Failure is a possible outcome for entrepreneurs, but are rates of entrepreneurial failure too high? Fairlie and Miranda (2016), for instance, report that 84.4% of US start-ups fail within seven years. Governments and the private sector invest considerable resources in the launch of new companies, so the question deserves our attention. Researchers and practitioners have developed frameworks to understand how improve the success rates of new firms, such as discovery-driven planning (McGrath and McMillan 2009), real option strategies (Adner and Levinthal 2004), design thinking (Martin 2009), lean start-up (Ries 2011), or business experimentation (Kerr et al. 2014, Gans et al. 2017).
The way that entrepreneurs predict the flow of future net revenues is also critical. On one hand, entrepreneurs employ rules of thumb, impressions, intuitions, or other semi-structured processes to predict future net revenues. On the other hand, they may apply a more scientific approach to understand and test the mechanisms that affect the performance of their new products or ideas, rather like a research scientist (Felin and Zenger 2009, Zenger 2016). Our research suggests that the scientific approach helps entrepreneurs make more precise decisions, and that they will be less likely to invest in business ideas that produce negative returns (Camuffo et al. 2017).
Case study: A tattoo search engine
Inkdome, an Italian start-up that planned to launch a search engine for finding the right tattoo artist online, provides a good example of the scientific approach. Inkdome set four main hypotheses:
- tattooed people use different tattooists,
- they search for them online,
- this takes time,
- they find online all the information they need.
It interviewed 50 potential customers, and decided to launch the search engine if all four hypotheses were corroborated at a 60% threshold. The test corroborated the first three, but not the fourth, and so Inkdome abandoned the idea. From these interviews with customers, it realised that they lacked information about reliable tattooists. So it pivoted to a new idea, which was an online consulting service to find the right tattoo artist. It tested this service using a similar approach, and concluded that it would be worthwhile to invest in the new idea.
Firms that do not adopt the scientific approach do not develop simple but clear theories. In consequence, they do not produce falsifiable hypotheses such as the four above, or clear decision rules. They tend to interview customers asking generic questions, such as “Do you like our idea?”, or “Would you buy the product?”. This creates confusion about the right decisions to make.
The precision effect
A simple numerical example can explain our approach. An entrepreneur must invest €1.15 in a business idea that can produce a discounted sum of future net revenues equal to -€1, €1, or €3. By conducting customer interviews, analyses of data, or discussions within the entrepreneurial team, the entrepreneur attributes probabilities to each of these outcomes.
In this case, we are not interested in whether an entrepreneur who adopts a scientific approach develops better ideas, but only in how she evaluates her idea compared to an entrepreneur who does not adopt this approach. Thus, we set the probabilities such that the average future net revenues predicted by the scientist and non-scientist entrepreneur are the same. The scientist-entrepreneur, however, is more precise in the sense that she is more certain about the average outcome – that is, she attributes lower probabilities to the two extreme outcomes (-€1 and €3).
Table 1 considers three potential ideas – a high-quality (H), medium-quality (M), and low-quality (L) idea – with the probabilities that a scientist and a non-scientist entrepreneur has attributed to the three outcomes (-1, 1, 3). A higher-quality idea exhibits a higher average outcome (1.8, 1.0, and 0.2). Both scientist and non-scientist entrepreneurs know that they cannot sustain negative net revenues, and thus they will never earn -1. In this case, they would earn nothing. Therefore, the last row reports the average net revenue that they would realise if they carried out the idea. The scientist-entrepreneur will always predict a lower realised average because she gives less weight to the extreme negative outcome.
Table 1 A numerical example of the scientific approach to entrepreneurial decision-making
Source: Camuffo et al. 2017.
Since both entrepreneurs make these predictions before investing 1.15 in the business idea, the scientist-entrepreneur would invest in only the highest-quality idea, while the non-scientist entrepreneur would invest in both high- and medium-quality ideas. More generally, a precise entrepreneur is more conservative, and will invest in ideas with higher average net revenues (1.8, but not 1.0). Moreover, if a scientific approach does not affect the frequency of the H, M or L ideas of the entrepreneur, the scientist-entrepreneur will abandon more ideas, and thus be more likely to close the firm or pivot to a new idea.
