What 20,000 People Taught Us About Entrepreneurship and Startups
- Spencer Greenberg & Markus Over
- 9 minutes ago
- 22 min read
Could you start a successful company?
Years ago, we wanted to help people answer this question about themselves. So, we extensively researched claims about what makes startups (and their founders) successful and compiled our findings into an interactive tool. Today, we're going to be telling you our findings about what is actually linked to successful entrepreneurship, by conducting an extensive analysis of the data from the tool.
If you haven’t tried the tool before and are interested in learning something about your aptitude to start a startup, it may be a good idea to try our tool before reading on. We’ve recently updated it, to incorporate the predictive models that we’re covering in this article – this will provide you with personalized predictions based on these results!
Key Takeaways of This Article
Much of the advice from successful multi-time founders appears solid: we found evidence to support a number of their claims, with only some claims not aligning with our findings.
Motivation matters: we found that mission-driven founders outperform self-interested ones.
Psychological traits are surprisingly predictive: traits like risk tolerance, emotional stability, proactivity, and honesty – typically studied in psychology – are better predictors than many business-centric variables.
Honesty is multifaceted: while self-reporting as “honest” and “not deceiving” were overall positive predictors, our data also indicate better startup outcomes for being able to “make something sound good, even if it's not.”
"Startup founder" is a recognizable psychological profile: The reasonably high accuracy (79% for differentiating a founder from a non-founder) shows there’s a detectable constellation of traits that distinguish founders from non-founders.
Data-driven tools can meaningfully support founder self-assessment: This research shows that well-designed personality-based tools, even relying on self-reporting, can provide valuable predictive insights.
Some of the strongest positive predictors of startup success are risk tolerance, taking initiative, seeking critical feedback, extraversion, enthusiasm, honesty.
Seeing marketing as the key to product success is negatively correlated with both raising funds and startup success, compared to seeing either a startup’s idea or its execution as the key.
Higher educational attainment is weakly but positively correlated with both starting a startup and succeeding.
Traits such as quarrelsomeness, anxiety, disorganization, and dishonesty are associated with lower startup success.
Self-perception can be (somewhat) informative: our models achieved some success in making predictions about startup success and fundraising success based on self-reported data. However, our models could only explain 10-14% of the variance in startup success and fundraising, so the traits we measured only captured a fraction of what determines a startup’s performance – though given the high degree of uncertainty and risk in startups – with the conventional view being that 9 out of 10 startups fail – this is probably not surprising that startup success is only somewhat predictable. It was easier to predict who would start a startup than how successful that startup would be.
Our approach in creating this tool was to look for advice from extremely successful multi-time founders (and startup investors) – people like Paul Graham and James Altucher. Their ability to produce successful companies repeatedly (and invest in successful startups) suggests they know what they're talking about when it comes to what makes a great startup founder.
As there was little reliable data available at the time, collecting and synthesizing the advice of such experienced founders was our best shot at providing benefit to our users. But our tool had a second purpose: it also allowed us to verify which of their claims accurately predict success! We asked users to let us know about their own history in starting and running companies. Over 20,000 users have gone through the tool, around a third of whom started a company in the past. This one-of-a-kind dataset has now allowed us to find out which traits, experiences, and views people hold really predict people’s likelihood of founding a successful company. And in this post, we’ll share these results with you, so buckle up!
What We Tested
Our tool was based on many interesting claims and advice from successful entrepreneurs, for instance:
Start-ups need to consist of people who could be described as an “animal” (Paul Graham), which, according to Paul Graham, might mean: “a salesperson who just won't take no for an answer; a hacker who will stay up till 4:00 AM rather than go to bed leaving code with a bug in it; a PR person who will cold-call New York Times reporters on their cell phones; a graphic designer who feels physical pain when something is two millimeters out of place..."
