Bold Conjectures and Severe Tests
Generate daring hypotheses then subject them to the harshest tests you can devise
Bold Conjectures and Severe Tests describes Popper methodology for scientific progress, which inverts the common assumption that knowledge grows through careful accumulation of observations. Instead, Popper argues that knowledge advances through a two-step process: first, generate bold conjectures—hypotheses that go far beyond available evidence, make surprising predictions, and risk being spectacularly wrong. Second, subject these conjectures to the most severe tests possible—experiments designed to maximize the probability of falsification. Theories that survive severe testing earn greater credibility precisely because they risked more and survived. Timid hypotheses that predict only what is already known, or tests designed to confirm rather than challenge, produce no new knowledge regardless of their outcome. Applied beyond science, this framework encourages intellectual courage in strategy and innovation: propose bold theories about your market, customers, or product, then design the toughest possible tests before committing resources. The combination of boldness in hypothesis and rigor in testing produces the fastest possible rate of learning, whether the hypothesis is confirmed or refuted.
- Bold conjectures that risk being spectacularly wrong generate more knowledge than timid hypotheses
- Severe tests that maximize falsification probability are more informative than gentle confirmatory tests
- A theory that survives a severe test gains more credibility than one that passes an easy test
- Knowledge grows through elimination of error, not through accumulation of confirmation
- Intellectual courage—willingness to be wrong—is a prerequisite for genuine progress
- Generate Bold ConjecturesPropose hypotheses that go significantly beyond available evidence and make surprising, specific predictions. A bold conjecture is one that most people in your field would consider unlikely, that makes predictions you did not expect yourself, or that challenges conventional wisdom in a testable way. In business, this might be our highest-churn customers are actually our most loyal long-term—they churn and return repeatedly. In product development, it might be removing our most popular feature will increase overall engagement. The bolder the conjecture, the more you learn whether it succeeds or fails.Pro tipTrain yourself to generate at least one bold conjecture per week that challenges your current assumptions—even if most are wrong, the habit of boldness accelerates learningWarningBoldness without testability is just recklessness—every bold conjecture must come with a clear falsification criterion
- Design the Most Severe Test PossibleFor each bold conjecture, design a test that maximizes the probability of falsification. A severe test is one where the hypothesis would clearly fail if it were wrong—not a test designed to produce ambiguous results that can be interpreted either way. If your conjecture predicts that removing a feature will increase engagement, run an A/B test with sufficient sample size and a predefined significance threshold. If your conjecture predicts that a new market segment exists, design a minimum viable test that would clearly reveal demand or its absence.Pro tipAsk yourself what result would force me to abandon this hypothesis—then design the test to produce exactly that kind of decisive evidenceWarningTests designed to produce ambiguous results are worse than useless—they consume resources while generating no real learning
- Learn from Both Confirmation and RefutationWhen a bold conjecture survives a severe test, it earns significant credibility—much more than a timid hypothesis passing an easy test. Celebrate this and investigate why the conjecture held despite its boldness. When a bold conjecture is refuted by a severe test, celebrate equally—you have eliminated a plausible but incorrect hypothesis and narrowed the space of possibilities. Both outcomes represent genuine learning that advances understanding. The only bad outcome is an ambiguous result from a poorly designed test that provides no clear signal.Pro tipCreate a team culture where refuted hypotheses are valued as learning events—the fastest learning organizations are those that refute the most hypotheses per unit of time
Einstein general relativity made the bold prediction that light from distant stars would bend by precisely 1.75 arc seconds when passing near the sun—nearly double the prediction of Newtonian mechanics. Arthur Eddington organized an expedition to observe a total solar eclipse in 1919, designing a severe test that could clearly distinguish between Einstein prediction and Newton. The observation confirmed Einstein prediction with remarkable precision. The significance was not merely that Einstein was right but that his theory had survived a test that could have destroyed it. Had the measurement shown Newtonian bending, general relativity would have been falsified.
Popper developed this methodology as an alternative to inductivism—the view that science proceeds by accumulating observations and deriving generalizations from them. Popper argued that this was both logically flawed (you cannot derive universal laws from finite observations) and historically inaccurate (the greatest scientific advances came from bold theoretical leaps, not careful data accumulation). Darwin did not accumulate enough observations to deduce evolution—he proposed a bold theory that made predictions far beyond available evidence. Einstein did not inductively generalize from observations—he imagined riding a beam of light and derived revolutionary predictions. Popper formalized this pattern into a methodology: conjecture boldly, test severely, and let the results drive progress regardless of whether they confirm or refute your conjecture.