A hypothesis is a testable claim about the world. Not a wish, not a vision — a proposition that can be disproved by evidence. "If we cut price 10%, volume will rise at least 15%." "Users abandon the funnel at step 3 because of load time." "This market will reach $500M in five years." Each is falsifiable. Run the experiment, measure the outcome, and you either reject the hypothesis or fail to reject it. The power of the frame is that it forces clarity. Vague goals ("we want to grow") don't tell you what to measure. A hypothesis does. It also protects you from confirmation bias: the point is to test, not to prove. You look for evidence that would kill the hypothesis. If you don't find it, the hypothesis survives for now.
In strategy and product, the discipline is to state the hypothesis before you act. What do we believe? What would have to be true for this bet to pay off? What would disprove it? Teams that skip this step optimise in the dark. They ship features and hope. Hypothesis-driven teams ship to learn. The difference is whether the decision is framed as "let's try this" or "we believe X; we'll know we're wrong if Y." The second frame produces better experiments and faster learning. It also forces prioritisation: you can't test twenty hypotheses at once. You pick the one that would change the decision most if it were wrong.
Section 2
How to See It
Hypotheses show up wherever someone is making a testable claim and then checking it. The diagnostic: is there a clear "if-then" or "we believe X; evidence would be Y"? Is the claim specific enough that a result could contradict it? Vague statements ("customers want quality") are not hypotheses. "Customers will pay 20% more for a two-year warranty" is. Look for the moment when the team defines success and failure before they run the experiment.
Business
You're seeing Hypothesis when a growth team runs a pricing test. The hypothesis: "Moving the annual plan to the top of the pricing page will increase annual conversions by at least 25% without hurting monthly revenue." They define the metric, the change, and the decision rule. After two weeks they have a result. The hypothesis is rejected or not. The next test targets the next belief. The loop is hypothesis → experiment → update.
Technology
You're seeing Hypothesis when an engineering team debates two architectures. Instead of arguing in the abstract, they state: "We believe approach A will reduce p99 latency by 40% with acceptable maintainability." They build a spike or a shadow deployment, measure, and decide. The hypothesis made the debate empirical. The best technical decisions are those where the team explicitly states what they believe and how they'll know they're wrong.
Investing
You're seeing Hypothesis when a thesis memo says: "We believe this company can reach 30% gross margin at scale because of X and Y. We're wrong if unit economics don't improve by series B or if churn stays above 5%." The hypothesis is the investable claim. The "we're wrong if" is the falsification condition. When the data comes in, the investor updates. Without the hypothesis, the thesis is a story that never gets tested.
Markets
You're seeing Hypothesis when a macro trader has a view: "Rates will stay higher for longer because labour remains tight; we're wrong if next month's payroll print is below 100k." The position is sized on conviction; the exit is triggered when the hypothesis is falsified. Disciplined traders write down the hypothesis and the kill condition before they put on the trade.
Section 3
How to Use It
Decision filter
"Before a significant bet — product, pricing, strategy, or investment — state the hypothesis. What do we believe? What would disprove it? If you can't state it, you're not ready to act. If you don't test it, you're not learning."
As a founder
Make hypotheses explicit for every major initiative. "We believe that [X]. We'll know we're wrong if [Y]." Run the smallest test that could falsify it. The mistake is building for six months and then checking. The move is to identify the critical assumption, design a cheap test, and run it early. Prioritise hypotheses by impact: test the one that would change the strategy most if it were wrong. Kill projects when the core hypothesis is rejected. Don't reframe the hypothesis to fit the data; update the strategy to fit the result.
As an investor
Every investment is a hypothesis. Write it down: what has to be true for this to work? What would prove us wrong? Set checkpoints — at 12 months, at 24 months — and compare reality to the hypothesis. When the hypothesis is falsified, act: double down if the surprise is positive, cut or exit if it's negative. The worst outcome is a thesis that drifts into a story with no test. The best is a hypothesis that gets updated with data.
As a decision-maker
Use hypotheses to resolve disagreements. When two sides disagree, ask: what would we need to observe to know who's right? That's the hypothesis. Then design the test. The discipline prevents endless debate. It also prevents the HiPPO from winning by default — the hypothesis and the data get a vote. The decision-maker's job is to insist on a testable claim before committing resource.
Common misapplication: Treating the hypothesis as something to confirm. Confirmation bias is the default. The discipline is to seek disconfirming evidence. Design experiments that could kill the hypothesis. If you only look for support, you're not testing; you're rationalising.
Second misapplication: Hypotheses that aren't falsifiable. "We believe we'll succeed" is not a hypothesis. "We believe we'll hit $10M ARR by Q4 if we close these five enterprise deals" is. The test is: can you imagine a result that would make you say "we were wrong"? If not, sharpen the claim.
Feynman insisted on stating hypotheses clearly and then attacking them. "The first principle is that you must not fool yourself — and you are the easiest person to fool." His approach: write down what you think is true, derive what would have to follow, then look for experiments that could contradict it. The habit transfers to any domain: make the belief explicit, find the weak point, test it. Feynman's legacy for operators is the discipline of not fooling yourself — and the hypothesis is the tool.
