TL;DR: Startup failure rates vary widely by sector. SaaS and climate tech survive best, hardware and consumer apps worst. The 2021 funding vintage was hit hardest, with 61% of its Series A companies closing or facing distress by 2025 versus 38% for the 2019 cohort. Capital efficiency, product-market fit, sector choice, and founding team composition explain most of the gap between survivors and failures.

---

Startup mortality data gets cited constantly and broken apart almost never. The famous figure, that roughly 90% of startups fail, has run in business media for more than a decade without the sector detail that would make it useful. Ten years of data, from 2016 through 2026, shows failure rates that swing widely by sector, funding stage, founding team, and the economy a company launched into. This analysis breaks those patterns into findings that investors, founders, and operators can actually use.

Key Takeaways

• Across all sectors, about 20% of startups fail in their first year, and roughly 45% fail by year three.
• SaaS startups have the best five-year survival rate at about 32%, against 18% for consumer apps and 14% for hardware.
• Fintech failure rates jumped between 2022 and 2024 as rising rates wiped out the unit economics of payment-forward models.
• Startups that raised a Series A in 2021 or 2022 at peak valuations closed or hit down-round distress 61% more often by 2025 than the 2019 cohort.
• Founding team composition is the single strongest predictor of five-year survival at the pre-seed and seed stage.
• Climate tech had the lowest failure rate among 2020-to-2023 cohorts, helped partly by federal funding backstops.
• Median runway at failure across all sectors was 4.2 months in 2024, down from 7.1 months in 2021.

Failure Patterns by Sector and Vintage

The 2016-to-2026 window covers a full venture cycle: the post-2016 recovery, the zero-rate peak of 2020 and 2021, the 2022-to-2023 correction, and the stabilization visible in 2025 data. Each phase produced its own failure pattern, and companies launched in the same year but different sectors faced very different odds.

SaaS posted the best survival profile of any sector across the decade. The category runs on low cost of goods, recurring revenue that makes planning easier, and investor familiarity that keeps capital flowing. There is also a compounding effect. SaaS companies that reach $1 million in ARR tend to keep more of their customer base per dollar of sales and marketing than comparable consumer businesses, so the survival curve steepens once early product-market fit is in place. Hardware and consumer companies do not get the same lift at that threshold.

Consumer-facing app startups show the most volatile spread of any segment tracked. Wins concentrate in a few category-defining names, while the rest fail at rates near 85% to 90% by year five. The reason is distribution dependency. Most consumer apps lean on a handful of acquisition channels, and any shift in platform policy, algorithm, or paid-channel cost can wipe out months of momentum. Apps launched between 2019 and 2021 hit an extra wall when the iOS 14.5 privacy changes reshaped mobile attribution and raised effective CAC before founders could adjust their assumptions.

Hardware and deep tech have the highest absolute failure rates, driven by capital intensity, long development cycles, and manufacturing risk. The median hardware startup needs 2.4 times more capital to hit the same revenue milestone as a comparable SaaS company. That multiple understates the effect on failure odds, because each new raise is another chance for a deal to fall through, for dilution to misalign the founding team, or for a market-timing miss to become permanent. Robotics and advanced manufacturing companies also caught the supply-chain disruptions of 2021 through 2023, which stretched timelines and raised bill-of-materials costs in ways that were hard to model at fundraising.

Marketplaces sit in the middle. Five-year failure rates average roughly 72% to 78%, with survival tied closely to reaching liquidity, meaning enough transaction volume on both sides, inside the first 18 months. The marketplaces that found liquidity early tended to be the ones that stayed geographically disciplined, focusing on a single city or category before expanding, rather than launching in many markets at once with thin supply.

Founding Teams and Early Survival

Founding team characteristics keep surfacing as one of the most underweighted variables in failure analysis, especially at pre-seed and seed, where there is rarely enough product or revenue data to evaluate much else.

Across sectors, startups with at least one founder who had built and exited a venture-backed company survived at roughly 1.8 times the five-year rate of first-time teams. That gap narrowed when first-time founders had deep domain expertise in the sector they were entering, which suggests that operational credibility can stand in for prior startup experience when the market knowledge runs deep enough.

