Three desks read one agricultural file, and the committee receives three different answers to the same ratio.
For banks and microfinance institutions that lend to agriculture. Each answer was computed correctly, from a different definition, by someone who had no reason to think there was another.
On a branch network, variation is not the exception. It is the distribution.
The reading a file receives depends on who picked it up, what else was on their desk that week, and whether anyone senior was free that afternoon. When the answer is nobody, the branch telephones head office, or it does not, and the paper still arrives at committee looking settled. On an agricultural book the spread is wider still, because a season, a price and a buyer all move together, and not every officer has lent against that crop before.
When the evidence is a photograph of a ledger
For a microfinance institution the constraint is rarely the officer. The record is thin, the proof is a phone photograph of a handwritten book and a mobile money history, the repayment depends on a harvest that has not happened yet, and the loan is too small to justify a senior reader on its own. So the parts that are hardest to read are the parts that end up assumed. That photograph is the evidence, not a poor substitute for it, and it is read as the page it is: the inputs bought before planting, the sales after the last harvest, the gap before the next one.
What works today: one standard on every desk
Today LitFin Intelligence does credit appraisal, underwriting and risk analysis for agricultural lending: for an individual loan, a group loan, or a whole portfolio, in English and Swahili. Deployed inside your institution, it reads every file the same way, against your own credit policy: a paddy grower's seasonal line, a poultry unit's working capital, a farmers' group's shared loan. Those are illustrations, not a list. It reads pages that arrived as photographs, and says plainly which ones it could not read. What reaches your committee is a number that comes with the basis it was computed on, instead of three of them. Tested on agricultural files with officers from a major agricultural lender that plays a large de-risking role for agricultural loans in Tanzania.
The senior reader beside every officer
That reading is beside the officer while the file is still open, in the branch or in the field. And when an officer moves on, the reasoning does not move with them. It stays with the file.
Your institution keeps every decision
LitFin Intelligence runs inside your institution, on your own files, against your own lending policy. It shows which evidence mattered and why, names every gap starting with the one that would most complete the file, and stands beside the officer while the file is still open. It does not lend and it does not decide. Your officers and your committee do.
Where this is going: the chain, the group and the season
A smallholder is rarely financed alone. She is financed as part of a group and of a value chain, and nobody sees all of it at the moment your committee decides. We are building the appraisal of a whole value chain as one case, groups a lender can finance with every member visible beneath the group, and a way to follow a financed season while there is still time to act, with village-based agronomists as the hands on the ground. This part is in design and early build. It is not live yet.