Entrepreneurs are taught to believe. Believe in the product, even when early users shrug. Believe in the market, even when the numbers are thin. Believe hardest of all in the moments when belief is least convenient, in the investor meeting that goes quiet, in the launch that underperforms, in the cofounder who starts to doubt. Conviction is the trait investors say they admire most in a founder. It is also the trait most likely to keep a founder married to an idea long after reality has filed for divorce.

There is another way to build a company, and it does not require less ambition, only a different relationship to belief. It means treating what you think you know about your business, who the customer is, what they will pay, why they will choose you, as a hypothesis rather than a settled fact. The question stops being whether the idea feels compelling. It becomes whether reality can prove it wrong.

Research spanning more than a thousand start-ups across Europe and the United Kingdom, produced by a group of scholars at Bocconi University, INSEAD, and Bayes Business School, suggests that this shift in mindset carries real consequences for how founders allocate their limited time, when they abandon weak ideas, and how much revenue they ultimately generate.

The value of thinking like a scientist, it turns out, is not that it helps founders predict the future more accurately. It is that it helps them discover, quickly and cheaply, when their current picture of the future is wrong.

What the Research on Scientific Decision-Making Actually Found

The best-known study in this body of work, led by Alfonso Gambardella, a professor of corporate management at Bocconi, recruited 759 start-ups across Milan, Turin, and London through social media, newsletters, and entrepreneurship events, and the results were published in the Strategic Management Journal with a companion write-up in the Harvard Business Review. Every founder in the study received training in standard strategic frameworks and evidence-gathering techniques such as interviews, surveys, and A/B testing.

The difference was that one group, the treatment group, was also taught to apply the scientific method on top of those tools: state a specific, falsifiable hypothesis, design a test capable of proving it wrong, and let the result decide what happens next. The other group, the control, was free to use the same tools however they liked.

One of the ventures the researchers tracked, an electric-moped sharing service called Mimoto, illustrates the mechanism cleanly. Its founders believed college students rushing between classes would be their core riders. They placed dozens of mopeds near an urban campus to test that belief and found something else entirely: usage was spread evenly across age groups, concentrated instead among people with unpredictable commuting patterns.

Rather than defend the original theory, the founders returned to it, formed a new hypothesis around young professionals, particularly lawyers shuttling between client meetings, and tested that instead. The pivot was not a guess. It was a conclusion the data had already reached.

Gambardella has described the origin of the research in blunt terms: many entrepreneurs were not very good at making predictions about their business, and that hurt them when it mattered most. Across the full sample, start-ups using the scientific method generated more revenue than the control group and were meaningfully more likely to walk away from ideas that were not working, a step most early-stage companies delay far longer than they should.

Among the top 25 percent of revenue-generating start-ups in the study, those trained in the scientific method earned an average of 28,000 euros more than their control-group counterparts over the course of the experiment. Among the top 5 percent, the gap widened to 492,000 euros. These are associations observed within a controlled field experiment, not guarantees, and the effect was strongest precisely among the ventures that were already performing best, a pattern worth sitting with rather than skipping past.

Act III: Osense, or How a Failed Idea Became a Better Business

No company in the research illustrates the process more vividly than Osense, a European sustainability start-up whose founders enrolled in the program with what felt like a fully formed idea. Their first concept was a peer-to-peer rental platform, built on the theory that people would rather rent goods than own them, and that reducing consumption this way would shrink the collective carbon footprint. To test it, they planned twenty-five field interviews asking whether people would actually reuse each other's belongings.

They needed only five conversations to see the idea was not going to work.

Left to their own instincts, the founders later admitted, they probably would have kept building anyway, spending months and real money on a prototype nobody wanted. Instead they returned to the one thread of positive feedback from those interviews, that people cared about sustainability, and formed a second hypothesis: a platform connecting rental car companies with customers interested in electric vehicles.

They tested that too, speaking with rental companies and their customers, hoping for the sixty percent approval that would justify moving forward. They got closer to twenty.

A second failed hypothesis would break most founders' resolve. Osense's did not, because the scientific method had already reframed failure as information rather than defeat. Shortly afterward, one of the cofounders came across a McKinsey report identifying Scope 3 emissions, the indirect emissions companies generate through their supply chains and employee travel, as the biggest unsolved problem for sustainability-minded corporations.

They asked a new question: could a tool give companies real-time visibility into that data? Three interviews with sustainability managers were enough to sense they were onto something. By the tenth, nine had been overwhelmingly positive. A six-month pilot followed, and the pilot partner became their first paying client.

Strip away the sustainability framing and what remains is the scientific method in its purest form: a hypothesis, a test, a rejection, a new hypothesis, a test, a rejection, and finally a discovery grounded in evidence rather than attachment. Osense did not succeed because its founders had a better idea than most entrepreneurs. They succeeded because they were willing to be wrong twice in a conference room instead of once, expensively, in the market.

Why Running Experiments Is Not the Same as Thinking Scientifically

It is tempting to conclude that any founder running tests is already doing this. They are not. A founder can run fifty A/B tests and still make decisions no more scientifically than one who runs none, if those tests lack a clear hypothesis, change several variables at once, or get reinterpreted after the fact to mean whatever the founder needs them to mean. Genuine scientific thinking requires a form of discipline most start-ups skip entirely: deciding, before the test runs, exactly what result would prove the idea wrong.

Call it pre-commitment. Before launching an experiment, a rigorous founder writes down what they believe, what evidence would support it, what evidence would contradict it, and what threshold separates success from failure. Without that step, it is almost impossible not to see confirmation in ambiguous results, since the mind is remarkably skilled at deciding, after the fact, that the data meant what it was hoping to find.