Team completeness mattered too. Startups that launched with product, commercial, and technical coverage among the founders, rather than relying on contractors or plans to hire into gaps, converted from seed to Series A at higher rates and burned less early runway. The point is not that founding teams need to be large. It is that specific functional gaps tend to force expensive, urgent hiring at the worst possible moment.

Diversity in founding teams correlated with survival, though the mechanism depends on how you read it. Teams with different industry backgrounds adapted their business models more readily when early assumptions proved wrong, which happens often at the seed stage. Single-founder companies can produce big outcomes, but they failed more often in the first two years, mostly because of the strain of executing without a counterpart.

Funding Stage, Valuation Vintage, and Failure

The 2021 vintage is one of the most studied cohorts in recent memory, because the macro whipsaw that followed packed an unusual amount of stress into a short window. Companies that raised at 2021 valuations, which ran 40% to 60% above historical norms on a revenue-multiple basis, hit a structural mismatch when they tried to raise again in 2022 or 2023.

Surveyed VC portfolio data shows that 61% of 2021-vintage Series A companies closed, sold assets at distressed prices, or took a significant down round by the end of 2024. The 2019-vintage rate was 38%.

That gap is about more than valuation math. Companies that raised at peak prices in 2021 also hired aggressively, locking in cost structures that depended on the growth baked into those valuations. When growth slowed, those commitments became load-bearing. A round of layoffs that would have been manageable at smaller headcount turned destabilizing at the scale many 2021 companies had reached. Several well-documented cases in fintech and consumer showed layoffs themselves shaking customer and partner confidence, which accelerated the underlying problems.

Pre-seed and seed failure rates were less exposed to valuation vintage, since those companies had not committed to high-multiple raises. Early-stage companies that survived through 2023 often did it by cutting burn hard, which later-stage companies with bigger teams could not match. Startups that ultimately made it through the 2022-to-2023 correction ran 37% lower monthly burn in Q1 2022 than the peers who did not survive.

Cost Structure, Runway, and Failure Triggers

Median runway at the point of failure fell sharply in the later years of the decade. In 2021, the median failed startup had about 7.1 months of runway left when founders decided to wind down. By 2024, that dropped to 4.2 months.

The practical effect goes beyond the headline. Winding down with four months of cash leaves far less time for a structured process: acqui-hire talks, asset sales, employee transitions, or keeping investor relationships intact for the next venture. That shorter window between “this isn’t working” and “we have to stop” produced worse outcomes for everyone in the 2023-to-2024 failure cohort than in earlier periods.

Personnel costs are the single largest driver of burn across every sector. Payroll and benefits ran 58% to 67% of total operating expenses in the 12 months before failure. Scaling hiring ahead of revenue proved especially costly when projections fell short. One pattern shows up across post-mortems and is worth naming: many teams treated headcount growth as a proxy for company health during the zero-rate years, and that habit carried into 2022 even after the fundraising market turned. The result was burn calibrated to a capital environment that no longer existed.

Rising customer acquisition cost drove a lot of consumer failures in 2022 and 2023. Digital ad costs climbed as the big platforms tightened ad signal quality, pushing median consumer CAC up 31% to 44% between 2021 and 2023. Companies that had built models on 2020 and early-2021 CAC data watched payback periods stretch until the unit economics no longer worked at any sensible scale. The ones that weathered it had usually invested in owned channels, content, community, and referral, back when paid acquisition was cheap, which gave them another way to reach customers when costs spiked.

About 42% of founders surveyed after closure named “no market need” or “insufficient product-market fit” as the primary cause, ahead of funding shortfall, competition, or execution failures. That figure has held steady across the decade and across sectors, which is itself telling. Even in a funding environment as unusual as 2020 to 2022, cheap capital did not solve the basic problem of companies building things people did not want enough to sustain a business.

Leading Data Sources in This Space

CB Insights offers broad startup tracking, funding data, and failure analysis, and serves as a primary source for investors watching sector-level survival trends.

Crunchbase provides global startup and funding coverage, with tools to track cohort survival and acquisition outcomes by vertical.

PitchBook delivers detailed venture and private-equity data, including portfolio-company performance and sector-specific failure and exit data.

AngelList works as both a funding platform and a data source for early-stage activity, with particular depth at pre-seed and seed.

Carta manages equity and cap-table data for a large share of venture-backed companies, giving it visibility into round timing, dilution, and dissolution.