"The most important quality in a startup founder is determination. Not intelligence — determination… Everyone who deals with startups knows how important commitment is, so if they sense you're ambivalent, they won't give you much attention" and "You have to be determined, but flexible, like a running back" (Paul Graham)
You need to understand your users (Paul Graham)
You need to seek feedback on your product early and frequently (Paul Graham)
You need to be “naughty,” in the sense of being willing to bend the rules to your advantage (Paul Graham and Sam Altman)
A startup’s structure, particularly in the beginning, should be closer to a monarchy than to democracy (Peter Thiel)
A startup’s founders are more important than the idea the startup is built on (Paul Graham)
You should have a product and customers before trying to raise VC money (James Altucher)
You should be very honest with your clients and business partners (James Altucher)
“The CEO is not only the communicator of the vision, the CEO is the consensus builder for that vision” (George Deeb)
We then asked users for their views or traits regarding these claims. We also asked users about their personal startup history. If they were involved in running a startup before, we further asked for some details about that, including about their fundraising, the market value of their startup, and exit conditions, if applicable.
It’s important to keep in mind that all the data in this study is self-reported. This might affect the views and character traits – because people might view and describe themselves differently from how others view them or how they really are. It also impacts the accuracy of the data people provide on their entrepreneurial history. On the plus side, this approach allowed us to gather relevant data from about 7,000 users who had previously started a company, yielding a unique and large dataset on startup success. Based on this, we could then train regression models to find the most reliable predicting factors for three questions:
Founder status: What predicts whether a person has started a startup before?
Capital raised: Among those who have founded a startup, what predicts how much funding they raised?
Startup success: Among those who have founded a startup, what predicts how successful their startup has been?
And this allowed us to put the intuitions of the successful entrepreneurs to the test! If their claims about what makes a good founder were right, then we should, in all likelihood, be able to identify these traits as good predictors for past successes in running a startup – and, thus, assuming that our findings generalize, also as good predictors for future success.
We created three predictive models, one for each of the three questions. These models allow us to quantify how predictive individual traits are for these questions. If you’re interested in the technicalities of what exactly we did and why, have a look at this article’s appendix, which includes more in-depth explanations. The appendix also contains a full list of all the coefficients and correlations we found, whereas the article itself will focus on the most interesting ones.
One more note before we dig into the details: a large number of people have taken this test, so even very small correlations would be "statistically significant", hence statistical significance is not a very useful way to approach these questions. Instead, we recommend being mindful of the size of the effects, as some of them are quite small predictors.
What We Found
Let’s look at the three questions we investigated one by one.
(1) Founder Status – What predicts whether a person starts a startup?
Out of the over 20,000 users who went through the tool, more than a third had started a company in the past. Thus, we had quite a large amount of data from which to explore predictors. We used the data of 16,760 participants to train our model and left the data of 4,188 participants in our test set for evaluation (i.e., we didn't train our predictive model on it; we just used it to measure how accurate the predictive model was).
Of course, the population that has taken our test is not very representative of the general public (since, for example, by virtue of taking this test, they were much more likely to be interested in starting a company than the average person), so the results are most relevant for the type of person that’s interested enough to take such a test. That being said, the users our data is based on are probably rather representative of the people reading this article (especially readers with an interest in entrepreneurship), as the audience of our tools and our articles naturally has a large overlap.
Here’s a small selection of the types of statements and questions (that people answered about themselves) that we included in our analysis to see if they predict whether people started a startup:
I see myself as extraverted and enthusiastic.
What’s the highest level of education that you’ve completed?
I see myself as sometimes deceptive and devious.
I consider myself to be more of a rule breaker. (Compared to a “rule bender” or a “rule follower”)
I am willing to take large risks.
Age
Which of these would you expect to be predictors for someone having started a startup, and in which direction (meaning does each factor rather predict that a person did start a startup before, or they didn’t)? Before we tell you the answers, see if you can guess them.