This is also where the study's second mechanism, what the researchers call methodic doubt, becomes important to define correctly. Methodic doubt is not the same as pessimism. A pessimistic founder expects the idea to fail before any evidence exists. A scientifically minded founder simply refuses to assume it will succeed, and treats that absence of assumption as a discipline rather than a mood.

The first begins with a conclusion already drawn. The second begins with a genuine question. Paired with what the researchers term efficient search, the ability to concentrate limited time on the ideas most likely to survive contact with a real customer rather than spreading it evenly across every possibility, methodic doubt is the habit that let Osense's founders notice a fatal flaw after five interviews instead of five years.

The first begins with a conclusion already drawn. The second begins with a genuine question. Paired with what the researchers term efficient search, the ability to concentrate limited time on the ideas most likely to survive contact with a real customer rather than spreading it evenly across every possibility, methodic doubt is the habit that let Osense's founders notice a fatal flaw after five interviews instead of five years.

Why the Best Founders Do Not Just Pivot, They Pivot Deliberately

One of the more counterintuitive findings in the research is that scientifically minded founders were not the ones pivoting most often. They tended to make one or two major strategic shifts over the course of the study, no more. Founders who never pivoted and founders who pivoted constantly both underperformed the group that pivoted deliberately, on the strength of specific evidence, and then held their new course.

The lesson is not that pivoting is virtuous. It is that evidence-driven pivots are focused, while instinct-driven ones tend to be either stubbornly absent or erratic.

This distinction is worth naming plainly, because founders often collapse two very different behaviors into one word. Persistence means continuing to pursue the underlying mission even as the specific approach changes. Stubbornness means continuing to defend the original hypothesis regardless of what the evidence says. Osense was persistent about building a sustainability company and, at the same time, entirely unstubborn about which company that would turn out to be.

It helps to think of the underlying process as a ladder, one that starts vague and ends specific enough to test. A belief such as customers want this becomes a hypothesis such as mid-market logistics managers will pay a recurring fee for real-time freight visibility. The hypothesis produces a prediction, a defined threshold against a defined sample, perhaps that at least six of twenty qualified prospects will agree to a paid pilot. The prediction is checked through an experiment, ideally one offered before the full product exists.

The evidence that comes back then forces one of three honest decisions: continue, modify, or abandon. Each rung matters, but the ladder only works if a founder climbs down it in order rather than skipping straight from belief to conviction.

The Limits of Scientific Thinking

None of this should harden into a new orthodoxy, and a more recent study by two of the same researchers, Chiara Spina and Elena Novelli, offers an important corrective. Their field experiment with 261 UK start-ups at varying stages of maturity, also published in the Strategic Management Journal and featured in the Harvard Business Review, found that scientific decision-making does not help every start-up equally, or in the same way.

More established ventures used the scientific method to optimize what was already working, refining a product listing or an advertising approach, and saw measurable gains almost immediately; one founder in the study raised profits by ten percent simply by testing different ways of describing the same product. Early-stage founders, by contrast, tended to point the same rigor at their most fundamental assumptions, and that kind of questioning came at a short-term cost.

Over the nine-month study window, early-stage ventures that adopted scientific decision-making saw revenue decline as they stepped back to re-examine their entire business model. Several founders described the dip as a deliberate trade, sacrificing near-term growth to avoid building confidently on top of a flawed foundation.

The researchers are careful not to frame this as a strike against the scientific method itself. Their conclusion is closer to a caution about timing: a mature start-up should use scientific rigor to sharpen what it already knows works, while an early-stage venture should expect that genuine, honest testing of its core assumptions may look like a step backward before it becomes a step forward.

A founder, or an investor evaluating one, who mistakes that exploratory dip for failure risks abandoning the exact process most likely to prevent a much larger failure later.

The Evidence Decides What Happens Next

The founders most likely to build the next big thing may not be the ones best at predicting the future. The research suggests they may simply be the ones best at discovering, faster and more cheaply than their peers, when their prediction was wrong. That still requires conviction. It is just not the kind that refuses to bend. It is the conviction to keep testing, to keep searching, and to let the evidence, not the pitch, decide what the company becomes.

The idea is only ever the beginning. What a founder does with the first piece of evidence that contradicts it is usually what determines everything after.

Primary Sources

Camuffo, A., Gambardella, A., Messinese, D., Novelli, E., Paolucci, E., and Spina, C., "A Scientific Approach to Entrepreneurial Decision Making: Large-Scale Replication and Extension," Strategic Management Journal, 2024 — https://onlinelibrary.wiley.com/doi/10.1002/smj.3580

Gambardella, A., "Why Entrepreneurs Should Think Like Scientists," Harvard Business Review, July 2024 — https://hbr.org/2024/07/why-entrepreneurs-should-think-like-scientists

Novelli, E. and Spina, C., "Making Business Model Decisions Like Scientists: Strategic Commitment, Uncertainty, and Economic Performance," Strategic Management Journal, 2024 — https://sms.onlinelibrary.wiley.com/doi/10.1002/smj.3636

Spina, C. and Novelli, E., "Research: When Does Scientific Decision-Making Benefit Startups?", Harvard Business Review, February 2025 — https://hbr.org/2025/02/research-when-does-scientific-decision-making-benefit-startups

Supporting Sources

Spina, C., "The Science of Successful Start-Ups," INSEAD Knowledge, September 2024 — https://knowledge.insead.edu/entrepreneurship/science-successful-start-ups

Spina, C. and Novelli, E., "Rapid or Right? Making Start-Up Decisions Like Scientists," INSEAD Knowledge, May 2025 — https://knowledge.insead.edu/entrepreneurship/rapid-or-right-making-start-decisions-scientists

‍

read - Hiring In 2026: Here Is What Talents Are Loking For In Startups Today