Somewhere in the mid-2000s, a statistic entered startup culture and never left: '90% of startups fail.' It appears in pitch decks, in entrepreneurship textbooks, in TED Talks, and in the opening paragraphs of countless articles about risk-taking. There is only one significant problem with it: the number does not appear to be true, the term 'startup' is rarely defined, and the term 'fail' is used interchangeably with 'close,' 'pivot,' 'get acquired,' and 'not achieve VC-scale returns.'
The actual data on new business survival — drawn from the Bureau of Labor Statistics' longitudinal employer data, the most comprehensive and methodologically rigorous source available — tells a different and more nuanced story. Approximately 20% of new businesses close in their first year. About 45% close by year five. Roughly 65% close by year ten. These are genuine and sobering figures, but they are not '90% within the first year' by any methodology that holds up to scrutiny.
The 90% figure may reflect something real about a specific subset of startups — specifically, those attempting to become high-growth, VC-fundable technology companies. In that narrowly defined category, the failure rate measured against venture expectations (10x return on invested capital, typically) is indeed very high. But most new businesses are not chasing venture capital. Most are small service businesses, retail operations, food establishments, and professional services firms, and their failure rates, while high, are not 90%.
This piece compiles the actual data on new business and startup survival, industry differences, VC versus bootstrapped outcomes, and the factors that research most consistently associates with survival.
"The 'most startups fail' narrative is both true and wildly overstated, depending on what you mean by startup and what you mean by fail. The data we have is better and more interesting than the number everyone keeps repeating." — Dane Stangler, Ewing Marion Kauffman Foundation, 2023
Key Definitions
New Business Survival Rate: BLS tracks the survival of 'new establishments' — businesses that opened in a given quarter — over subsequent years. This is the most rigorous longitudinal data on business survival in the United States, covering all industries and all sizes. It does not distinguish between 'startups' (venture-aspiring tech companies) and small businesses (local service providers).
Startup Failure (VC Context): In venture capital, a startup 'fails' if it does not return investors' capital. This includes companies that shut down, companies that are acquired for less than invested capital, and companies that operate indefinitely at below-threshold scale. This is a much stricter definition than BLS's 'establishment closure.'
Premature Scaling: The Startup Genome Project's term for the pattern of companies that expand spending, hiring, and marketing before achieving product-market fit. It is the most common identified failure pattern in their analysis of over 10,000 startups.
Product-Market Fit (PMF): The condition in which a product satisfies a strong market demand — characterized by organic user growth, high retention, and customers who would be 'very disappointed' if the product disappeared (typically measured via the 'Sean Ellis test,' where 40%+ of users expressing strong disappointment signals PMF).
Runway: The number of months a startup can operate before running out of cash, assuming current burn rate. A common heuristic is to maintain 18-24 months of runway; falling below 6 months without a clear path to profitability or fundraising is considered a danger zone.
What the BLS Data Actually Shows
The Bureau of Labor Statistics' Business Employment Dynamics (BED) data tracks the survival of all new business establishments in the United States over time. It is the most methodologically consistent and comprehensive dataset available for answering the question 'how long do new businesses last?'
The data, analyzing cohorts of new establishments from 1994 through 2019 (the most recent fully analyzed cohorts), shows strikingly consistent survival curves across decades:
Year 1: approximately 80% of new establishments survive. Year 2: approximately 70% survive. Year 3: approximately 62% survive. Year 5: approximately 55% survive. Year 10: approximately 35% survive.
These figures represent survival of any kind — the business is still operating in some form, regardless of its financial performance, ownership structure, or industry. 'Survival' does not mean 'thriving' or 'achieving its founders' goals.' It means the establishment has not closed.
Importantly, BLS data also captures that the 2010s showed modestly better survival rates than the 1990s and early 2000s, consistent with research suggesting that improved access to business tools, cloud computing, and low-cost digital marketing have reduced some barriers to early-stage survival. The improvement is meaningful but not dramatic — a few percentage points across the 5-year and 10-year survival windows.
Survival Rates by Industry
The BLS data allows for industry-level analysis that reveals significant variation. The industries with the lowest 5-year survival rates are consistent over time:
Construction has among the lowest survival rates, with approximately 42% of new construction firms surviving 5 years. The combination of project-based cash flow, intense competition, thin margins, and vulnerability to economic cycles makes construction one of the most challenging industries for new business formation.
