On paper, a sourcing company in Karnataka had 5,300 contract growers on 6,000 acres, managed by 36 field officers across 2 districts. Neither number held up once the officers geo-tagged every boundary.
When they came to us, their procurement team was planning purchases on field reports that were 8 days old. Some of those reports did not even agree with each other.
Settlement week showed the damage most. The same grower would sit in the register twice, under 2 spellings and 2 different acreages. Those duplicate entries leaked money at settlement in 2 ways. Non-contracted growers sold unverified produce at the collection centre under duplicate names. And some contracted growers took seed and fertiliser on credit under a second entry, then side-sold their harvest to a local trader, leaving the input advance unrecovered. When the procurement head asked which farms had actually been visited last month, all anyone had was the officers' word.
What changed for procurement
Officers geo-tagged every farm boundary, and each one was checked against 3m satellite imagery. That turned up 2 numbers the registers had hidden.
257 of the 5,300 farmer records were duplicates, leaving 5,043 real growers. Each duplicate was a route for non-contracted produce or unrecovered input credit.
And the 6,000 declared acres came to 4,380 when officers walked the boundaries. The other 1,620 acres existed only on paper, and contracted volume had been set against them. This surfaced mid-season, after planting, with that season's contracts already set.
The company had contracted 3,600 tonnes against the 6,000 declared acres. It procured 2,680 tonnes from the 4,380 verified acres. The real acres delivered. The shortfall traces back to the 1,620 acres that did not exist.
Grower IDs and verified acreage now export from the app to the company's ERP and payment system, along with every input issued on credit, so finance settles from the same records the field team keeps. Rollout took 4 to 6 weeks from signing to the field team logging in the app.
For digital farming technology in India to hold up at settlement, it must start with the field officer. Every number procurement sees begins as an entry on his phone.
A field officer's day before the app
Srinivasa, one of the field officers, covers a cluster of villages on a two-wheeler. He knows which grower's borewell gives trouble in summer, and which one is likely to side-sell to a local trader if the company's payment is delayed.
His day starts with a register in the bike's side box. Past the last village tank, his phone loses signal. So everything goes on paper: the planting date, the acreage he measured walking the bund, what he told the farmer about spacing, which inputs the farmer says he used.
In the evening, he copies his daily activity into a spreadsheet or photographs the register pages and sends them to his supervisor on WhatsApp. With 36 officers and long days, some of it always slips to the weekend. By the time it reaches head office, the data is 8 days old.
And there is no proof of the visit. When a grower later complains he has not seen an officer all month, it is Srinivasa's word against a register.
Management could not see which farms were in trouble, so the default was blanket advisory: the same recommendations and spray schedule for every farmer, whether his crop needed it or not.
Building the app around the field officer's week
Follow an officer like Srinivasa for a week and the work comes down to 6 jobs:
- Geo-tagging a farm boundary, often in a patch with no mobile signal
- Capturing farmer KYC and land details on a small screen, in the sun, with the farmer waiting
- Recording what the farmer did on the plot: planting, irrigation, sprays, fertiliser
- Logging the crop advisory he gave
- Recording which inputs were issued and how they were used
- Building enough trust to stop side-selling
We set up Crop Analytica's farm management app around that week and cut anything that did not fit it.
The officers' first question was about GPS. Was head office going to track their movement? We walked them through the geo-tagged visit log: when a grower says no officer came, the GPS timestamp is their proof of visit.
Their farmers trust them, and they did not want a phone getting in the way of that. So the app stays in the officer's hand, and crop advisory at the field stays face to face. Today, 24 of the 36 officers log every visit in the app daily. The other 12, who had spent years on paper registers, moved from 8-day backlogs to logging every 2 to 3 days, with their supervisor following up.
The app works offline. Entries save on the phone and sync once the officer is back in network coverage. Most screens are taps and dropdowns, because typing on a small screen in the sun, with a farmer waiting, is exactly where entries get skipped.
The 5 digital farming tools that stayed in use
The Karnataka team opens these every week.
1. Offline farmer registration and farm geo-tagging
Srinivasa walks the boundary with the phone, the app geo-tags the plot, and the farmer's details go in once. Each boundary is then checked against 3m satellite imagery, so a phone GPS error on a 1-acre plot does not turn into an acreage error. Where 2 farmer records landed on the same plot, the platform flagged them, and the team checked each case before removing it. That is how the 257 duplicate records came to light. Each grower now exists once in the system, with one acreage tied to a verified boundary.
2. Farmer activity log
Every visit adds to the same record: crop stage, irrigation, spacing, fertiliser, weeding. When an officer is transferred to another cluster, his replacement inherits the full history on day 1.
3. Input issue and usage records
Seed, fertiliser and crop protection supplied on credit get logged against the grower the day they are handed over. Input recovery is where contract farming money usually leaks, and now finance can match every issue when the produce comes in.
4. Visit and advisory tracking
Each visit carries a timestamp, a GPS point and the advice given. Managers can see which farms got a visit this week. For Srinivasa, that record works in his favour too.
5. Satellite crop monitoring
A weekly satellite check on every farm, so officers know where to go first.
How do you know which farms need a visit
With 5,043 growers and 36 officers, the hard question every week is which farms to visit.
Crop Analytica's precision monitoring answers it with weekly crop health and growth readings for every contracted plot. The readings come from 3m x 3m high-resolution satellite imagery, processed by our AI and ML models. No hardware goes into the field. Our crop health and growth readings are validated at over 90% accuracy on this programme's farms in Karnataka, kharif 2026.
