Most EUDR commodities are grown by people who have never exported a coordinate in their lives. Coffee, cocoa and rubber consignments routinely aggregate hundreds to thousands of farms under 4 hectares, and Article 9 wants the geolocation of every one. This is a field-operations problem more than a GIS problem, and it has a known playbook. This guide is that playbook, distilled from how exporters, cooperatives and their EU buyers actually run collection campaigns.
First, shrink the problem: points beat polygons
Plots of 4 hectares or less can be declared as a single latitude/longitude point at six decimal places. For smallholder chains that converts an impossible surveying exercise into a capture exercise: one person standing inside each plot with a phone. Polygons remain better evidence (they define exactly which pixels answer for the plot), so capture boundaries where plots are large, disputed or near forest, and points everywhere else. The full legal rules are in the geolocation requirements guide; the routes to polygons are in the polygon guide.
Design the campaign around existing structures
Nobody visits ten thousand farms from Rotterdam. Collection runs through the aggregation points that already touch every farmer: cooperative agents, washing stations, buying clerks, lead farmers. The pattern that works:
- One enumerator per cluster (a village, a co-op section) with a smartphone and a capture app that records point, GPS accuracy, farmer identifier, and declared area in one form.
- Capture inside the plot, ideally near its centre, never at the farm gate, the buying station or the enumerator's home. Duplicate coordinates across farmers are the signature of shortcut capture and are easy to detect later.
- Record the area honestly as a number in hectares. In TRACES GeoJSON, a point without an area defaults to 4 hectares, and an area typed as text is silently read as zero, so this one field causes outsized damage. The format guide covers the trap.
- Pay for completeness, verify for quality. Enumerator incentives per plot produce fast, sloppy data unless a random-sample re-visit is part of the deal.
Accuracy: good enough is genuinely good enough
Six decimal places is a recording precision, not an accuracy demand: consumer phone GPS lands within 3 to 10 metres in the open, which is ample for locating a plot for satellite screening. What degrades fixes is canopy and slope, so under shade (Ethiopian forest coffee is the extreme case) let the receiver settle, and capture the app's reported accuracy alongside the point so downstream QA can weight it. Record coordinates at full precision and never round or re-type them by hand.
Validate the day data arrives
Run every batch through automated checks before it enters your master file: points in the sea, in towns, outside the declared country; duplicates across farmers; areas of zero or exactly 4.0 everywhere (a default, not a measurement); accuracy outliers. Every failure is a cheap conversation while the enumerator is still in the village and an expensive one at filing time. Then screen the clean points against the 2020 cutoff immediately: a flagged plot identified in March is a sourcing decision; in December it is a blocked shipment.
Make it contractual and keep it alive
Plot data is a delivery item, not a favour: put format ("GeoJSON, WGS84, plot-level, numeric areas"), completeness and update duties into supplier contracts. Farmers join and leave cooperatives every season, so a supply-shed file decays at a few percent a year; the sustainable model is the supplier maintaining a living register you receive per lot, with versions kept in your five-year file. And resist the shortcut of dropping unmapped farmers: exclusion concentrates supply and raises prices, while funding their mapping converts a compliance risk into secured volume. Buyers of Ivorian cocoa and Colombian coffee both have national infrastructure to lean on before mapping anything themselves.