Graham has pushed founders to "make something people want" and to learn from users. The YC playbook is hypothesis-driven: state what you believe about the market and the product, then get in front of users and see if it holds. The famous "do things that don't scale" is a way to test hypotheses fast — manual, direct contact gives signal before you build. The hypothesis frame turns feedback into tests: "We thought X; users said Y; we're updating."
Section 6
Visual Explanation
Hypothesis loop: State belief → Define falsification condition → Run test → Update or reject.
Section 7
Connected Models
Hypothesis sits at the centre of scientific method, experimentation, and decision-making. The models below either define how to test (scientific method, falsification), support the loop (A/B testing, critical assumptions), or warn against the opposite (confirmation bias).
Reinforces
Scientific Method
The scientific method is observe, hypothesise, predict, test, update. Hypothesis is the core unit: the testable claim that the method operates on. The reinforcement: the method gives the process; the hypothesis gives the content. You can't run the method without a hypothesis; you can't test a hypothesis without the method.
Reinforces
Principle of Falsification
Popper's criterion: a claim is scientific only if it can be falsified. Hypothesis is the vehicle for that. You state the claim in a form that evidence could contradict. The reinforcement: falsification is the standard; the hypothesis is the statement that meets the standard. A good hypothesis is one where you can specify what would prove it wrong.
Tension
Confirmation Bias
Confirmation bias is the tendency to seek and overweight evidence that supports your view. Hypothesis-driven testing is the antidote: you explicitly design for disconfirming evidence. The tension: left to themselves, people confirm; with the hypothesis frame, they're pushed to test. The discipline is to run experiments that could kill the hypothesis.
Tension
Critical Assumptions
Critical assumptions are the beliefs that have to hold for a strategy to work. The tension: you have a list of assumptions, but until you turn them into hypotheses and test them, they're just a list. Hypothesis is the action: take the most critical assumption, state it as a testable claim, and run the test. Assumptions without hypotheses stay untested.
Section 8
One Key Quote
"A theory which is not refutable by any conceivable event is non-scientific."
— Karl Popper, The Logic of Scientific Discovery (1934)
Popper's point: the mark of a scientific claim is that it can be wrong. The same applies to strategic and product hypotheses. A claim that no evidence could contradict is not a hypothesis; it's a belief. The practitioner's discipline is to state hypotheses that are refutable — and then to run the tests that could refute them.
Section 9
Analyst's Take
Faster Than Normal — Editorial View
State the hypothesis before you build. The teams that ship without a clear hypothesis are optimising in the dark. They'll get data, but they won't know what decision it informs. The move is to write down "We believe X. We're wrong if Y." before the sprint. Then build the smallest test. The hypothesis is the lens that turns data into decisions.
Test the assumption that would change the strategy. Not all hypotheses are equal. Some are nice to know; others would flip the plan. Prioritise the critical one. If you're wrong about it, everything else is secondary. Run that test first. Kill or pivot when it's falsified.
Seek disconfirmation. The default is to look for evidence that you're right. The discipline is to design experiments that could prove you wrong. If you can't imagine a result that would change your mind, your hypothesis isn't testable — or you're not really testing.
Update in public. When the hypothesis is rejected, say so. Update the strategy doc, the roadmap, the thesis memo. The cost of being wrong is lower when the hypothesis was explicit. The cost of hiding the result is that the organisation keeps acting on a dead hypothesis.
Section 10
Summary
A hypothesis is a testable, falsifiable claim. Use it by stating the belief and the falsification condition before you act, then running the smallest test. Prioritise hypotheses by impact; kill or pivot when the critical one is rejected. Avoid confirmation bias by designing for disconfirming evidence. Connected ideas include scientific method, principle of falsification, A/B testing, and critical assumptions.
The foundational text on falsifiability. Popper argues that the demarcation of science is that its claims can be refuted by evidence. Essential for understanding why hypotheses must be testable.
Ries popularised hypothesis-driven product development: build-measure-learn, with explicit hypotheses about what will happen. The business application of the scientific method to startups.
Kahneman on confirmation bias and how we seek evidence that supports our views. The hypothesis discipline is the antidote; the book explains why we need it.
Cagan on product discovery and testing assumptions with users. Aligns with hypothesis-driven development: state the belief, test with the smallest experiment.
Leads-to
A/B Testing
A/B testing is the operational form of hypothesis testing in product and growth. You have a hypothesis ("variant B will convert better"), you run the experiment, you get a result. The lead-to: once you think in hypotheses, A/B tests are the natural way to test them. The hypothesis comes first; the test design follows.
Leads-to
First Principles Thinking
First principles thinking breaks a problem into fundamental truths and rebuilds. Hypotheses can be built from first principles: "If X is true and Y is true, then Z should hold. We'll test Z." The lead-to: first principles give you the candidate hypotheses; testing gives you the evidence. The two work together — reason to generate, experiment to validate.
