Why startups fail: lessons from 26 years of Harvard research

Executive overview

Most startup failures are avoidable — they stem from predictable mistakes, not bad luck. Tom Eisenmann's research at Harvard Business School identifies three recurring failure patterns that founders repeat without recognising them.

The antidote is thinking like a scientist: form hypotheses, test assumptions quickly, and treat failure as data rather than personal verdict.

Customer discovery done before building is worth ten times the time it saves.

The false start: building before understanding

• Founders skip customer discovery to start building — the most common early-stage failure.
• Lean startup is misread as "build fast"; the overlooked half is deep customer discovery first.
• Skipping four weeks of discovery leads to four months of wasted building and a painful pivot.
• Non-technical founders hire engineers, then feel pressure to keep them busy — so they build the wrong thing.
• Technical founders fall into the same trap: engineers love to build, so they build before the problem is fully understood.
• The fix: talk to potential customers before writing a line of code.

Bedfellows: relationship failures inside the venture

• Bad relationships between co-founders, team members, or investors are a second major failure pattern.
• Founders choose their co-founders and investors — responsibility for those relationships sits with them.
• Blaming partners without self-reflection is a sign this pattern is at work.

False positives: misleading early signals

• Early customers signal strong demand, but their needs can differ sharply from mainstream customers.
• Dropbox's first adopters were software engineers with complex file needs — not the mainstream user the founder originally envisioned. He built for mainstream and it worked, but that choice isn't always right.
• When doing customer discovery, talk to both early adopters and mainstream customers; identify whether their needs diverge.
• If needs differ, decide explicitly: build for early adopters and evolve, or build for mainstream and accept early-adopter risk.

Testing assumptions to catch mistakes early

• Maintain a written list of your core assumptions: customer needs, solution appeal, acquisition path.
• Your job is to test those assumptions continuously — confirm, pivot, or cut.
• Some entrepreneurs change direction too fast (zigging and zagging); others are too stubborn. Neither extreme works.
• Once the business is large enough, run controlled tests. Until then, rely on structured customer conversations.

Learning from failure without distortion

• Fundamental attribution error: we blame others for our failures, but attribute others' failures to their character.
• Founders who fail often blame co-founders, investors, regulators, or the market — ignoring their own role.
• The stubborn, blame-deflecting founder is the more dangerous type: they repeat mistakes in new contexts.
• Founders who over-blame themselves often quit entirely — also a loss, because one failure doesn't define ability.
• The right approach: let emotions settle, alternate between distance and reflection, then own your specific mistakes.

First-time founder success rates

• First-time founder success rate (financial return for investors): 21%.
• Second-time founders who succeeded first: 30% success rate.
• First-time founders who failed, then tried again: 22% success rate — nearly identical to first-timers.
• The lesson: failure does not meaningfully lower your odds next time if you learn from it.
• Serial entrepreneurship over a career with a 70–80% failure rate per venture almost guarantees at least one failure.