On October 4, 2010, at 10:43 in the morning, Evan Williams was vomiting into a wastebasket minutes before telling 300 Twitter employees he was stepping down as chief executive. The blog post said he chose to hand power to Dick Costolo. He had not chosen. Per Nick Bilton’s account in Hatching Twitter, Williams was forced out in a boardroom coup run by investors and people he had considered friends, among them Jack Dorsey, the first CEO, whom Williams himself had pushed out two years earlier. That single scene contains almost everything worth knowing about cofounder breakups: the friendship, the betrayal, the rotation of who holds the knife, and the fact that none of it was a surprise to anyone paying attention.

Founders tend to assume their own relationship is the exception. The public record says otherwise, and the record is far richer than most people building companies realize.

The data nobody wants to be part of

The canonical source is Noam Wasserman, the Harvard Business School professor whose book The Founder’s Dilemmas drew on quantitative data covering almost 10,000 founders, published by Princeton University Press in 2012. Per the HBS abstract, “people problems are the leading cause of failure in startups.” The figure that gets quoted, sometimes loosely, is that 65 percent of high-potential startups fail due to conflict among cofounders. As the firm Servanda notes in summarizing the research, the precise reading is that of startups failing at the founding-team level, 65 percent do so because of interpersonal dynamics rather than bad products or weak markets.

Two of Wasserman’s other findings are sharper and less cited. Per a widely-circulated list of statistics from the book compiled by Dharmesh Shah, in 73 percent of founder-CEO replacements the founder was fired rather than voluntarily stepping down, and founding-team turnover increases dramatically when the startup raises its first round of financing. The chances of a founder-CEO being replaced rise with each new round. The money that saves the company is the same money that destabilizes the people who started it. The HBS abstract draws on named cases including Evan Williams of Twitter and Tim Westergren of Pandora, so the data and the anecdotes describe the same phenomenon from two angles.

The four documented shapes

Read enough public accounts and the breakups sort into a small number of shapes.

The first is the erased cofounder, written out of the origin story early, before the value exists. Twitter’s clearest casualty was not Williams or Dorsey but Noah Glass, who, per Bilton’s account, persuaded Williams to invest in the podcasting startup Odeo that became Twitter’s vessel, and was pushed out so completely that most people do not know his name. He was a cofounder by contribution and a non-person by the time it mattered.

The second is the idea-credit dispute, where someone present at the creation claims they were cut out before the cap table was set. Snap is the textbook case. Frank “Reggie” Brown, a Stanford fraternity brother of Evan Spiegel and Bobby Murphy, claimed he came up with the disappearing-photo concept in 2011 and was pushed out, per CNET’s reporting, seeking a one-third stake. He filed suit in February 2013. The settlement, disclosed only later in Snap’s S-1, was $157.5 million, per TechCrunch, with $50 million paid in 2014 and $107.5 million in 2016. Notably, Snap publicly conceded, per BuzzFeed News, that Brown “originally came up with the idea” while at Stanford, an acknowledgment Mark Zuckerberg never gave the Winklevoss twins.

The third is the founder-CEO ouster, the Williams-Dorsey rotation, where control passes from one cofounder to another and then away from both. This is the shape Wasserman’s 73-percent figure describes, and it is the one most tied to financing events.

The fourth is the slow vision divergence, the breakup with no villain, where two people who finished each other’s sentences end up, in Servanda’s description, “arguing about equity splits over email” while the product stops shipping. There is no lawsuit and no coup. There is just two people who wanted different companies and never said so out loud.

The honeymoon-to-year-two curve

The timing is consistent enough to plan around. The breakups rarely happen in the first months, when the relationship “feels good” and the hard conversations get skipped precisely because skipping them is pleasant. Servanda’s framing is that most cofounders avoid the difficult discussions early, then “discover irreconcilable differences under pressure.” The pressure arrives on a schedule: the first financing round, the first real hire who reports to one founder and not the other, the first strategic fork where the company can only go one way.

Tom Blomfield, who has cofounded multiple startups, puts the stakes plainly in his rules for startups: “If your startup is successful, you’ll spend the majority of your waking time with these people for the next 5 years. Your relationship will be like a marriage, but even more intense.” The marriage metaphor is overused, but the timeline it implies is right. The disputes that end companies tend to surface around the same point a marriage discovers whether it was built on more than the wedding.

The money that saves the company is the same money that destabilizes the people who started it.

The boring instruments are the ones that work, and they all exist to handle the breakup before anyone wants to think about a breakup.

Vesting is the central one. The standard structure, per a practical guide from Promise Legal, is four-year vesting with a one-year cliff: nothing vests until month 12, when 25 percent vests in a lump, and the remaining 75 percent vests monthly over the next 36 months. For founders this is typically done through restricted stock with reverse vesting, which is what protects the company from the cofounder who leaves in month eight and walks away with a quarter of the equity for a year of half-effort. The Reggie Brown outcome, a nine-figure settlement, is partly a story about what happens when contribution and ownership were never papered cleanly at the start.

Written agreements from day one are the other half. Blomfield’s rule is that all deals go into writing immediately, covering money invested, equity split, and salaries including deferred compensation, and that “there are plenty of open-source co-founder agreements published by law firms.” He quotes an investor’s maxim he came to respect: “I won’t ask you to trust me.” The point of writing it down is not distrust. It is that memory diverges, and the version each founder remembers is the version that favors them.

The equity split itself sits oddly outside the usual frameworks. Paul Graham’s famous Equity Equation, per a breakdown of the essay, gives founders a clean formula for trading equity for money or talent, 1/(1 - n), but says nothing about how cofounders should divide ownership among themselves. That silence is the tell. The hardest equity decision is the one the best framework in startups declines to model, because the split is a judgment, not a calculation: a bet on who will still be carrying weight in year four, made in year zero when everyone is carrying weight equally.

The signals that precede the blowup

The accounts agree that the warning signs are visible months ahead. Servanda’s read names three triggers that compound each other: vision misalignment, financial disputes, and unequal effort. Unequal effort is usually the first to show and the last to be discussed, because raising it feels like an accusation. One founder starts logging the other’s late starts and early exits. Nobody says anything. The resentment compounds quietly, exactly the way the Twitter friendships did before October 2010.

The structural signals are easier to read than the emotional ones. Turnover risk spikes at the first financing round, per Wasserman. The arrival of a board changes who can fire whom. A cofounder who stops showing up to the hard meetings has often already left in every way except the cap table. These are not subtle once you know the curve, and the founders who survive their cofounder relationships are mostly the ones who treated the boring paperwork and the uncomfortable early conversations as the actual work rather than the friction around it.

The thing the public record makes plain is that the breakup is rarely about the moment it happens. Williams pushed out Dorsey, then was pushed out himself by the same mechanism two years later, in front of 300 people, into a wastebasket. The instruments that would have made each exit cleaner, vesting schedules and written terms, were cheap and available the whole time. The cost of skipping them shows up years later, denominated in lawsuits, erased names, and the specific kind of betrayal that only happens between people who were once close enough to finish each other’s sentences.