Science and software, on the same payroll.
Built on the conviction that scientific and digital expertise — not more land — is what lifts agricultural upstream productivity.
What we learned in five years.
We set out as a market-linkage business and ended up an operator — because every problem we tried to solve from the outside turned out to live inside the field.
We started in 2021 with a simpler thesis than the one we hold now: that the variability in Indonesian horticulture was a market problem. Find the farmer a better buyer, a fairer price and a reliable off-take, and the rest would follow. We built the trading and market-linkage product to prove it, and it worked — right up to the point where the crop itself became the constraint. Access moves the price. It does not move the yield, and it cannot rescue a lot the field was never run to spec in the first place.
So we moved upstream. NextBio gave us inputs we controlled — seed and biologicals developed with Indonesian research institutions for Indonesian conditions — and we spent two seasons writing down what our best agronomists actually did, until it was a set of SOPs anyone could follow. Around 15,000 farmers came through that programme and the yields moved. But the lift travelled only as far as an agronomist could drive: practice held on the blocks somebody watched and drifted on the blocks nobody did. The bottleneck had stopped being knowledge and become execution.
That is why we built VarmOS and stopped advising. An operator does not hand over a recommendation and hope. It dispatches the work order, records what was done on which block on which date, and holds a QA gate before anything leaves the packhouse. Through the 2024–2025 pilots we ran that loop on our own managed blocks, and the thing we cared about most finally began to behave: Grade-A stopped being a good season and started being the output of a process.
Five years in, the conviction is narrow and hard-won. In Indonesian agriculture the scarce input is not land, capital, science or market access — all four are available to anyone willing to go and get them. What is scarce is the ability to run the same standard on the thousandth hectare as on the first. That is the whole of what we are building, and it is what we ask to be paid on. The next measure of it is roughly 3,000 hectares under management, at grade, against a buyer SLA.
Access is necessary, not sufficient
A better buyer raises the price of what you already grow. It does nothing about what comes out of the ground, and it will not save a lot that missed spec.
The inputs have to be ours
Seed and biologicals developed with Indonesian research institutions for Indonesian conditions, QA-ed to one specification — so the same protocol returns the same crop, lot after lot.
Advice does not scale. Systems do.
Fifteen thousand farmers taught us that knowledge transfers easily and discipline does not. Consistency arrives when work is dispatched, recorded and checked — not when it is recommended.
Execution at scale is the moat
A variety or a protocol can be copied in a season. Running them to the same standard across scattered blocks, cycle after cycle, is the part that takes years to build.
The fee should follow the result
If the standard is the product, we should carry the risk of missing it. That is why our contracts are profit-share or KPI-linked rather than a flat invoice.
We are paid when the yield goes up.
Profit share or a KPI-linked fee — not a flat invoice that runs whether the harvest lands or not. If the numbers do not move, that is our problem too.