We tested a total of 73 variables and have attached a detailed table of all the coefficients and correlations we found at the end of this article. Many of them were not meaningfully predictive, but some of them were, so let’s, for now, focus on the most interesting ones we identified. The "correlation" column quantifies the relation between a trait and the predicted metric in isolation (not taking into account the other variables), whereas the "coefficient" column answers the question of how the given trait predicts the outcome when holding all other traits constant (loosely speaking). So, the coefficient tells us how much the trait predicts starting a startup above and beyond the effects of the other predictors. You can find a more detailed explanation of these columns and their interpretation (including how values are standardized) in the appendix.
As for the six factors mentioned above, it turns out all of them were at least somewhat predictive about whether a person started a startup before, and they were all positive predictors except for "I see myself as sometimes deceptive and devious," which is a negative predictor:
Question / Statement | Association with Starting a Startup | Coefficient from the full logistic regression (with L1 regularization) | Individual Correlation |
|---|---|---|---|
I see myself as extraverted and enthusiastic. | Positive | 0.19 | 0.23 |
What’s the highest level of education that you’ve completed? | Weakly positive | 0.06 | 0.14 |
I see myself as sometimes deceptive and devious. | Weakly negative | -0.12 | -0.17 |
I consider myself to be more of a rule breaker. (as opposed to a “rule bender” or a “rule follower”) | Weakly positive | 0.08 | 0.11 |
I am willing to take large risks. | Positive | 0.37 | 0.29 |
Age | Positive | 0.48 | 0.26 |
ing is the fact that being deceptive is negatively correlated with starting a startup – particularly given that “rule breaker” types are more likely to have founded a startup than rule followers.
Being willing to take risks is a particularly strong predictor of starting a company. In fact, it is the strongest predictor we found! Well, besides a person’s age. Age is somewhat less interesting as a predictor due to being both obvious (you just had more time to have founded something in the past) and not actionable (you can work on your willingness to take risks, but you can’t just age faster – though you can decide at what age you start a startup). It makes sense that being risk tolerant is a potent predictor, as it is widely believed that about 90% of startups fail. Creating your own company is a very high-risk (and potentially high-reward) life strategy, so naturally, some selection for risk-tolerant founders is happening. James Altucher argues that “Starting a business is not about taking risks. It’s 100% about risk mitigation.” – yet, you need first to tolerate the inherent risk before you can work on mitigating it.
Here are some more predictors for founder status:
Question | Association with Starting a Startup | Coefficient | Correlation |
|---|---|---|---|
I handle myself very effectively in situations that others would find highly stressful. | Positive | 0.07 | 0.22 |
People might describe me as an "animal." | Positive | 0.14 | 0.20 |
I can usually talk my way out of anything. | Weakly positive | 0.04 | 0.13 |
I see myself as critical and quarrelsome. | Weakly negative | -0.06 | -0.11 |
We’ve now covered some of the strongest predictors of starting a company, but let’s also look at the other end of the spectrum: factors that, perhaps surprisingly, turned out not to make our models more predictive, and hence were dropped from the regression:
“I feel highly driven to have enormous achievements during my lifetime.” With an individual correlation of 0.17, this was, in isolation, a fairly decent predictor of starting a company, yet it added little value to the regression beyond the other independent variables.
“I believe my future is dictated far more by how I choose to behave than by luck or fate.” had a correlation of 0.11, but also did not improve the logistic regression.
“When handling a crisis situation, I…” with the answer ranging from “I take charge” to “I look to the others in the group to find out what they need from me.” The answers to this question, too, were somewhat predictive in isolation, with a correlation of -0.17 (meaning that agreeing more with “I look to the others in the group to find out what they need from me” than with “I take charge” was negatively associated with having started a startup), but were not able to improve the regression model.
Taking all of the predictors we identified (including some we didn’t mention here, which you can find in the table in the appendix) into account, our logistic regression model achieved an AUC score of 79%, which is quite good. This means that, if given two people, one of whom started a startup and one of whom didn't, it could correctly identify which is which 79% of the time! This indicates that the factors we identified indeed capture many relevant aspects that allow us to make an informed guess about whether a particular type of person has founded a startup before (or, to the extent that we can extrapolate from this, which kind of person is likely to found a startup in the future).