Retail — particularly single-location retail — shows approximately 44% five-year survival. The structural headwinds from e-commerce, rising commercial rents in viable locations, and competition from national chains have worsened retail survival rates over the past decade relative to the longer historical average.
Food and beverage / restaurants are frequently cited with survival rate statistics that vary wildly. The actual BLS data puts independent restaurant survival at approximately 40% after five years — challenging, but not the 90% failure rate often cited in popular media. The popular overstatement appears to trace to a 1990s study by Professor H.G. Parsa that measured failure rates differently and in a specific local market.
Professional and technical services (consulting, engineering, software) show substantially better survival rates — approximately 58% at five years. Lower capital requirements, scalable business models (particularly for software), and demand from enterprise clients contribute to better survival outcomes.
Healthcare and social assistance businesses show among the highest survival rates (approximately 61% at five years), benefiting from steady demand, insurance reimbursement systems, and barriers to entry that limit competition.
VC-Backed Startup Outcomes
Venture-funded startups represent a small fraction of all new businesses — approximately 0.5% of new businesses in the US receive any venture capital in a given year — but they receive the overwhelming majority of attention in startup culture, business media, and entrepreneurship education.
The VC math starts with the expectation of a power law distribution of returns. VC fund models typically assume that most portfolio companies will either fail outright or return minimal capital, a small number will provide modest returns, and one or two will generate the dominant share of the fund's profits. A commonly cited heuristic in VC is the 'rule of thirds': approximately one-third of portfolio companies fail (return zero capital), one-third return invested capital or a modest multiple, and one-third generate meaningful returns.
Cambridge Associates' benchmarking data, which aggregates venture fund performance across hundreds of funds, shows that the median venture-backed startup returns less than the original invested capital to its investors. The best-performing quartile of VC-backed startups accounts for the overwhelming majority of industry returns.
CB Insights' analysis of US venture-backed startups funded between 2008 and 2018 — a period spanning both the post-financial-crisis recovery and the bull market — found that approximately 67% of startups that raised a Series A round ultimately did not return investors' capital through any exit route (IPO, acquisition, or secondary sale). The percentage returning capital meaningfully to all investors (including early-stage) was lower still, approximately 10-20% depending on fund vintage and sector.
The unicorn — a venture-backed company reaching a $1 billion valuation — has become a widely tracked metric. As of Q4 2025, approximately 1,500 companies globally hold unicorn status, per CB Insights. The US leads with approximately 700 active unicorns, followed by China (approximately 300) and India (approximately 70). Notably, the unicorn list is crowded with companies that achieved the $1 billion valuation threshold during the 2020-2021 bull market and have seen valuations collapse in subsequent down rounds or markdowns — so the list overstates the count of currently thriving billion-dollar companies.
Why Startups Fail: The Research
CB Insights' post-mortem analysis is one of the most frequently cited sources on startup failure reasons. Their analysis of over 110 startup post-mortems identifies the following as the most common failure factors:
No market need (42% of cases): The product solved a problem that either did not exist at meaningful scale or that potential customers did not prioritize sufficiently to pay for. This is the 'brilliant solution looking for a problem' failure mode.
Ran out of cash (29%): The company exhausted its capital before reaching a sustainable revenue base. Commonly associated with premature scaling, extended sales cycles (particularly in enterprise sales), and failure to raise follow-on funding.
Not the right team (23%): Skill gaps, co-founder conflict, and inability to attract talent are all captured in this category. Research consistently shows that team quality is the factor most predictive of investor decisions at early stages.
Got outcompeted (19%): The product was viable, but a better-resourced, better-executed, or earlier-moving competitor captured the market.
Pricing/cost issues (18%): The unit economics never worked — either the cost of acquiring customers exceeded lifetime customer value, or margins were insufficient to sustain a viable business, or pricing was misaligned with willingness to pay.
The Startup Genome Project's research, which has analyzed data from over 10,000 startups, identifies premature scaling as the underlying pattern in the majority of failures — often appearing before or alongside the specific reasons listed in CB Insights' analysis. Companies that scale prematurely — hiring ahead of revenue, expanding marketing before proving product-market fit, building operations for a scale they have not reached — are 2.3x more likely to fail than those that scale after establishing product-market fit.
