Survival rates: what the data actually shows

"90% of startups fail," "92% of SaaS companies fail within three years," "95% fail within five years" — these figures, and a dozen close variants, appear across an enormous share of SaaS and startup content, routinely attributed to "Harvard Business Review," "a Startup Genome study," or no source at all. This lesson runs the same check this course has applied to every other claim in it, and the result is worth stating plainly: almost none of these specific figures trace to a real, disclosed, current methodology, and the sources that would give you a genuinely reliable answer say something less dramatic, and less SaaS-specific, than the number that usually circulates.

#Why the popular figures don't hold up

The "90% fail" figure most commonly gets attributed to Harvard Business Review, but no HBR-published study stating that exact figure, with a disclosed methodology, could be traced in this research. The 2011 Startup Genome Report — a real, published analysis of over 3,200 high-growth technology startups — is the source some versions trace back to, but that report studied a self-selected sample of high-growth technology startups specifically opted into that study's own program, not a representative sample of all SaaS or software startups, and being unrepresentative in exactly that way isn't a small caveat — it's the difference between "startups broadly" and "one specific, non-random cohort a decade-plus ago." [Speculative] — this course treats every specific percentage in this genre (90%, 92%, 95%, and the many other close variants) as unverified until proven otherwise, precisely because the same figure keeps reappearing attributed to different, vaguely-cited, or entirely absent sources — the classic pattern of a number that's been repeated so many times it's acquired the texture of fact without ever having a traceable origin.

This matters beyond pedantry: a founder deciding whether to start a SaaS business, or when to give up on one, deserves a real number to weigh the decision against — not a repeated-until-authoritative one that happens to be dramatic enough to generate content-marketing clicks.

#What's actually known: government business-survival data, not SaaS-specific, but real

The US Bureau of Labor Statistics tracks survival of every private-sector business establishment in the country through its Business Employment Dynamics program, using unemployment-insurance tax records rather than self-reported survey data — this is about as close to primary, disclosed, methodologically-transparent data as exists on business survival. [Established]. The most recent published cohort data: of private-sector establishments that opened in 2013, 34.7% were still operating ten years later, in 2023; roughly 78% survive their first year, and just over half (about 51%) survive five years, across all industries combined.

The Information sector — the broadest government classification that includes software and technology businesses, alongside publishing, telecom, and media — has historically shown among the lowest survival rates of any sector BLS tracks, in cohort studies going back to the late 1990s. [Directional] — this pattern (Information sector underperforming most other sectors on survival) shows up consistently across multiple BLS cohort studies spanning more than two decades, which is real corroboration, but the specific historical figures available for that sector (drawn from a cohort that opened in the late 1990s) are now dated enough that citing an exact current percentage for the sector specifically would overstate precision this research doesn't have; treat "the Information sector runs below the all-industry average" as the solid, repeated finding, and any specific percentage for that sector alone as approximate.

The honest gap this course won't paper over: BLS's Information sector classification is far broader than "SaaS company" — it includes newspapers, broadcasters, and telecom providers alongside software firms, none of which share SaaS's specific unit-economics or distribution dynamics — and no government or other primary-tier source this research could find publishes a disclosed, methodologically-transparent survival rate specifically for subscription software businesses. The precise "SaaS startup failure rate" you're looking for does not appear to exist as a traceable, disclosed number, at any tier. Anyone stating one with confidence is either citing something this research couldn't verify, or repeating a number that's drifted loose from any real source.

#The more useful reframe

Given that gap, the kill-switch framework in the previous lesson is a more honest tool than any borrowed survival percentage: it doesn't need a base rate for "SaaS companies in general" to be useful, because it evaluates your specific business against mechanisms (distribution, unit economics, retention, capital runway) that are true regardless of what percentage of other companies clear them. A precise industry base rate would be a nice-to-have for calibrating expectations going in — and the honest answer is that all-industry BLS data (roughly half of new businesses surviving five years, across every sector, not SaaS specifically) is the most defensible anchor available, adjusted downward somewhat for the historical pattern of the Information sector underperforming that average, and adjusted again for the fact that "survival" in BLS data means the establishment still exists, not that it's a good business anyone would want to have built. That's a real, if imprecise, answer — a meaningfully more honest one than a specific, dramatic, untraceable percentage borrowed from content that never checked where it came from.