Most support software is sold by the seat. That made sense when the work was done by the people in the seats. It makes less sense when an agent does part of it.
We spent a long time on this question before we sold anything, and we ended up somewhere simple: you pay when Kestrel resolves a ticket, and not otherwise.
What a seat price gets wrong
A seat price charges you for access. Whether the product closes one ticket or ten thousand, the bill is the same. That puts all the risk on the buyer. If the agent turns out to be bad at your kind of ticket, you have still paid for it.
It also sets up an awkward incentive. A vendor paid per seat does well when you hire more people. A vendor selling automation does well when you need fewer. Charging per seat for a product that reduces seats means working against your own customer.
What counts as resolved
The hard part of pricing by outcome is agreeing on the outcome. Our definition has three parts:
- Kestrel closed the ticket itself, with no person involved.
- The customer did not reopen it within seven days.
- The action it took stayed inside the limits you set.
A draft that a person sends is not a resolution. A ticket handed to your team is not a resolution. A refund you declined is not a resolution. None of those are billed.
Why seven days
We tried a shorter window first. Twenty-four hours looked reasonable until we watched what happened to tickets that were closed on a Friday. A week turned out to be the point where the reopen rate stops moving for almost every team we looked at.
What it changed for us
Pricing this way changed what we build. A feature that makes the agent sound better but does not close more tickets earns us nothing, so we do not build it. The roadmap is mostly a list of reasons a ticket could not be closed, in order of how often they happen.
It is a narrower way to run a product. We think it is the honest one.