Why Startups Fail: A New Roadmap for Entrepreneurial Success

By Tom Eisenmann

10 min read

Why does it matter? Because the advice that makes great founders is also what gets them killed.

The conventional story about startup failure goes one of two ways: bad luck, or the wrong founder. Both feel complete. Neither is much use if you're trying to not be next.

Tom Eisenmann spent twenty-four years teaching entrepreneurship at Harvard. He mentored thousands of founders, catalogued the failure patterns of hundreds of startups, and personally invested in two that failed. That last part mattered — it turned academic interest into something more uncomfortable: a genuine need to understand how failure actually works, not the myth or the postmortem spin, but the mechanism.

What he found: failure isn't random, and it isn't usually incompetence. It follows six recognizable patterns. Almost all of them are triggered by founders doing exactly what the playbook says — moving fast, staying passionate, scaling hard. The advice isn't wrong. It's just that at certain moments, in certain doses, it becomes the thing that kills you.

Most Startup Postmortems Are Wrong Before They Begin

Failure is a pattern you can recognize in advance, not a verdict on character delivered after the fact. That reframe is what Tom Eisenmann spent years building, starting from a problem he couldn't solve about his own investments.

He encouraged and personally invested in two student ventures. Both failed. What shook him was that he couldn't explain why — not in a way that would help the next founder avoid the same fate.

Every startup postmortem makes the same two moves: find the fatal flaw, find who's responsible. Both are wrong. When you try to explain why a startup died, two cognitive errors corrupt the analysis before it begins. The first is the single-cause fallacy: we spotlight one reason (bad timing, the wrong hire, a competitor's move) when outcomes almost always have multiple drivers. The second is the fundamental attribution error: outside observers blame the founder's character ("he was too stubborn"), while the founder blames circumstances ("the market shifted"). Neither account is reliable, and both distort whatever lesson you thought you extracted.

So most startup postmortems produce confident explanations that are probably wrong in the ways that matter. What came back, across hundreds of founders and two decades of case work, were six recurring patterns: documentable, largely avoidable, and spread across a startup's early and late stages.

Half the Lean Startup Method Is Worse Than None of It

What if the Lean Startup method, applied diligently, is exactly what kills a startup?

Sunil Nagaraj did everything a Lean Startup founder is supposed to do. He launched a minimum viable product, got it in front of real users, watched what worked, cut what didn't, and pivoted. Three times in fourteen months. By the end he'd burned $630,000 of a $750,000 seed round and shut the company down without ever finding a market. The more useful measure of runway isn't months of cash remaining — it's the number of pivots possible before the money runs out. A false start consumes exactly the capital that would have funded the correction. The mistake and the cost of fixing it come out of the same account.

The irony: Nagaraj wasn't failing because he ignored the playbook. He was failing because he followed half of it.

Triangulate started as a matching engine, software to predict romantic compatibility from users' digital behavior rather than self-reported questionnaires. His first pivot was Wings, a Facebook dating site where friends could vouch for singles as wingmen. It reached 35,000 users, but almost none paid, viral growth ran at 0.03 new users per user against a projected 0.8, and the wingman feature mattered far less to daters than it did to Nagaraj. He dropped it. Second pivot: strip out the matching engine nobody valued. Third: DateBuzz, which let people vote on text and interests before seeing photos. Each pivot felt like progress. It was the mechanism of failure.

What went wrong? Nagaraj never seriously studied what online daters wanted before he started building. He ran a 150-person survey — but a survey isn't customer discovery. He never sat across from someone and asked what frustrated them about dating online, or whether they'd want a friend vouching for them. "In retrospect," he said, "I should have spent a few months talking to as many customers as possible before we started to code."

Lean Startup has two steps: customer discovery (understand the problem before designing anything), then MVP testing (put a prototype in front of real people and watch). Founders under pressure to ship, especially technical ones, tend to skip step one and treat step two as the whole method. The result is an MVP machine that iterates efficiently but never asks whether it's solving the right problem. "Ready, fire, aim" isn't a corruption of Lean Startup. For many founders, it's what Lean Startup feels like from the inside.