Performance depends, instead, on whether the true probability distribution of outcomes is closer to the one predicted by the scientist. If so, the scientist performs better because scientists and non-scientists perform equally well when ideas are H or L (they make positive or no profits), but for M ideas scientists earn zero and non-scientists suffer losses. This suggests that the wedge between the scientific and heuristic approach is most pronounced when entrepreneurial ideas are neither very good nor very bad. These are the vast majority of business ideas in a world in which entrepreneurship is not restricted to one category or group. The widespread adoption of a scientific approach to entrepreneurial decision making might reduce the failure rate of entrepreneurial activities.
Application to start-ups
Our randomised controlled trial engaged 116 Italian start-ups from different industries in an eight-week training programme. We collected data about their actions and performance at 16 points during one year. We gave all firms the same quantity and quality of training, but we pushed 59 randomly selected firms to make decisions after developing and testing rigorous hypotheses, as Inkdome had done.
We found that 24 start-ups exposed to the scientific approach dropped out, compared to 20 in the control group. More importantly, the start-ups exposed to the scientific approach pivoted 26 times, compared to 12 times in the control group. We cannot exclude that scientific training helped our firms to learn how to produce better ideas.
We found strong evidence of a precision effect:
- Pivoting. Of the 19 scientist-entrepreneurs who pivoted, five pivoted a second time, one a third time, and one four times. Of the 11 non-scientist entrepreneurs who pivoted, only one pivoted a second time. If the scientific approach provided only learning, scientists ought to pivot fewer times after the first pivot, because they would have pivoted to better ideas that were more likely to be pursued.
- Big decisions. We asked the entrepreneurs who survived to the last round of data collection whether in creating a new start-up they would feel confident in making drastic decisions, such as closing the firm. The answer of the scientist-entrepreneurs was overwhelmingly positive compared to the others.
- Average revenue. Scientist-entrepreneurs earned higher average revenue during our analysis. At the same time, the number of scientist and non-scientist entrepreneurs who earned non-zero revenue was similar (nine and eight). This is consistent with precision. Learning ought to induce more scientist-entrepreneurs to earn revenue, while precision implies that many scientist-entrepreneurs enjoy higher performance because they avoid bad ideas. When they carry out an idea, it is more profitable.
Managerial and policy implications
For start-ups, a scientific approach emphasises better screening in the search for better ideas as a way to improve performance. It raises the question whether the approach could be usefully adopted in larger firms. But this approach assumed the superiority of the scientific approach – in particular, it assumed that the probabilities of outcomes predicted by the scientific approach are closer to the 'true' probabilities. This could, however, push entrepreneurs to focus on ideas that can be assessed using the method, restricting the range of ideas that they pursue. A naïve non-scientist entrepreneur might generate successful breakthroughs by pursuing projects in which extreme positive values have higher probability of success than a scientist-entrepreneur would predict.
Adner, R and D Levinthal (2004), “What Is Not a Real Option: Considering Boundaries for the Application of Real Options to Business Strategy”, Academy of Management Review 29(1): 74–85.
Camuffo A, A Cordova and A Gambardella (2017), “A Scientific Approach to Entrepreneurial Decision-Making: Evidence from a Randomized Control Trial”, CEPR Discussion Paper 12421.
Fairlie, R and J Miranda (2016), “Taking the Leap: The Determinants of Entrepreneurs Hiring their First Employee”, Journal of Economics and Management Strategy 26(1): 3-34.
Felin, T and T R Zenger (2009), “Entrepreneurs as Theorists: On the Origins of Collective Beliefs and Novel Strategies”, Strategic Entrepreneurship Journal 3: 127–146.
Gans, J, S Stern and J Wu (2016), "Foundations of Entrepreneurial Strategy", available at SSRN.
Kerr, W R, R Nanda and M Rhodes-Kropf (2014), “Entrepreneurship as Experimentation”, Journal of Economic Perspectives 28(3): 25–48.
Martin, R (2009), The Design of Business: Why Design Thinking is the Next Competitive Advantage, Harvard Business School Press.
McGrath, R G (1997), “A Real Options Logic for Initiating Technology Positioning Investments”, Academy of Management Review 22(4): 974–996.
Ries E (2011), The Lean Startup: How Today’s Entrepreneurs Use Continuous Innovation to Create Radically Successful Businesses, Crown Publishing Group.
Zenger T R (2016), Beyond Competitive Advantage: How to Solve the Puzzle of Sustaining Growth While Creating Value, Harvard Business School Press.