What launched
In April 2026, JDE Peet's, Louis Dreyfus Company, Sucden, Neumann Kaffee Gruppe, Touton, Sucafina and Tchibo launched the Coffee Canopy Partnership: an industry-funded effort, built with Airbus as technical partner, to produce the first openly accessible high-resolution map of global coffee-growing regions. The pilot covers East Africa — Ethiopia, Tanzania, Kenya, Uganda, Burundi and Rwanda, roughly 1.2 million square kilometres — combining Pléiades and Pléiades Neo satellite imagery (down to 30cm resolution) with AI models trained to distinguish coffee plots from surrounding forest. Worldwide coverage is planned for 2027. The initiative has backing from the UK's FCDO and endorsement from the UN FAO.
Laurent Sagarra of JDE Peet's has described the project as pre-competitive by design — data shared across the supply chain rather than held inside any single company. That framing matters, and it's worth taking at face value: this is genuinely different from a vendor selling a proprietary compliance tool. It's closer to shared infrastructure the whole sector can build on, which is precisely why it's worth engaging with seriously rather than treating as a competing product.
What the mapping actually resolves
Reports on the pilot note that the underlying models can identify farm boundaries and separate coffee plots from forest with real precision, even across the fragmented smallholder landscapes typical of East African highlands. That's a meaningful technical achievement, and it directly targets a problem we've written about before: coarser satellite data has a long history of misclassifying shade-grown and agroforestry coffee as forest loss, putting smallholders who never touched a natural forest at risk of losing EU market access. A shared, high-resolution reference layer is a real fix for that specific failure mode, at a scale no single importer could fund alone.
What a shared map isn't
Here's where the distinction has to be precise, because it's easy to blur. Coffee Canopy Partnership produces a landscape reference — a view of where coffee-growing land sits, and how that land has changed. It doesn't, and isn't designed to, generate the specific attestation an individual operator has to file: that this exact plot, tied to this exact supplier relationship, is the origin of this exact shipment. An AI model that can draw a plausible farm boundary from a satellite image hasn't confirmed that the boundary belongs to the supplier named on your invoice. That link — plot to shipment to declaration — is still the operator's job, built from their own supplier records, not something a shared landscape dataset can attest to on their behalf.
Even where the two overlap — a farm that sits inside the mapped landscape, correctly identified as coffee, not forest — the operator still has to turn that into a submission: a GeoJSON file, properly structured, geometrically valid, tied to the right country and the right supplier. That step doesn't disappear because the surrounding landscape is well mapped.
Why we know that step still gets missed
We tested this directly. Earlier this month we ran eight geolocation features — one valid, seven with deliberate errors — against the EUDR Information System's own acceptance server. Swapped coordinates, a self-intersecting polygon, a plot declared in the wrong country: all eight were accepted without a single warning. None of that is a landscape-mapping problem. High-resolution imagery of an entire region doesn't catch a coordinate transposed at the point of entry into someone's spreadsheet. That failure happens downstream of any map, at the moment a specific file gets built for a specific submission — which is exactly the layer landscape initiatives, by design, don't touch.
A better map of the forest doesn't validate the file that gets filed for one farm inside it. Those are two different layers, solving two different problems, and both need to exist.
TraceBean sits at that second layer: the point where a specific supplier's coordinates become a specific operator's submission. We validate geometry, check country-boundary consistency, and flag structural errors before that file ever reaches the Information System — regardless of how good the landscape data around it is.
Initiatives like Coffee Canopy Partnership make the surrounding picture clearer. They don't remove the need to get each individual file right.
Coffee Canopy Partnership details are drawn from the joint JDE Peet's / Airbus press materials published 22 April 2026, and subsequent industry coverage of the pilot's technical scope.
← Back to Blog