Most farm software is written for a connected farm with a salaried
agronomist and one crop. That is not the farm we started from. Ours
has a cooperative between the farmer and the buyer, a field officer
covering four parishes on a motorcycle, and a valley where the network
drops for the whole afternoon.
So the thing we are trying to beat is not another platform. It is the
hardback register at the buying centre — which is fast, trusted, works
in the rain, and cannot tell you which garden a sack came from. Any
screen slower than that page gets abandoned, and rightly.
We started with the boundary, because everything else hangs off it,
and we made the system refuse things rather than warn about them. That
is a harder product to build and a slower one to sell. It is also the
only kind whose records are worth anything to a buyer, a lender or an
auditor two seasons later.