(2) Capital Raised – What are the main predictors of successfully raising money for a startup?
In the previous section, we discussed factors that allow us to predict whether a person is likely to have founded a startup based on some characteristics that are plausibly causally upstream of that. However, maybe a question that’s even more interesting is this: given a person starts a start-up, what predicts their startup’s performance?
One way to quantify this is to use their success at raising money as part of their entrepreneurial career, which we also gathered data about in our survey.
Based on this data, we created a linear regression model (not a logistic regression this time because we’re trying to predict a continuous number rather than a binary classification) to predict how much money individual founders raised in the past, based on traits and opinions that are likely causally upstream of their entrepreneurial career. To be precise, our regression model doesn’t predict the money raised directly but rather the logarithm of that (since money raised is an unbounded variable that can span many orders of magnitude, taking the logarithm is important to prevent outliers from distorting the results – it's essentially like predicting the number of orders of magnitudes of money raised, instead of the money itself).
The predictions of this model are based on the 7,672 users who had started a company before and have provided the necessary information we needed to feed into the model. 57% of these users (or 45% out of all users who claimed to have startup experience) had successfully raised money for their startup:
Looking at the most predictive of the 73 variables we tested, at the top of the list (which, again, you can find in full at the end of this article), we find several of the answers to an interesting question: “If you made a billion dollars from your startup, which of these would you be most likely to do?”. Users could select a variety of answers, as well as “None of the above.”
Here are the coefficients our linear regression model came up with for the individual answers to it:
Response | Association with Capital Raised | Coefficient from the full linear regression (Lasso method) | Correlation |
|---|---|---|---|
Become an investor. | Positive | 0.32 | 0.11 |
Extensively research how to optimize my impact on the world. | Weakly positive | 0.29 | 0.05 |
Make another billion dollars. | Weakly positive | 0.25 | 0.05 |
Donate to the charitable causes I’m most passionate about. | Weakly positive | 0.16 | 0.04 |
I don’t know… ask me when I get there. | Mixed | 0.16 | -0.05 |
Retire and enjoy life! | Mixed | 0.04 | -0.14 |
All these coefficients can be understood as being relative to the “None of the above” option (which is most associated with failure to raise capital – which does not have its own coefficient because it serves as the point of comparison).
It’s not obvious how to interpret these findings. Maybe it’s just that this question, and the top three answers above in particular, correlate with something like “ambition,” which may, for a variety of reasons, be helpful in raising money for your startup. Or maybe "None of the above" is such a bad option (and hence, all the other answer options have positive coefficients relative to it) because it suggests a lack of goal or driver. Or maybe there’s more to it – it’s hard to draw any firm conclusions based on this data alone.
The “Become an investor” answer looks like a clear-cut case of a positive predictor of raising capital, but there's some danger it is affected by reverse causation! It’s possible that having interacted with investors in the past (due to having founded a company before) makes it more likely for people to choose this response. This points to how careful one needs to be when interpreting coefficients cautiously when it comes to assuming that they are causes.
Another interesting question was this: "If you were going to start a new company now, and you could choose what the result is 5 years in the future, which of these outcomes would you truly choose?"
Here are its answers and their corresponding coefficients:
Response | Association with Capital Raised | Coefficient | Correlation |
|---|---|---|---|
Impact | Positive | 0.11 | 0.18 |
Money | Negligible | 0.02 | 0.02 |
Fame | Weakly negative | -0.04 | -0.06 |
Fun | Negative | -0.10 | -0.21 |
So, study participants who were in it to have a positive impact on the world raised more money on average than those who cared more about fame or having fun (note that these were not measured relative to any other default option). This seems like a positive sign, as it seems best for society if impact-oriented founders get more funds than those motivated by money or fame.
Paul Graham once wrote that he “got being a good startup founder down to two words: relentlessly resourceful.” So naturally, we tested this as well. We asked participants which description best matched them: relentlessly resourceful, painstakingly practical, defiantly decisive, or inspiringly idealistic.
