Divisi
Divisi for Academia & Research
Primary data collection
is the line that eats the grant.
Season-long records from working smallholdings, with a disclosure guarantee you can write straight into a protocol.
Pricing an enumerator round for a season that starts before the grant clears? Applying is free, and a person reads it within a working day or two.
The enumerator round is the largest line in the budget, and the recall data it returns is thin by the time it is cleaned. The panel loses households between waves, and the ethics application takes a term of its own — all to observe a season that was happening anyway.
a look inside
Your areas, shaded by what is actually in the ground
No field is drawn and nobody is named. Where an area holds fewer than ten farms it merges into its parent and is republished at that level, so the map stays honest and never goes blank.
The aggregate view at ward resolution, with the demographic split beside it. Illustration — the layout, colours and figures a partner account shows, drawn rather than sampled, because a real cohort belongs to the farmers in it.
the ground you are standing on
Two numbers, neither of them ours
The sample behind the finding that management alone lifts Sub-Saharan maize from about 2 to 4.3 t/ha. Assembling a panel at that scale is precisely the cost this is meant to lower.
Nature Communications, 2024The share of Sub-Saharan Africa's food calories from farms under five hectares — the least-instrumented majority of the system most of this literature is about.
Samberg et al., 2016
Why the series does not break
An instrument that exists only to collect gets abandoned once the incentive stops, which is where a lot of smallholder panel data loses its third wave. Divisi is a free farming app in its own right: what to plant on this field and when, the job due this week and why, the pests to scout for. The farmer also keeps a diary for their own purposes. Observation is a by-product of a tool that pays for itself in their hands.
Read a crop guide as a farmer sees it →See it on your own phone → One link, both stores — it opens the right one for your device.
impact, and the balance sheet
The questions this is already being asked
Climate adaptation, the yield gap, gender and youth in agriculture, and market participation are all bottlenecked on the same thing — affordable, continuous observation of real management decisions.
01
Climate adaptation, observed rather than modelled
The adaptation literature is thick with models and thin with panels. This is a panel, and the baseline it is measured against is a published method rather than a proprietary score.
How Planting dates, operations and condition recorded through the season against a documented FAO GAEZ baseline for each field's own soil, terrain and climate.
02
The yield gap, at the decision level
The gap decomposes into decisions rather than sitting as a residual, which is the step most yield-gap work has to approximate.
How Attainable yield per field beside what was actually done on it — which operations were completed, which lapsed, and when.
03
Gender and youth participation
The inclusion variables a funder asks for, on a definition that matches national policy, without a separate instrument to collect them.
How Counts of people rather than plots, split by gender and by age band, with the same ten-farm floor applied to the group being described.
04
An ethics story that starts in the right place
A review board is asked to approve a mechanism that already exists and is auditable, which is a shorter conversation than approving one you have designed for the study.
How Aggregates stand on at least ten farms and roll up rather than suppress; per-field access exists only through a consent the farmer gave per field and per season, revocable from their phone.
How this maps to the Sustainable Development Goals — read the alignment →
Why this passes a review board
A disclosure guarantee you can write into a protocol
Every published aggregate stands on at least ten farms, and the cell adapts rather than the threshold: a thin ward folds into its district and is retested. The rule is conservative about differencing — if any child cell is short, the parent is published and none of the children are — and it is a pure function with property tests asserting no published set permits recovering a suppressed one.
Observation instead of recall
Planting dates, field operations, inputs, photographs and yields are recorded when they happen, by the farmer, because the record is theirs and useful to them. Boundaries are walked with GPS, so hectares are measured rather than estimated — which removes one of the larger error terms in smallholder yield work before the analysis starts.
A documented, citable method underneath
Suitability, planting windows, water adequacy and attainable yield come from an implementation of the FAO GAEZ v5 cascade over named public layers — ISRIC SoilGrids 2.0 soils, iSDAsoil nutrients, GAEZ terrain, long-run climate normals, NMME seasonal anomalies. Named datasets with versions, not a proprietary score.
It opens in your own GIS
Your cohort is served as GeoJSON from the dashboard, at whatever resolution your account is granted, so it opens directly in QGIS, R or Python. A dataset that cannot leave is a dataset that cannot be cited.
How a research relationship works
Tell us the question
Apply with the institution, the study area and what you are trying to observe. A person reads it. Accredited institutions get aggregate access free.
Define the study area
Choose districts and wards from the administrative tree, draw a catchment on the map, or upload the project boundary you already hold as GeoJSON or a shapefile.
Read it every week, not every wave
The cohort is continuous rather than three cross-sections. Condition, crop mix, operations completed and missed, and the age-band split, all moving through the season.
Go per-field only where farmers agreed
If the protocol needs individual fields, that is a programme farmers join and consent to by field and by season — the same mechanism a contract scheme uses, and the same one your ethics committee will recognise.
what your account shows
Resolution follows the relationship
Your account reads area figures across the region it is approved for. Where you run a programme farmers join, those fields resolve individually as well.
Area figures
District and ward totals
Counts, hectares, crop mix, condition and the demographic split for the areas your mandate covers — grouped by the administrative boundaries your own statistics already use. Verification is what moves an account from district totals to ward totals.
Field detail
One field, its season, its evidence
The crop, its walked boundary, its projection, the work done and missed, and the dated photographs — for the farmers who joined your programme and agreed to it, per field and per season. Enrolment is how an account reaches this, which is why the invitation flow is part of the product.
The floor under every published figure
Every area figure stands on at least ten farms. Where a ward holds fewer, it merges into its district and is republished there rather than being hidden — so the map sharpens as coverage grows and never goes blank, and no published set lets a suppressed one be recovered by subtraction. That floor is what makes serving these figures legitimate, and it is why a data-protection review of this is short.
What it costs an institution
Free for accredited institutions
Universities, colleges and research institutes: the map, ward and district figures, the demographic splits and the GeoJSON, with no seat limit. Teaching use is free and we would rather encourage it.
Per-field cohorts are scoped case by case
They require farmers to be recruited and to consent, which is real work with a real cost and is usually a line in the grant rather than a licence fee. Curriculum development and workshops are quoted separately.
We ask one thing back, and it is not money
Cite the method, and send us the paper. Published work using this is worth more to us than the fee we would have charged for it.
What you can count on
The cohort selects itself: farmers who chose to install Divisi. We would rather hand you that characterisation for your study area up front than have a reviewer find it. It is also why the series does not break — nobody is paid to report, so nobody stops when the funding does. Use it for what dense observational data is uniquely good for: real management decisions, dated and geolocated, at a cost no enumerator round reaches. Let your own design carry attribution.
Questions worth asking
not quite you?
Same platform, different job
Plenty of organisations are two of these at once. The access model is identical; what changes is what the figures are for.
Start with the map, not the contract.
Tell us who you are, where you work and what you are trying to see. We read every application ourselves and scope the rollout with you.
Universities and agricultural colleges, national and CGIAR research institutes, agricultural economics and rural development faculties, climate and food-systems programmes.
Already a partner? Sign in to your dashboard