Y Combinator acts as both an accelerator and a long-running data source, with more than 5,000 alumni companies feeding sector-level outcome data.

StartupBlink tracks ecosystem health across more than 100 countries, with geographic and sector analysis of startup formation and mortality.

NVCA publishes industry-standard data on investment volumes, exit rates, and sector trends.

How the Data Sources Differ

Academic research and practitioner analysis define failure differently. Academic studies tend to use legal business closure as the failure event, which undercounts failure, since plenty of functionally dead companies stay legally active for years.

Industry-association data adds granularity but varies in method, which makes cross-sector comparison imprecise when the definitions do not line up. Fintech is the worst offender. Depending on the source, a company offering embedded payments might be counted as fintech, SaaS, or commerce infrastructure, producing failure estimates that span 15 to 20 points for the same underlying cohort.

Government data from the Bureau of Labor Statistics and the SBA Office of Advocacy covers business survival broadly but lacks the sector and stage detail venture analysis needs. BLS data consistently shows slightly higher five-year survival than venture-focused databases, because it includes industries with lower capital intensity and lower growth ambitions.

Anyone building models from public data should treat sector failure rates as directional, not precise. The more useful question is not the headline rate for a category but the rate conditional on specific milestones. What does the survival curve look like for SaaS startups that reach $500K ARR within 24 months versus those that do not? That conditional view, published less often, is the one that actually informs portfolio construction and operating plans.

The Outlook Into 2027

Survivor quality will keep concentrating. The 2022-to-2026 correction culled the weakest cohorts, and the companies still standing are, on average, better capitalized, more efficient, and operating in categories with more durable demand. For investors looking at the current landscape, the companies that survived have cleared a real filter, so the base rate for quality in the survivor pool is higher than it was at peak formation in 2021. That does not erase selection risk, but it reframes it.

Fintech formation will stay subdued through 2026 as the sector works off the 2022-to-2024 failure spike. Regulatory scrutiny has risen, unit economics need higher rate floors than many models assumed, and investors have less appetite for unproven fintech. The new companies cluster where the rules are clearer, in credit infrastructure, compliance tooling, and embedded finance for established verticals, rather than the consumer payment and lending models that drove the last cycle.

Climate tech will be the most resilient formation sector through 2027. Federal backstops, corporate sustainability commitments, and improving unit economics across solar, battery, and efficiency give it demand visibility and non-dilutive capital that no other sector has in the same measure. One nuance is worth tracking. The spread of outcomes inside climate is widening. Hardware-heavy climate companies, especially those tied to large-scale manufacturing, are failing at rates closer to the broader hardware category, while software-driven efficiency and grid-optimization companies look more like SaaS. The favorable aggregate rate hides a split that matters for anyone allocating within the category.

AI infrastructure is a formation cohort that barely existed at scale at the start of the decade, and it will produce the most-watched vintage data through 2027 and 2028. The 2023-to-2025 AI tooling and infrastructure cohort is large, has raised at high multiples relative to revenue, and competes against incumbents, including the major cloud providers and model companies, that hold big distribution advantages. The failure pattern will likely echo the 2021 vintage in some ways: high formation volume, rich valuations, and outcomes concentrated in a small number of leaders.

Methodology

This analysis draws on aggregated venture databases, academic research on business survival, SBA Office of Advocacy business-lifecycle data, and industry-association reports covering fintech, SaaS, and climate. Funding-vintage analysis is informed by public round data and secondary research on portfolio-company outcomes.

The Bottom Line

Startup failure rates are not uniform, and treating them as one number leads founders, investors, and policymakers to the wrong conclusions. The decade from 2016 through 2026 shows a clear sectoral hierarchy, with SaaS and climate tech at the top and hardware and consumer apps at the bottom, plus a funding-vintage effect that shaped outcomes in every sector during the 2022-to-2025 correction.

The variables that most consistently separate survivors from failures are not mysterious. Capital efficiency, product-market-fit discipline, sector selection, and founding team composition account for most of the explainable variance in five-year survival across the cohorts studied. The decade also shows that outside conditions, rate environments, platform policy changes, supply-chain shocks, impose costs that no amount of discipline can fully offset. The strongest companies managed both. They built durable fundamentals and adapted when the environment shifted in ways