Here are the results in terms of predicting startup capital raised:
Response | Association with Capital Raised | Coefficient | Correlation |
|---|---|---|---|
Relentlessly resourceful | Weakly positive | 0.05 | 0.12 |
Defiantly decisive | Negligible | 0.04 | 0.03 |
Inspiringly idealistic | Weakly negative | -0.01 | -0.08 |
Painstakingly practical | Weakly negative | -0.04 | -0.09 |
This gives some credibility to Paul Graham’s claim, as indeed, people who consider themselves to be relentlessly resourceful were, on average, notably better at raising money for their startup! Of course, because this is based on self-report, we don't know that these people really are more relentlessly resourceful. It also could be that founders who were more successful at raising money were more likely to be aware of Paul Graham's work and more likely to apply this phrase to themselves because of this added awareness.
One more interesting finding came from the question: “When it comes to a new product, what do you consider to be the biggest factor of success?”
Here are the results:
Response | Association with Capital Raised | Coefficient | Correlation |
|---|---|---|---|
Idea | Negligible | 0.02 | 0.01 |
Execution | Mixed / Weakly positive | 0.00 | 0.07 |
Marketing | Weakly negative | -0.07 | -0.08 |
So, while we cannot infer that much about the difference between valuing a startup’s idea or execution more, putting too much weight on marketing seemed to be linked to less fundraising success.
Again, there were also some factors that, surprisingly to us, were not predictive of money raised:
“Which of these two statements describes you better?” Answers ranged from ”I'm not afraid to ask for what I need, regardless of whom I'm asking it of" to “I'm careful not to make other people uncomfortable.” A correlation of -0.11 indicates that choosing the first answer over the second was, in isolation, predictive of raising more money, yet this question was not able to meaningfully influence the linear regression. So, it appears that other factors sufficiently account for the information value captured by this question.
"Which of these two statements describes you better?” with answers ranging from "It's easy for me to persuade people that I'm right" to "I respect other people's perspectives." Persuasiveness, too, seems important, but was not predictive in this case. Even individually, with a correlation of only -0.04, this question did not prove to be predictive.
All things considered, our linear regression model, which predicts the amount of money raised among users who have started a startup before, achieved an r score of 0.37 (R² = 0.14). This means that our model’s predictions have a 0.37 correlation with the logarithm of actual money raised (and was able to account for about 14% of the variation in values). So, the model does have some ability to predict capital raising success, but it's extremely far from perfect in its predictions.
In practical terms, this means that the model can offer a modest amount of help in identifying traits that tend to be present among more startup founders who were successful at raising capital. For example, if you were evaluating a set of startup teams and trying to make a rough guess about which ones might go on to raise more money, our model could help you to spot signs that would improve your odds over random guessing.
Of course, the predictive power of the model is limited: many successful founders won’t fit its patterns. As with all the models discussed in this article, the best way to use it is as a source of directional insight; a tool for reflection and hypothesis testing rather than definitive prediction or absolute validation.
(3) Startup Success – What are the main predictors of startup success?
Lastly, we sought to identify the best predictors of startup success in general, beyond just the money raised. After all, some startups become very successful without raising a lot of (or any) money. So we created another linear regression model, but this time, the quantity it predicted was an overall measure of the success of their startup, produced as a combination of several factors that the founders provided to us, namely:
Funds raised, if any
What the company was sold for, if it was
The latest valuation of the company, if not sold
We then combined these values into a numeric “success” metric, designed in such a way that higher values would represent a higher degree of success, which our model could then try to predict. You can find the precise definition of this (quite complex) metric in the appendix.
Obviously, this is not a perfect proxy for success. Plenty of startups achieve meaningful impact without raising much money, being acquired, or reaching a high valuation – especially those that are mission-driven or operating in niche markets. Our metric also can't capture qualitative aspects of success, such as customer satisfaction, innovation, team growth, or founder fulfillment.
That said, the metric still provides a useful proxy for success because it captures outcomes that are commonly used as benchmarks. Funding raised, acquisition price, and valuation are tangible, quantifiable signs that a company has achieved external validation, resources, or interest – all of which influence future opportunities for growth, visibility, and impact.