The first report of the season came in during the growth stage. It flagged 962 farms. Officers checked each one, by phone or in person, and confirmed 893 as poor-performing.
Siddappa's plot was one of them. The satellite showed its crop health falling behind the plots around it. Srinivasa walked in and found the problem inside the canopy. The plant population was too dense because spacing had not been followed, and a top dressing of fertiliser had been skipped. He told Siddappa what to fix, standing in the field, and logged the advice before he left. 7 days later he was back to check.
Lingaraju's plot showed water stress in the same report. He was not on the farm when a field officer, Ganesh Rao, came by, so Ganesh called him, fixed an irrigation schedule over the phone and logged the call. A week later he was standing in that field again.
That became the routine for all 36 officers. Advice in person if the farmer was on his plot, a logged call if he was not. Either way, a recheck within 7 days. Officers also dropped in on farms at random to see whether the advice was being followed.
Across the 893, the causes were what any agronomist would expect: pest and disease pressure, water stress, poor plant stand, and growers skipping parts of the package of practices. Farms that stayed poor report after report went on a focus-farm list and got visited more often.
Spray schedules changed too. Crop protection now goes to plots where the satellite and the officer both see a need.
What did it change at harvest
By harvest, farms in poor condition were down from 893 to 329. That is 564 farms recovered in a single season. At the weighbridge, those 564 farms delivered 15% more produce per acre than the 329 that stayed poor. Part of that gap comes from the 168 rainfed plots among the 329, which the dry spells hit hardest.
The 329 deserve an honest look.
Thimmappa's farm was one of those. Sunil, his field officer, logged the same interculture advice visit after visit, and the weeding never happened. The farm stayed on the focus-farm list all season and finished in the 329. Anyone who has managed a contract farming programme knows a Thimmappa.
Across all 329 farms that stayed poor until harvest, only 26 were growers like Thimmappa who heard the advice and did not act. Some lacked labour for weeding at peak season; others trusted a neighbour's method. The rest were up against the season and their own field conditions:
- 168 were rainfed plots hit by El Niño dry spells: long dry stretches struck at peak growth stage, and these farms had no source for protective irrigation when the satellite flagged moisture stress week after week.
- 76 had weak nutrient management: top dressing went in late or not at all because the grower ran short of cash mid-crop.
- 59 were planted too close: once the plant stand is set too densely, spacing cannot be fixed mid-season.
For chronic plots where a grower is sceptical, officers now bring in the cluster supervisor and walk the farmer over to a neighbour's recovered plot, so he can see the advice working in a standing crop.
| Metric | Before | After |
|---|---|---|
| Farmer records | 5,300, duplicates unknown | 257 duplicates found and removed |
| Acreage | 6,000 declared | 4,380 verified by geo-tagging |
| Procured vs contracted | 3,600 t contracted on declared acres | 2,680 t procured from 4,380 verified acres |
| Field report lag | 8 days, often inconsistent | Same day for 24 of 36 officers; 2 to 3 days for the other 12 |
| Farms in poor condition | Unknown until someone visited | 893 confirmed mid-season, 329 at harvest |
| Yield, recovered vs stayed-poor farms | Not measured | 15% more per acre |
| Crop protection | Blanket spray schedule | Sprayed where flagged and confirmed |
| Field visits | Same field visit schedule for all 5,300 registered farmers | Targeted to flagged farms, rechecked in 7 days |
| Field officer time | Routine rounds on every farm | Flagged farms and farmer meetings first |
| Regional yield forecast accuracy | Unmeasured, shortfalls identified at the weighbridge | About 86% at region level, 4 weeks before harvest, measured against weighbridge arrivals |
Officers spent their days on the farms the satellite flagged, with a reason for every trip. Routine rounds on healthy farms dropped off the schedule.
That time went back to the growers. Officers used it for farmer meetings and one-to-one visits to hear what was holding each crop back. That contact builds the trust that keeps a grower's produce coming to the company's collection centre at harvest.
For the procurement head, settlement week looks different now. Verified acreage and crop status by village sit on one dashboard, next to the in-season yield forecasts. The forecasts landed 4 weeks before harvest at about 86% accuracy across the region, measured against weighbridge arrivals. Field data from 24 officers lands the same day, and the other 12 officers' clusters update every 2 to 3 days. Collection centre schedules and processing capacity get planned a month before trucks reach the weighbridge, and so do raw material commitments.
What should a sourcing team do first in 2027
Geo-tag before you contract. This team set its contracted volume on 6,000 declared acres, and 1,620 of them did not exist.
Start with the field officer. If an officer will not use the app by day 3, the dashboard at head office stays empty.
Plan for no signal. Most contracted farms sit in patchy network coverage, and an entry that needs signal to save usually ends up back in the register.
Use the satellite to pick which farms get a visit, then send the officer in to diagnose. On Siddappa's plot, the imagery led Srinivasa to the farm and he found the spacing problem on the first visit.
Tie the data to one decision before rollout. For this team it was the procurement plan, and the test was simple: did the plan get faster and more accurate?
And go back in 7 days. The recheck shows whether the farmer acted on the advice.
If you manage a sourcing, contract farming or agri-lending programme, here is a question for your next review: what share of your contracted farms are poor-performing right now, and how sure are you of that number?
If you would like to see how the Karnataka programme is set up, book a walkthrough with Crop Analytica.