Why We Kill Ventures

Every serious studio has a graveyard. Ours will have one too. That is not a flaw in the model. It is the model.

The discipline of killing ventures is, in my experience, the most misunderstood part of building things. Founders hold on too long. Investors reward hold-on behaviour up to the point where it destroys them. Boards circulate slide decks explaining why the next quarter will be different. I have sat in many of these rooms. I have been the person circulating the deck. The regret curve is ugly, and the worst version of it is the regret you feel not because you failed but because you waited six months longer than you should have before admitting you already knew.

The sunk-cost trap has a specific shape

Most founders I know are intelligent, committed, and self-aware. And most of them, at some point, hold on to a venture past the point where the evidence argues for stopping. The mechanism is almost always the same. The venture has consumed enough of their identity that stopping it feels like admitting the wrong thing about themselves rather than the wrong thing about the market.

This is where single-founder psychology fails. And it is where a studio has a structural advantage, provided the studio is willing to use it.

A studio runs multiple bets. The institutional decision to stop any one bet is not an admission of personal failure. It is a reallocation of capital to the bets with stronger signal. The emotional weight distributes across a portfolio rather than collapsing onto a single founder. Done well, this produces better decisions earlier. Done badly, it produces a holding company that collects zombie ventures while the operators tell themselves they have an investment thesis.

A studio that cannot kill a venture is not a studio. It is a holding company with extra steps.

Why the standard prescriptions make me nervous

There is a version of stage-gate discipline, widely circulated in venture-studio circles, which proposes specific numerical thresholds for killing or continuing a venture. Twenty customer interviews. Five measured conversions. A particular conversion rate. A specific unit economics sketch. A named trigger at each stage.

I understand the appeal. Numbers feel defensible. Numbers remove judgment from a conversation that is otherwise soaked in emotion. Numbers let you point at a page rather than at a founder when the time comes to stop.

I am cautious about this approach, for two reasons.

The first reason is that the right evidence threshold depends on what you are testing. Twenty customer interviews might be reasonable evidence for a pre-launch consumer idea aimed at a broad, noisy market. It is almost certainly insufficient for a proposed behavioural shift inside a company with fifty thousand existing customers. In that second case, twenty interviews tell you about twenty people's reactions to a stimulus, not about what the system of the existing customer base will do when the shift is attempted at scale. Sometimes the only way to generate the relevant evidence is to build a small, limited version and deploy it. Sometimes you need many more than twenty conversations because the variance in the population is high. Sometimes you need fewer, because you are testing a proposition that ten sophisticated buyers either confirm or deny. Evidence threshold is not a universal constant. It is a property of the specific question.

The second reason is that prescriptive gates can produce a particular kind of false confidence. A team that passes the twenty-interview threshold congratulates itself and moves to the next gate, even if the interviews themselves were shallow, leading, or sampled from a biased pool. A team that fails the threshold kills a venture that might have been salvageable with a better-designed test. The gate becomes a ritual rather than a truth procedure.

Principles instead of thresholds

What I propose, instead of numbers, are principles. Four of them, each of which does more work than a gate number while requiring more judgment to apply.

Evidence proportional to ambition. The evidence you need scales with the consequence of being wrong. A low-cost test aimed at a narrow market needs less than a high-cost pivot aimed at a broad one. The question to ask is not "have we done twenty interviews" but "is the evidence we now have sufficient to justify the next increment of commitment, given the cost of being wrong." Usually the answer is no, and usually the fix is to design a smaller next step, not to demand a bigger dataset.

Test before you build, wherever the test exists. Most ideas can be tested, in some partial form, before they are built. A landing page. A scripted prototype. A concierge service. A conversation with ten buyers holding a written contract. The studio's job is to insist on the cheapest test that generates the relevant signal, even when the founders want to skip testing because building feels more productive. Building without testing is a way of transferring risk from your head onto your balance sheet.

Separate "is this a real problem" from "is my solution right." Most kill decisions conflate these two questions, and the conflation produces the wrong answer in both directions. A venture can have the wrong solution to a real problem, and the fix is to change the solution, not to kill the venture. A venture can have a plausible-looking solution to a problem that nobody has, and no amount of solution iteration will save it. The studio discipline is to test these two layers independently. It is surprising how rarely founders do this.

Build the muscle for stopping early. Stopping is a skill that improves with practice and atrophies with avoidance. A studio that has never killed a venture will not know how to kill one when the moment arrives. A studio that kills small bets quickly, and communicates the reasoning transparently, builds the organisational muscle to kill larger bets when the evidence warrants it. The kill is not the failure. The failure is the stubbornness that precedes the kill.

What killing looks like in practice

When we kill a venture, we do not simply close a file. We extract what the venture produced. The infrastructure code, if any, goes into the studio's reusable base. The customer research feeds into the next opportunity map. The operational playbook for what worked in go-to-market, in hiring, in procurement, gets distilled into notes. The relationships we formed with clients, advisors, and potential hires become part of the studio's expanding network, available for the next venture.

The venture stops. The learning does not.

This is the structural argument for the studio model. A solo founder who kills a venture loses most of the accumulated investment. A studio that kills a venture keeps the investment in the form of judgment and infrastructure that compounds into the next build. Over time, each kill improves the speed and quality of the subsequent builds. The pattern recognition gets sharper. The taste for which ideas deserve a next step gets more reliable. The graveyard becomes an asset.

The emotional contract

I will not pretend any of this is easy. Every kill decision involves real people who have committed real effort to an idea that mattered to them. The studio has to set the emotional contract from the first week. This venture might not pass the first honest test, and that is not a failure of the team. It is the process working as designed. The studios that cannot have that conversation openly end up with zombie ventures that consume resources, attention, and morale while the leadership pretends otherwise.

I have closed companies. I have watched colleagues hold on too long, out of conviction, loyalty, fear, or the simple inability to face the alternative. The regret curve is ugly, and I do not wish it on anyone.

The best studios are not the ones that never fail. They are the ones that fail early, learn quickly, and turn every stopped venture into a starting advantage for the next one.