Three ventures we killed — and why
The kill criteria that saved us eighteen months of building.
We publish the ones that died for the same reason we publish the ones that shipped: the reasoning is the asset. Three ventures reached the paid-prototype stage in 2025 and did not make it to incorporation. Here is what stopped each one.
One: the insurance claims copilot
The product worked and the adjusters loved it. The problem was that nobody could buy it — claims automation sat under a transformation budget that three directors jointly controlled, and every pilot renewal required all three to re-approve. We killed it at week fourteen after the second pilot took eleven weeks to convert a £9,000 renewal.
The criterion it failed was the one we now weight most heavily: a single budget holder who can say yes alone.
Two: the freelancer payments wedge
Real problem, real frequency, and a wedge we could build. It failed on distribution. Every acquisition path we modelled routed through platforms that were already building the feature in-house, and we had no privileged relationship with any of them.
"We can out-build an incumbent. We cannot out-distribute one from a standing start."
Three: the clinic scheduling marketplace
This one hurt, because the paid pilot closed in nine days. But the second and third customers never came at the same price — each subsequent clinic wanted bespoke integration work that made the unit economics resemble an agency, not a product. We archived the thesis and kept the customer research, which later shaped Practice OS.
What we changed
All three deaths pushed the same edit into the playbook: the buyer-power question moved from week ten to week two. It is now the first thing the disproof sprint tests, and it has already ended two theses before a prototype existed.