Your Most Dangerous Signal Is an Unusually Good Start

The winter Baroo launched pet care services at Ink Block, a new luxury apartment building in Boston's South End, 70 percent of the building's pet owners signed up. Lindsay Hyde, Baroo's founder, was thrilled. She took it as proof that her market existed everywhere.

It didn't. Three anomalies had combined to produce that number, and none of them would travel.

Ink Block was brand-new; every resident had moved in within weeks of each other, which meant none of them had a dog walker yet. Switching costs in pet care run high (a walker learns your dog, your schedule, your household routines), but these residents had nothing to switch from. Hyde later admitted the pre-launch survey asked the wrong question: would you use a nearby pet service, not would you replace your current one. Meanwhile, a Hollywood film crew had rented a block of units while shooting nearby; they had pets, generous daily stipends, and no time to walk anyone. And that particular winter, Boston received 94 inches of snow in 30 days, a record. Nobody wanted to be outside. Baroo was doing multiple walks per day for households that normally needed one.

Hyde's takeaway: "If we could operate through that winter, we could do anything." Four cities and $4.5 million later, Baroo shut down in February 2018.

Eisenmann calls this a false positive: adoption numbers that look like product-market fit but are actually local conditions wearing its clothes. What makes it dangerous isn't that founders are careless. It's that the signal is real. Baroo genuinely had demand. Hyde wasn't imagining it. The false positive is seductive precisely because something true is happening — just not the thing you think.

The antidote is one uncomfortable question, asked at the exact moment things look best: What is unusual about these specific customers? If the answer involves a snowstorm, a film crew, or a building where everyone moved in last month, the number needs discounting — not scaling.

The Investors Most Excited About Your Growth Are the Ones Most Likely to Destroy You

Jason Goldberg was fielding congratulations when he felt sick.

In June 2013, his online design retailer Fab.com had just closed $165 million in venture capital at a $1 billion valuation: unicorn status, the benchmark every founder tracks. His phone kept lighting up. He answered, said the right things, and spent the whole time knowing something none of the callers knew: he needed $300 million to execute his plan. He'd raised just over half.

Three months later, in an internal memo to his team, Goldberg wrote: "We spent $200 million and we haven't proven out our business model."

That sentence contains the whole failure. Fab's pre-launch cohort of design fanatics bought frequently, referred friends, and cost almost nothing to acquire. Every subsequent wave of paid-acquisition customers performed measurably worse: lower repurchase rates, higher acquisition costs, less enthusiasm for the products. Goldberg ran the cohort analysis and watched the numbers degrade in real time. Senior managers had recommended cutting European operations in half months before the final reckoning; he waited until after the fundraise closed. By October 2013, Fab was burning $14 million a month. The company laid off 80 percent of its U.S. staff. Its assets eventually sold for roughly $30 million in stock.

The board had voted almost unanimously for aggressive expansion. Only one member argued for pulling back to the U.S. market at a sustainable pace. This wasn't delusion — it was the logic of the capital structure. Late-round investors had paid high prices for their equity. For them to earn a return, Fab had to keep growing. Patient, profitable consolidation would have worked for early investors, who'd bought in cheap. Late investors needed the rocket ship.

Venture capitalist Fred Wilson estimates that two-thirds of the startup failures he's observed came from the same mechanism: overfunding a promising idea before the underlying business model was resolved. The higher the price paid to get in, the more they need the company to grow its way out of a valuation that only makes sense at scale.

Eisenmann frames this as structural, not personal. Goldberg wasn't reckless; he was responding rationally to a board responding rationally to its own incentives. The dangerous signal isn't investor alarm — it's investor enthusiasm. The founder who satisfies late-stage investors is often the one who destroys the company. Goldberg had the cohort data. He watched each wave perform worse than the last and kept growing. The survival path runs through that signal: use it to set the growth rate, not the other way around.

