The cartridge is the size of a credit card and about as interesting to look at. Rosalind Achebe holds it up between two fingers, without ceremony, the way you would hold a bus ticket. Inside it there is one channel, one reagent, and one answer, which arrives in eleven minutes and is either yes or no. It cost about $9 to make. It is, by a distance, the least sophisticated object her company has ever produced, and it has paid for everything else in the building.
On the shelf behind her is the other thing. Forty channels. One sample. A machine that was going to answer forty questions at once and change how a whole category of testing worked. Kestrel Assay spent three years and, by Achebe’s reckoning, about $2.1 million on it, and for most of that time she believed the single-channel cartridge was a distraction being pushed on her by people who did not understand the science.
“I was right about the science,” she says. “The forty-channel machine works. It has always worked. I was wrong about literally everything else, which is a distinction I was too pleased with myself to notice for about two years.”
The ambitious thing
The pitch for the platform was genuinely good, and Achebe can still deliver it. One sample, forty assays, eleven minutes, at a marginal cost that made the existing workflow look like an antique. Every lab she showed it to agreed. Nobody bought it, and the reason they gave was so consistent that she eventually stopped writing it down.
A lab that buys a forty-channel machine is not buying a machine. It is rewriting its validation, retraining its staff, revisiting its accreditation, and explaining to somebody senior why the old process — which is not broken, which nobody has been fired over — should be replaced by an instrument from a company nobody has heard of. The purchase is not a purchase. It is a project, with a sponsor, and the sponsor has to be prepared to be wrong in public.
We built something that required a customer to be brave. There is no such customer. There are customers with a problem and a budget, and bravery is not a line item.
The boring thing
The single-channel cartridge came out of a conversation Achebe now says she attended in a bad mood. A veterinary practice manager, asked what the forty-channel machine could do for her, said that she did not have forty questions. She had one question, she had it about nine hundred times a month, and the answer currently took two days to come back from a lab forty miles away, by which time the decision it informed had already been made without it.
One question. Nine hundred times a month. No accreditation project, no committee, no bravery. Kestrel stripped thirty-nine channels out of a machine it had spent three years building, and shipped the fortieth as a $9 disposable with a reader that cost $1,200 and did nothing clever whatsoever.
Achebe describes the internal reaction as the worst month of her professional life. Two of her scientists thought she was liquidating the company’s reason to exist and told her so in a meeting she says she deserved. The objection was not stupid. The cartridge is not novel. It could have been built in the 1990s, and roughly was. Its entire innovation is that it is in the room where the question gets asked instead of forty miles away.
- The platform: three years, $2.1M, forty assays, eleven minutes, zero units sold.
- The cartridge: eleven weeks from decision to first shipment, one assay, eleven minutes, $9.
- First-year cartridge volume: 61,000. This year it will pass 400,000.
- Readers placed: 340. Achebe gives them away below cost and says the reader is a coat hook.
- Share of revenue that is now consumable rather than capital: 91%, which she says is the only number she would defend in a fight.
The last one is the mechanism. A machine is a purchase somebody has to approve. A cartridge is a supply somebody reorders without thinking, out of a budget that already exists, on a Thursday, by email. Kestrel stopped selling a decision and started selling a habit, and the habit compounds in a way Achebe admits she had considered beneath her.
What the dull one bought
The forty-channel machine is not dead. It is in the same building, being worked on by the same two scientists who objected, funded entirely by cartridges. Achebe is careful about the causality here, because she thinks founders tell this story backwards. The cartridge did not merely pay the bills while the platform waited. It fixed the platform. Four years of a narrow test running nine hundred times a month in real practices produced a volume of failure data that no amount of laboratory work would have generated — every way a reagent degrades in a van, every way a busy person mishandles a sample at 8 a.m. The forty-channel machine on the shelf today is a better instrument than the one they could not sell, and it is better for reasons that only exist because they shipped the dull thing.
She is not sentimental about the lesson and refuses to make it general. The Corrigan Institute has argued that most diagnostics startups die holding a validated instrument nobody has bought, and Achebe thinks the reason is rarely the science. It is that the ambitious version asks a customer to change, and the boring version asks a customer to reorder. She would still build the platform. She would simply not have spent three years believing that a thing being possible was the same as a thing being wanted. Then she puts the cartridge down on the bench, where it looks like nothing at all.