To start, here are five traits our model identified as positive predictors for startup success:
Trait | Association with Startup Success | Coefficient | Correlation |
|---|---|---|---|
I am willing to take large risks. | Positive | 0.12 | 0.21 |
I am a very proactive person who takes initiative. | Positive | 0.07 | 0.22 |
I see myself as extraverted and enthusiastic. | Weakly positive | 0.04 | 0.17 |
I see myself as always honest and truthful. | Weakly positive | 0.03 | 0.12 |
What’s the highest level of education that you’ve completed? | Weakly positive | 0.02 | 0.10 |
Interestingly, these traits were already predictive of who started a company. And now, additionally, among just those who started a company, they are also predictive of their company’s degree of (numeric) success! So, it appears that these traits are not only more prevalent in individuals who are willing to found a startup, but also, in some way, helpful for or indicative of running a startup well.
Here are some other positive predictors:
Trait / Response | Association with Startup Success | Coefficient | Correlation |
|---|---|---|---|
When you’ve created something new, how often do you actively seek critical feedback from others about it? | Weakly positive | 0.06 | 0.14 |
I see myself as calm and emotionally stable. | Weakly positive | 0.02 | 0.08 |
As well as some negative predictors, that is, ones that are indicative of lower degrees of success:
Trait / Response | Association with Startup Success | Coefficient | Correlation |
Agreeing more with “I do my best to avoid using other people” rather than with “If there’s someone I want to meet, I will find a way to make it happen” | Weakly negative | -0.07 | -0.17 |
Agreeing more with “I tell it like it is, I don’t sugar-coat things” than with “I know how to make something sound good, even if it’s not” | Weakly negative | -0.05 | -0.06 |
I see myself as anxious and easily upset. | Weakly negative | -0.03 | -0.17 |
I see myself as reserved and quiet. | Weakly negative | -0.03 | -0.16 |
I see myself as disorganized and careless. | Weakly negative | -0.02 | -0.14 |
I see myself as critical and quarrelsome. | Weakly negative | -0.01 | -0.08 |
The question mentioned earlier, to test Paul Graham’s claim that the best founders are “relentlessly resourceful,” also turned out somewhat predictive here (at least, in the sense that it wasn't an answer option in that question that was negatively predictive). The order of predictiveness of these answer options is the same as in the model predicting money raised: being relentlessly resourceful again comes out on top, as claimed by Paul Graham:
Response | Association with Startup Success | Coefficient | Correlation |
|---|---|---|---|
Relentlessly resourceful | Mixed / weakly positive | 0.00 | 0.10 |
Defiantly decisive | Negligible | 0.00 | 0.01 |
Inspiringly idealistic | Weakly negative | -0.04 | -0.06 |
Painstakingly practical | Weakly negative | -0.04 | -0.08 |
Finally, one question led to somewhat different results than in our fundraising model: “When it comes to a new product, what do you consider to be the biggest factor of success?”
Response | Association with Startup Success | Coefficient | Correlation |
|---|---|---|---|
Execution | Weakly positive | 0.06 | 0.11 |
Idea | Mixed | 0.00 | -0.07 |
Marketing | Mixed | 0.00 | -0.07 |
When predicting money raised, idea and execution performed similarly well, with marketing trailing behind. In this model, “Execution” turns out to be the best predictor of overall success.
Two factors that were not predictive of startup success:
“When approaching problems, I tend to be better at coming up with surprising new ideas and tactics.” and “When approaching problems, I tend to be better at solving problems quickly using tried and true methods.” were both part of the survey. While creativity may be an important quality in a founder, the responses to this question were not predictive in our model, as the correlations were merely 0.05 and -0.05 respectively, and the linear regression did not improve when accounting for these questions.
“Which of these best describes what motivates you to be a startup founder?” Possible responses were fame, money, friends, and passion. The correlations for the four options were -0.06, -0.08, -0.05, and +0.12 (respectively), so they had some predictive power. But they did not improve the linear regression beyond the other available factors.