Bootstrapped vs. Venture-Backed Outcomes
The comparison between bootstrapped (self-funded) and venture-backed startups is complicated by selection effects: the types of businesses that bootstrap and those that seek venture capital are fundamentally different in their ambitions, models, and industries.
Profitable, sustainable bootstrapped businesses are common and largely invisible in business media — they do not make headlines because they neither raise money publicly nor go public. Indie Hackers, a community platform for bootstrapped founders, documents thousands of profitable software businesses generating $10,000 to $1 million+ in monthly recurring revenue with no external investment. These businesses have high survival rates precisely because profitability means they are not dependent on external capital cycles.
Academic research by Robb and Robinson (2014) found that bootstrapped startups showed lower short-term survival rates than funded startups (because capital access provides runway), but among survivors, bootstrapped businesses were more likely to be genuinely profitable after 3-5 years. The funded businesses had higher survival but more often were 'alive but walking dead' — burning capital without a clear path to profitability.
The question of which model produces better outcomes for founders depends entirely on what founders want. Venture capital optimizes for magnitude of outcome at the expense of probability — you might get a large exit, or you might get zero. Bootstrapping optimizes for probability of a stable, profitable outcome at the expense of potential magnitude. Neither is objectively superior.
What Predicts Survival: The Evidence
Several factors emerge consistently across the research literature as the strongest predictors of startup survival.
Founder experience is the most reliable single predictor. Research by Gompers, Kovner, Lerner, and Scharfstein (2006, published in Journal of Finance) found that entrepreneurs who had previously founded a venture-backed company that went public had a 30% success rate on their next venture, compared to 21% for first-time entrepreneurs. The learning from failure and the networks built through prior startup experience are both significant.
Co-founder configuration matters. Partnerships (specifically complementary technical and commercial co-founders) outperform solo founders and functionally-similar co-founder pairs. Research published in Harvard Business Review found that two-founder companies raised 30% more investment, grew customer base 3x faster, and were 19% less likely to scale prematurely than solo-founded startups.
Market timing may matter more than product quality. Peter Thiel and Bill Gross (in a 2015 TED Talk using Idealab's portfolio data) both identified timing as one of the most important factors in startup success — being too early is almost as fatal as being too late, because an early-to-market company exhausts resources before the market is ready.
Revenue orientation from day one is strongly associated with survival. Profitable small software companies (SaaS bootstraps, micro-SaaS, consulting-led product companies) show dramatically better survival rates than companies that delay revenue focus in pursuit of scale. The Lean Startup methodology's emphasis on validated learning and early customer revenue is supported by the empirical survival data.
Practical Implications
For aspiring entrepreneurs, the most important implication of the data is that failure is common but not inevitable, and that specific factors under founders' control significantly affect the odds. Building with a complementary co-founder, validating market need before building extensively, and maintaining tight cost control during the pre-product-market-fit phase all improve odds substantially.
For investors, the power law distribution of VC returns is a mathematical feature of the asset class, not a failure of execution. Portfolio construction strategies that maximize the number of shots at high-return outcomes — while maintaining discipline on entry valuation — remain the core VC investment logic.
For policymakers, the BLS data shows a relatively stable rate of new business formation and survival across decades, with small improvements in recent years. The research identifies access to early revenue (government procurement programs, university technology transfer) and mentorship (accelerator programs, SBIR grants) as the most evidence-backed policy interventions for improving startup survival.
References
- Bureau of Labor Statistics. (2025). Business Employment Dynamics: Survival of Private-Sector Establishments. bls.gov.
- CB Insights. (2021, updated 2024). The Top 12 Reasons Startups Fail. cbinsights.com.
- Startup Genome Project. (2019, updated 2023). Global Startup Ecosystem Report. startupgenome.com.
- Gompers, P., Kovner, A., Lerner, J., & Scharfstein, D. (2010). Performance Persistence in Entrepreneurship. Journal of Financial Economics, 96(1).
- Robb, A.M., & Robinson, D.T. (2014). The Capital Structure Decisions of New Firms. Review of Financial Studies, 27(1).
