It is not about the data. It is about the model.
There is a version of this business that sells storage and dashboards: get everything into one place, put charts on it, and insight will follow. We are not that. Charts describe. They do not tell you what happens if you raise the night temperature by one degree for the next six weeks.
The model is already there
Walk a greenhouse with a good agronomist and you get a running commentary that is, structurally, a model. Fruit load is high, so the plant will pull assimilates away from the head, so the next truss sets short, so in five weeks there is a dip. Push the EC and the fruit gets firmer and sweeter, but you pay in kilos, and there is a point where you pay in blossom end rot.
That is a set of states, rates and trade-offs. It is a good model. Its weaknesses are that it lives in one person’s head, it cannot be run forward for eight scenarios before lunch, it cannot be argued with using numbers, and it retires when they do.
Writing it down does three things. It becomes testable, so it can be wrong in a specific way instead of vaguely. It becomes shareable, so the grower, the crop manager and the owner argue about the same object. And it becomes runnable, so “what if” takes seconds.
Different people, different mental models
The disagreements we get called into are usually not about the data. Everyone has the same sensor readings. They are about which mental model is right.
- The grower steers on the plant in front of him, on a horizon of days.
- The crop manager steers on the plan, on a horizon of a season.
- The owner steers on the contract, the energy position and the labour bill.
A model that only speaks plant physiology cannot settle that argument. So ours does not stop at plant physiology. The same run that produces dry matter partitioning also produces kilos per week, labour hours, energy and water use, and the margin at the price you actually get. When the grower and the owner disagree, the disagreement becomes visible as a difference in two scenarios rather than a difference in temperament.
Biology, steering, outcome
Three layers, one run:
- Biology — states and rates. Photosynthesis, water status, reserves, growth, stress, mortality.
- Steering — what a human or a climate computer can actually change. Setpoints, feed rate, irrigation, planting date, stocking density, harvest moment. In the equations these are controls, not wishes.
- Outcome — the KPIs somebody is measured on. Yield and quality, cost per kilo, margin, water per kilo, energy per kilo, hours per hectare, mortality, welfare and emission indicators.
Most models in the literature do the first layer well and leave the other two to the reader. Most commercial tools do the third layer and guess at the first. The value is in keeping the chain unbroken, because that is the only way a biological insight turns into a decision somebody is willing to sign.
What we do not do
- We do not need your data lake first. If you have one, fine, we will use it.
- We do not sell a black box that has “learned” your farm. If it cannot be explained on a whiteboard, it does not go in.
- We do not start with a dashboard. A dashboard on top of a model nobody believes is an expensive way to lose an argument.