Moonshots Aren't a Character Test — They're a Probability Calculation

Think of it as betting that a coin lands heads five times in a row. Even at coin-flip odds for each — and for some of these breakthroughs, 50 percent is generous — the probability is about 3 percent. That's not bad execution. That's not failure of vision. That's a roulette table with different decorations.

Eisenmann calls this the Cascading Miracles problem: a category of startup where every required breakthrough is genuinely plausible, but the product of their odds isn't. Better Place is the case study.

Shai Agassi founded Better Place in 2007 to blanket Israel and Denmark with electric vehicle charging infrastructure — charge spots in parking lots, robotic stations that swapped depleted batteries in five minutes. He raised $900 million, signed a deal with Renault-Nissan, and got letters of intent from 400 Israeli corporations to electrify their fleets. When the company declared bankruptcy in 2013, it had sold roughly 1,000 cars across both countries.

The postmortem temptation is to blame Agassi — his rigidity, his management style, his refusal to listen. Eisenmann's analysis goes a different direction. Better Place needed five things to go right simultaneously: mainstream demand for electric cars, consumer preference for swappable batteries over fast charging, partnerships with multiple automakers, deployment across several national markets at once, and sustained investor support through years of capital-intensive scaling. Any single miss would destroy the company. Five coin flips at 50 percent each: 3 percent odds. Not a character problem. A structure problem.

Agassi was the kind of storyteller who fills a TED stage and makes $900 million feel like a reasonable bet. That same storytelling locked the company into a single narrative. When costs exploded, the gaps were not rounding errors. Charge spots projected at $200–$300 each came in at $2,500. Battery swap stations projected at $300,000–$500,000 cost over $2 million. The survey suggesting 400,000 buyers produced roughly 1,000 actual sales. Each was a pivot signal. Six years of selling the same vision made turning away almost impossible.

The charisma that attracts the capital is structurally opposed to the recalibration the capital requires. You don't raise $900 million by being the first to say the idea doesn't work.

Every Piece of Startup Advice You've Heard Has a Built-In Failure Mode

The six pieces of advice every first-time founder hears — just do it, be persistent, bring passion, grow fast, focus, be scrappy — are not wrong. They're worse than wrong: they're right in the wrong situations.

Eisenmann maps each one to a specific failure mode. "Just do it" before customer research produces false starts — founders build solutions before understanding the problem. "Be persistent" shaded into stubbornness delays the pivots those false starts demand. "Bring passion" tips into overconfidence that makes research feel unnecessary, amplifying false positives. "Grow fast" drives premature scaling before unit economics hold. "Focus" on early adopters means ignoring the mainstream customers who come next, which is exactly how strong early traction becomes a ceiling. "Be scrappy" means refusing to hire expensive specialists, which is how startups end up with the wrong team at the moment they most need the right one.

The pattern: these are heuristics designed for low stakes and time pressure. They work when mistakes are cheap and reversible. They become dangerous at inflection points (shifting from exploration to building, deciding whether to pivot, choosing investors) where mistakes are expensive and hard to undo.

Kahneman showed that the brain defaults to quick intuition for routine decisions and deliberate analysis for complex ones. The failure mode for founders is reaching for intuition exactly when the situation requires deliberate thought. Eisenmann's prescription follows: treat the maxims as defaults for ordinary decisions, and switch modes at the ones that can kill the company. Sleep on it two nights. Write out the options. Share the analysis with people who know the situation.

The Question Worth Asking Before You Step on the Gas

The six patterns Eisenmann documented don't begin with obvious mistakes. They begin with the obvious moves: ship the MVP, scale what's growing, take money from people who believe in you.

What Eisenmann is really asking you to do is learn to recognize the pattern before you're inside it. Most calls you make as a founder should be fast — speed is genuinely an advantage. But a handful of moments deserve something slower: when you shift from exploring to building, when the numbers look unexpectedly good, when the biggest check comes with the most urgency. Write it down. Sleep on it twice. Ask not just is this working but do I actually understand why.

The ambition stays. The heuristics stay. You just stop applying them equally to everything — because the decisions that kill startups look, at the moment you make them, exactly like the decisions that build them.