as noise, errors, or irrelevant exceptions. The discipline of treating anomalies as signals rather than noise is the foundation of strategic insight.
- Form a HypothesisBased on the anomaly, form a clear hypothesis about why the anomaly exists and what it implies for strategy. The hypothesis should be specific enough to be testable and should suggest clear actions. Frame it as: 'If we do X, we expect Y to happen because Z.'Pro tipSchultz's hypothesis was that Americans would value the Italian espresso bar experience if it was presented in the right way. This was specific enough to test through a pilot store and clear enough to guide decisions about store design, product offerings, and pricing.
- Test Through ActionImplement the strategy as an experiment. Design the implementation so that it generates maximum learning regardless of whether it succeeds or fails. Define in advance what you expect to observe if the hypothesis is correct and what you expect to observe if it is wrong.Pro tipSchultz tested his hypothesis by opening Il Giornale, a pilot espresso bar in Seattle. He observed what worked (the espresso bar atmosphere), what needed adjustment (the constant opera music was irritating, menu items needed Americanization), and used this learning to refine the concept before scaling.WarningThe biggest mistake in experimentation is not learning from the results. Many organizations implement strategies, observe the results, and then explain away any disconfirming evidence rather than updating their hypothesis.
- Capture Proprietary KnowledgeActively document and internalize what you learn from the experiment. This proprietary knowledge becomes a strategic asset that competitors cannot easily acquire. It accumulates through repeated cycles of hypothesis, experiment, and learning.Pro tipStarbucks' decades of operational learning about store location, customer preferences, barista training, and supply chain management created a body of proprietary knowledge that no competitor could replicate simply by copying the visible elements of the business model.
Schultz noticed the anomaly of Italian espresso bar culture during a 1983 trip to Milan. He formed the hypothesis that Americans would value a similar community coffee experience. He tested this through a pilot espresso bar called Il Giornale. He learned that the Italian concept needed significant adaptation for American tastes (less opera, more comfortable seating, different food). He used this proprietary knowledge to build Starbucks into a global brand.
Rumelt worked with Hughes Electronics engineers who approached strategy as they would an engineering problem: with hypotheses, experiments, and data. They recognized that their satellite communication strategy was a hypothesis about future demand, technology evolution, and competitor behavior. By framing it this way, they designed their strategy to generate learning regardless of whether specific bets paid off.
Rumelt developed this framework by drawing parallels between scientific methodology and strategic practice, inspired by his work with Hughes Electronics engineers who had risen to management positions. These engineers understood hypothesis testing in their technical work but did not apply the same discipline to strategic thinking. The Starbucks example showed how Howard Schultz formed a hypothesis about the European espresso bar experience, tested it in Seattle, and built proprietary knowledge that could not be easily replicated by competitors who had not undergone the same learning process.