how the numbers are made

An estimate you can check.
The yield, the range, and what limited it.

Divisi works out the yield a crop can attain on your own field, how that changes in a dry year and in a wet one, and which factor is holding the rest back. A season is too expensive to plan on a guess, so every number below comes with the working behind it.

See what grows on my farm Get the app

Divisi's verdict for one field: grows well in Home field, rated for a normal season, with what to correct

Also consider: ranked alternatives for this plot — hot pepper, robusta coffee and garlic, each able to grow here with care

How it works: an implementation of the FAO's Global Agro-Ecological Zones framework, run against published soil, climate and terrain data for your actual field.

the inputs

Published science, read at your own field

Every recommendation starts at one actual field: soil properties from published soil mapping, thirty-year monthly climate normals, terrain from elevation models. These are the datasets agricultural research runs on. The work is reading them correctly at the scale of one smallholding, and knowing what they cannot resolve.

the method

A suitability cascade, not a score somebody invented

The crop's potential is cut back by each thing the field imposes — temperature through the cycle, water against demand, soil chemistry, terrain — in the order the FAO framework sets out. Out comes a suitability class and the yield that crop can attain here, with the reasons attached: not '85 out of 100', but which factor cost what.

Get this worked out for your own field — set up your farm in 3 minutes.

Two Divisi crop picks for one field, each with a photo, a grows-well-here verdict and the yield to expect

the uncertainty

A range, because a single number would be a lie

Climate averages say nothing about the year you are about to have. So Divisi runs the same plan — same field, same crop, same date — through a drier and a wetter season, and reports the band. A crop with a narrow band behaves whatever the year does, which is often the more useful thing to know.

What happens when a farmer taps 'check my farm'

Read the field

Soil, climate and terrain for the field you pinned are assembled into one environmental picture of it.

Run every candidate crop

Each crop in the catalogue is run through the cascade against that picture, at each plausible planting date, and ranked.

Return the reasons

The farmer gets a class, an expected yield, a range and a planting window — with what limited the crop, in words.

it doesn't work alone

One season, seen from every side

Divisi is one thing, not a course. Here is what this connects to — and why that matters more than any single feature.

Planting calendar

The most visible thing the cascade produces: a planting window for one field.

Farm records

Where the farmer's ground truth comes back in, and where the model's limits stop mattering.

Farmer Hive

What a model cannot see, a farmer standing in the field can.

Get Divisi — set up your farm in 3 minutes

Pin your farm once, then add your workers or your family — everyone works from the same crop plan, the same guides and the same farm diary.

Dated and photographed Works offline Runs on the phone already at the farm Light on data

What you can count on

The method is published, the reasoning is on the page, and every dataset is named. Soil and climate resolve to a grid cell, not to your boundary — Divisi knows your neighbourhood better than it knows one corner of one field, which is why it asks you for what the data cannot see. Yields are a ceiling to aim at. We are looking for a research partner to run a randomised trial and publish what it finds.

Questions worth asking

Divisi gives you a figure for the crop you asked about, on the field you pinned: the yield that crop can attain there under good management. On either side of it sit a drier-year and a wetter-year figure. With them comes the reason the number is not higher — water, temperature, a soil property, the slope. It is a ceiling to aim at, not a promise about one particular season, which is why the range matters more than the middle number.

No. The recommendation engine is a deterministic agronomic model: the same field, crop and season produce the same answer every time, and every answer decomposes into the factors that produced it. Language models are used in Divisi for writing and translation work, never to decide what will grow on somebody's field.

Soil from ISRIC SoilGrids 2.0 and iSDA's African soil mapping; climate from TerraClimate, as monthly normals for 1991 to 2020 covering temperature, rainfall and evaporative demand; terrain from digital elevation models; and crop parameters compiled from published farming research including FAO and CABI material. Live weather for forecasts and alerts comes from a commercial provider for your actual field. The rating tables we build on top of these are ours; every input underneath them is public, and you can go and read it.

It is precise at the scale it works at, which is a grid cell of a few kilometres — sharp enough to tell two farms in the same district apart, which is already far beyond the district-wide advice most farmers are given, and short of resolving the wet corner of one plot. That is why Divisi asks the farmer for the things the data cannot see, such as whether a field is irrigated or under a greenhouse, and why the diary matters: the farmer's own record is the ground truth the model does not have.

Yes. We will walk an agronomy or M&E team through the cascade, the data sources, the calibration and the failure modes, and we would rather do that before a programme starts than after. Get in touch and we will set it up.

Because published together across the whole catalogue they are the engine, and rebuilding it from them is a weekend's work for a competitor. The method is public, the reasoning behind any individual recommendation is visible to the farmer who received it, and the crop, pest and disease guidance is free to read on this site. The rating tables underneath stay ours.

All questions answered →

For organisations

Buying from farmers, or running a farmer programme?

See every field you buy from or support on one map, measured by the farmers who walked it, and which ones are falling behind while there is still time to help. Free for the public sector and research.

Explore Divisi Partners →

An estimate you can check. The yield, the range, and what limited it.