- Parsa, H.G., et al. (2005). Why Restaurants Fail. Cornell Hotel and Restaurant Administration Quarterly, 46(3).
- First Round Capital. (2019). The 10 Year Project: Insights from 10 Years of Investing. review.firstround.com.
- Cambridge Associates. (2024). US Venture Capital Index and Selected Benchmark Statistics. cambridgeassociates.com.
- Kauffman Foundation. (2023). State of Entrepreneurship 2023. kauffman.org.
- Ries, E. (2011). The Lean Startup. Crown Business. (Follow-up research cited throughout.)
- CB Insights. (2025). Global Unicorn Club. cbinsights.com.
- OECD. (2023). Entrepreneurship at a Glance 2023. oecd.org.
Frequently Asked Questions
What is the actual startup failure rate?
The widely cited '90% of startups fail' figure is not well-sourced and appears to conflate different definitions of 'startup' and 'failure.' Bureau of Labor Statistics data on new business survivability — the most comprehensive and methodologically rigorous US dataset — shows that approximately 20% of new businesses fail in their first year, 45% by year 5, and 65% by year 10. These are broadly consistent with OECD data across advanced economies. 'Failure' in BLS terms means business closure, which includes intentional closures, acquisitions, and mergers that the owner considers a success. Startups seeking venture capital have a different profile: CB Insights analysis of VC-backed startups shows approximately 75% do not return investors' capital, though most of these companies do not 'fail' in the BLS sense — they simply do not generate VC-scale returns.
What are the most common reasons startups fail?
CB Insights' 'The Top Reasons Startups Fail' report, which analyzed post-mortems from over 110 startup failures, consistently identifies no market need as the single most common reason — cited in 42% of cases. Running out of cash is second (29%), followed by not having the right team (23%), getting outcompeted (19%), and pricing or cost issues (18%). The Startup Genome Project's research adds important nuance: premature scaling — expanding operations, hiring, and spending before achieving product-market fit — is the single most common pattern in startup failure, underlying many of the specific reasons listed above. Their analysis found that companies that scale prematurely are 2.3x more likely to fail than those that scale after reaching product-market fit.
Do VC-backed startups fail more or less than bootstrapped startups?
VC-backed and bootstrapped startups fail for different reasons and at different rates, making direct comparison complex. Venture-funded startups have higher absolute survival rates in their first 1-3 years because funding provides runway. However, VC-backed startups face a higher-stakes definition of 'failure': a company that generates modest sustainable profits and employs 20 people indefinitely would be considered a failure by venture capital standards (insufficient returns), while bootstrapped founders might consider it a success. Data from First Round Capital's analysis of its own portfolio and published academic research suggests approximately 10-15% of VC-funded startups achieve the strong exits (IPO or acquisition over 10x invested capital) that produce meaningful venture returns. Most of the remainder return some capital, shut down, or operate indefinitely at sub-scale.
Which industries have the highest and lowest startup failure rates?
Industry significantly affects startup survival. BLS data shows that construction, transportation, and retail have the highest 5-year failure rates, consistently above 50%. Information technology and professional services startups show slightly better survival rates, approximately 40% failure within 5 years. Healthcare and pharmaceutical startups have high capital requirements but benefit from regulatory barriers to competition once established — successful healthtech companies show high retention. Restaurant startups are frequently cited as having very high failure rates; the actual data puts approximately 60% of independent restaurants out of business within 5 years, though this is lower than the 90% figure often cited in food media. SaaS (software as a service) startups with established recurring revenue show significantly better survival rates than consumer app startups, where hit-driven economics dominate.
What factors most predict startup survival?
Multiple studies identify several consistent predictors of startup survival. Founder experience is the strongest single predictor: second-time founders succeed at approximately 30% higher rates than first-time founders with no prior startup experience, per research published in Harvard Business Review. Market size and timing matter significantly — companies addressing markets that are growing, rather than trying to displace entrenched players in flat markets, show higher survival rates. Team composition predicts success more reliably than idea quality: startups with complementary co-founders (technical plus commercial) outperform solo founders and single-skill founding teams. Profitability orientation — defined as reaching breakeven or near-breakeven before raising external capital — predicts survival better than the amount of capital raised, per Startup Genome Project data.