One explanation for our findings is that, indeed, seeing the idea as most important correlates with being able to raise more money, while seeing execution as most crucial leads to greater success in general.
Overall, this third linear regression model achieved an r score of 0.3 (R² = 0.10), indicating that the model’s predictions have a correlation of 0.3 with our success metric, explaining 10% of the variation in values. Similar to the model for capital raised, this indicates that our model captures some relevant factors, but is still far from perfect.
This means that (again similar to the model for capital raised) this model can offer a modest amount of help in identifying traits that tend to be present among more successful startup founders (as measured by our success metric). It can help you improve your chances at spotting startups that will go on to reach higher valuations or exits, and it can be useful for reflection and hypothesis testing, but individual predictions will still be unreliable.
Conclusion: What Predicts Entrepreneurship and Startup Success
As with any study of this nature, it’s important to emphasize that our findings are correlational rather than causal. Just because a particular trait or belief is associated with startup success doesn’t necessarily mean that cultivating it will directly cause you to succeed. However, even if some of these factors aren’t strictly causal, they may still serve as useful predictors of entrepreneurial outcomes based on patterns we’ve observed across thousands of founders. So, while our findings may not tell founders what to do to improve their odds, they can suggest causal hypotheses, and they may still provide some insight into whether a given person is likely to be a good founder.
Keeping this in mind, our findings do lend empirical support to several of the claims made by renowned entrepreneurs and startup accelerators. For example:
Claim | Verdict | Notes |
|---|---|---|
You need to seek feedback on your product early and frequently (Paul Graham) | ✅ | We found evidence that seeking user feedback early and often predicts startup success (see Appendix….). |
You should be very honest with your clients and business partners (James Altucher) | ✅ | Seeing oneself as honest and truthful was a rather robust positive predictor in our models, not only for starting a startup but also for it being successful. However, the fact that agreeing with “I know how to make something sound good, even if it's not” was also somewhat predictive of startup success indicates that things are more complicated. |
A startup’s founders are more important than the idea the startup is built on (Paul Graham) | ✅ | We found some weak evidence for this, as participants valuing the importance of a startup’s idea the highest were, on average, less successful than those putting more weight on execution (though this is only an indirect test of Graham's claim) |
Our data also lends empirical support to some more common-sensical views on entrepreneurship:
✅ More proactive and extroverted people are more likely to start a company, and given they do, are more successful on average
✅ Emotional stability and dealing well with stress are likewise predictors of founding a startup as well as, albeit more weakly, of its success
✅ Risk-tolerant people are more likely to start a company, and even among founders, the more risk-tolerant ones tend to be more successful
At the same time, some of the claims by startup gurus that we investigated did not quite match our findings:
Claim | Verdict | Notes |
|---|---|---|
Start-ups need to consist of people who could be described as an “animal” (Paul Graham) | ❌ | While this was a good predictor in our data for who did start a company, it did not turn out to predict startup success or money raised among founders (as a self-reported indicator – perhaps ratings by people who work with the respondent, rather than self-report, would have yielded a different result) |
You need to be “naughty,” in the sense of being willing to bend the rules to your advantage (Paul Graham and Sam Altman) | ❌ | Same here: describing oneself as a rule breaker was predictive for starting a startup in the first place, but not for founder success. |
However, this does, of course, not necessarily mean these claims are wrong – we have to be mindful of our study's limitations.
Ultimately, while there’s no perfect formula for predicting startup success, our research suggests that certain traits, mindsets, and behaviors are associated with whether someone is likely to found a company and, among founders, whether they’ll raise substantial funding or build a successful business. We also found evidence in favor of many of the claims of successful entrepreneurs, while a few of their claims were inconsistent with our data.
If you found these insights intriguing, you might want to try our interactive tool yourself. It’s designed to help you explore how your own traits and experiences align with those of successful entrepreneurs. Who knows? You might just discover that you have what it takes to build the next great startup!