How Agerpoint helped a global food corporation get ahead of a cocoa crisis, and what it took to make plant-level data actually work in the field
Overview
Cocoa is in everything. Chocolate bars, confectionery, flavourings, it's a commodity so embedded in food supply chains that most companies take it for granted. Until a few years ago, disease swept through cocoa farms across West Africa and prices jumped 130% almost overnight.
For one major global food and beverage corporation, the crisis exposed something uncomfortable. They had built an entire business on cocoa from these regions, but they had almost no visibility into what was actually happening on the farms. They didn't know exactly where their cocoa was coming from. They couldn't tell whether disease was spreading toward their suppliers. They had no reliable way to check whether cocoa from a new region would meet their quality standards.
The data existed, in theory. It just couldn't be collected fast enough, or at the right scale, to be useful.
This is the problem Agerpoint set out to solve: how do you get precise, plant-level observations from remote farms to the people who need to act on them, without it taking months and costing a fortune?
Why This Is Hard
Before any technology enters the picture, it helps to understand what agricultural measurement actually looks like on the ground.
A specialist visits the farm, does a concentrated round of data collection, and leaves. They walk rows of cocoa trees with a tape measure and calipers, count pods by hand, measure each one individually to check ripeness, and look closely at leaves for early signs of disease. By the time the report reaches anyone who can act on it, weeks have passed.
And that's the optimistic version. In many of the regions where cocoa grows, specialist agronomists aren't available locally at all. So companies end up measuring a small sample of trees and hoping it reflects what's happening everywhere else. Entire sections of a farm go unexamined. Problems stay invisible until they're too big to ignore.
For a corporation trying to manage supply risk, this isn't just inconvenient. It means operating blind. You find out about a disease outbreak when prices spike, not when the first trees start showing symptoms.
One Approach in Practice
Agerpoint's first question was practical: what do people in remote growing regions already have access to? The answer, for the vast majority of people globally, is a smartphone.
Rather than building new hardware or training new specialists, Agerpoint built software that turns an ordinary smartphone into a precision measurement tool. A local field worker walks around a cocoa tree, taking photos and short video clips while the app guides them through each step. The instructions are simple enough that no agronomy background is needed, and the app works in multiple languages so it can be used by workers across different regions. The whole capture takes a few minutes.
Where previously the corporation would have needed to fly in a researcher to measure a tree by hand, now a local worker with a phone can do it, consistently, repeatedly, and across far more trees than any visiting expert could manage.
Those images are sent to Agerpoint's cloud platform, where the software does the work that used to require hours of manual counting. It pieces the overlapping photos together into a three-dimensional model of the actual tree. Then machine learning analyzes that model: counting every pod, estimating the size and ripeness of each one, and flagging any leaves or pods that show signs of disease.
The contrast is stark. Before, someone would count pods one by one and measure each with calipers. Now, the model does it automatically, across dozens of trees at once, producing results in a fraction of the time.
But knowing what's happening on individual trees still isn't enough. To manage a farm, or a supply chain, you need to understand patterns across the whole landscape.
This is where aerial imagery comes in. Agerpoint layers the ground-level smartphone data with images from drones or satellites. The aerial view acts as an early warning system. When one section of a farm starts showing stress, it shows up as a color shift in the imagery: green for healthy canopy, red or pink for something that needs attention. Instead of sending workers to walk the entire farm hoping to stumble across a problem, teams can now see exactly where to look before anyone steps into the field.
For the corporation, this combination changed what was possible across several parts of the business.
On disease, for the first time they could track the spread of infection tree by tree across a farm, rather than discovering the full extent of damage after the fact. By layering in weather and wind data, they could model where disease was likely to move next, creating an early warning system that gave them time to respond. Treatments could be targeted at specific affected sections. Sourcing could be diversified before a farm-wide crisis hit.
On quality, when the corporation began exploring new cocoa suppliers in Latin America to reduce its dependence on West Africa, it faced a familiar problem: how do you know if the cocoa will meet your standards without sending teams to every farm? Agerpoint let them verify pod sizes and ripeness remotely, checking quality against their requirements before signing any procurement contracts.
On compliance, the EU's deforestation regulation requires companies to demonstrate that the commodities they source haven't contributed to forest clearing. Separately, sustainability teams needed tree-level data to calculate carbon stocks. In the past, both of these would have meant separate data collection exercises. Because Agerpoint's ground-level measurements capture location data and tree dimensions as a matter of course, the same dataset could serve both needs, with procurement, agronomy, sustainability, and compliance teams working from shared information rather than each commissioning their own fieldwork.
Image Credits: Agerpoint
What This Enables, and Where It Falls Short
What changes:
The corporation went from reacting to crises to having a more continuous picture of what was happening across their supply chain. Decisions that used to be made on guesswork: when to harvest, where disease risk was building, whether a new sourcing region was viable, could now be made on actual evidence.
The other shift was internal. Teams that had been operating in silos, each running their own data collection, found themselves working from shared information. That has real cost implications, but it also changes how a company responds when something goes wrong. Everyone is looking at the same picture.
Where it falls short:
The quality of what comes out depends on the quality of what goes in. If field workers don't follow the capture process consistently, results become harder to compare across trees and visits. Getting that consistency right takes investment in training and workflow design, it doesn't happen automatically.
The machine learning models that detect disease are trained on existing images, which means they perform best in conditions similar to where that training data came from. A model built on data from one cocoa variety or one region may be less reliable when applied somewhere new. The system needs to be tested and validated in each new context before it can be fully trusted.
Biodiversity is a more open question. Agerpoint's tools can support structured data collection by local communities, with GPS tagging and multi-language support. But biodiversity measurement depends heavily on local ecological knowledge, what to measure, what counts as healthy, what species matter in a given landscape. The tools provide infrastructure. The expertise still has to come from people who know the place.
What Others Can Take From This
Visibility has to come before strategy. The corporation's problem wasn't that they lacked a sustainability strategy or supply chain ambitions. It was that they had no reliable picture of what was actually happening on the farms they depended on. Any organization serious about managing nature risk needs to start with that question: what can we actually see right now, and what are we operating blind on?
The tool is only as good as the workflow around it. Agerpoint's technology works in practice because it was designed for the conditions that actually exist in the field: limited connectivity, workers with different languages and skill levels, farms that are nothing like a controlled research environment. A tool that works perfectly in a lab or a pilot but breaks down at the point of real deployment is not a solution. Test for field conditions, not ideal conditions.
Look from above before you send anyone in. Using aerial imagery to identify where problems might be, before directing ground teams to investigate, makes field data collection far more effective. Without that first pass, sampling is random. Random sampling across hundreds of acres will miss things.
One data collection effort should serve more than one team. The separation between agricultural data, sustainability data, and compliance data is often an organizational habit rather than a technical necessity. If the right information is captured once, it can flow to procurement, sustainability, and compliance simultaneously. That requires deciding upfront what everyone needs, not retrofitting it later.
The buyer and the user are rarely the same person. The corporation bought the technology. The field workers used it. Both groups had to be served for the system to work. Any organization deploying nature measurement tools needs to think carefully about usability at the point of collection, not just value at the point of reporting.
Case Classification:
Measurement · Modeling · Computer Vision · Digital Reconstruction Image-derived plant-level observations · Perennial Crop Systems · Agricultural Supply Chains
