How to get the plot geolocation EUDR needs, out of a chain that’s never been asked for it.

Last updated: 23 August 2026

Why collection, not format, is the bottleneck

The geolocation requirements themselves are not complicated: a point for a plot of four hectares or less, a polygon for anything larger, coordinates to six decimal places, packaged as GeoJSON. A competent person can learn the whole specification in an afternoon. What takes months is getting real coordinates for real plots out of a supply chain that was never built to produce them.

That is the work to plan around. The regulation asks for the geolocation of every plot of land where your coffee or cocoa was produced, not the farm as a dot, not the supplier, and certainly not the warehouse. For a single-estate lot that is one boundary. For a cooperative lot it can be hundreds of smallholder plots, each its own geometry, each needing to be tied back to the sacks you bought. Assembling that is a coordination problem long before it is a data problem.

If any one plot behind a lot shows post-2020 deforestation, the whole lot is disqualified, so the plot list must be complete, not a representative sample. Collection is about coverage, not just accuracy.

Start by mapping your chain

You cannot ask the right person for coordinates until you know who that person is. Before any data request, sketch the chain from the plot to your door and find where plot-level knowledge actually lives. It usually falls into one of three shapes:

• Direct from an estate or a single farm. The simplest case: one producer who knows their own boundaries. Here you are asking the grower directly, and the geometry is typically one or a few polygons.
• Through a cooperative or producer organisation. The common case for smallholder coffee and cocoa. The cooperative is your natural counterpart: it knows its members and, increasingly, already holds farm-mapping data. Your request goes to the co-op, which aggregates from its members.
• Through an exporter or trader. The hardest case for visibility. An intermediary sits between you and the plots, and you depend on them to pass through plot-level data they may hold loosely or not at all. This is where you must push hardest, and earliest, and where “we’ll sort it out later” most often fails.

Write down, for each lot or supplier, who holds the plot knowledge and who you will ask. That map is the backbone of the whole exercise, and it doubles as part of the supply-chain description your due diligence needs anyway.

What to ask for, exactly

Vague requests produce vague data. Ask for specific things, in specific units, and say plainly what you cannot use. A good request to a supplier or cooperative covers:

• Coordinates of each plot, not the farm centre or the office. Make it explicit that you need the production plots: the fields the beans were grown on.
• A point with its area in hectares for small plots; a boundary polygon for plots over four hectares. The four-hectare line decides which. Ask for the area figure even on points: if it is missing, the system silently assumes four hectares, which can misstate a small plot.
• Which plots map to which lot. Coordinates you cannot connect back to the coffee or cocoa you actually bought are of little use. Ask for the link (a producer ID, a lot reference) alongside the geometry.
• The production period. Article 9 wants the date or time range of production, so collect it with the coordinates rather than chasing it later.
• How the coordinates were captured. Standing on the plot? Drawn on a map? From an existing farm register? Knowing the source tells you how far to trust it.

Say the unit out loud in every request: “the plot, not the farm and not the supplier.” The most common bad data isn’t wrong coordinates, it’s the right coordinates for the wrong thing.

How a farmer or cooperative actually captures it

Most of your suppliers do not need specialist equipment. For the great majority of smallholder plots (genuinely at or under four hectares) a single point is legitimate, and a point is easy to capture well:

• A smartphone standing in the plot. An ordinary phone’s GPS, read from the middle of the field (not the farmhouse, not the road), gives a point good enough for a small plot. Any basic maps or GPS app shows the latitude and longitude.
• Walking the boundary for a polygon. For a plot over four hectares, someone walks the perimeter with a phone app that records a track, closing the loop back to the start. Many farm-mapping apps used by cooperatives and certification programmes do exactly this.
• Existing cooperative mapping. A great many co-ops have already mapped member farms for certification or traceability projects. That data can often be reused. Check what format it is in and whether it is per-plot.

Give suppliers a simple, written instruction: “stand in the middle of each plot, record the location, note the area, send us the number for each plot with the producer’s name”, and you will get far better data than from a request that assumes they know what you mean. Precision beyond the plot is not the goal; the goal is the right plot, captured honestly.

Data you may already be able to reuse

Before commissioning fresh mapping, check what already exists. Two sources save real effort:

• Certification-scheme data. If your beans are Rainforest Alliance, Fairtrade or organic, the scheme or your supplier may already hold farm polygons or points. It won’t make you compliant on its own (a certificate is not a substitute for due diligence), but the geodata behind it can be a genuine head start, if it is per-plot and carries the area.
• Cooperative or exporter traceability systems. Suppliers investing in EUDR readiness increasingly maintain plot registers. Ask specifically whether they can export per-plot coordinates in a standard format, and in what geometry.

Reused data still has to be checked against the same standard as fresh data (the format traps below apply equally), but it can turn a months-long collection into a verification exercise.

The collection mistakes that sink a filing

Most rejected or unreliable geodata fails for a handful of predictable reasons. Watch for these before they reach the file:

• The office or warehouse pin. Coordinates taken at the co-op building, the wet mill, or the exporter’s yard, not the plot. The single most common error, and an obvious one to an inspector: the point sits on a town, not on farmland.
• One point for a whole cooperative. A single coordinate standing in for dozens of farms. It fails the “every plot” requirement outright and hides exactly the plots that might carry risk.
• Swapped latitude and longitude. GeoJSON records coordinates as [longitude, latitude], the reverse of how people say them. Swap the two and a Colombian plot lands in the Indian Ocean. Silent, and very common.
• An unclosed or self-crossing polygon. A boundary whose last point doesn’t return to the first, or whose edges cross, is not a valid plot geometry and is rejected.
• A missing area on a point. No Area value means the system assumes four hectares, which can quietly overstate a half-hectare plot or push a borderline one across the line where a polygon was required.
• Coordinates with no link to the lot. Perfect geometry that nobody can tie back to the beans you bought is evidence of nothing.

The full mechanics of geometry and the six-decimal rule (including where extra precision can actually invalidate a plot) are in the geolocation requirements guide. The point here is that almost all of these are collection failures, not format failures: they are decided the moment someone stands in the wrong place or asks the wrong question.

Sanity-check before you trust it

Do not treat received coordinates as finished. A few quick checks catch most problems before they become a bad filing:

• Does it land on farmland? Drop each point on a satellite view. If it sits on a rooftop, a road or the sea, it’s wrong, usually an office pin or a lat/long swap.
• Is the area plausible? A “point” standing for a plainly large estate, or an area figure that doesn’t match the geometry, is a flag.
• Is the coverage complete? Count plots against what you know of the lot. A cooperative lot with three plot points is not complete.
• Then screen for deforestation. Only once the geometry is trustworthy does the deforestation screen mean anything, and for shade-grown coffee and cocoa that screen has its own well-known trap.

Collecting it responsibly, and early

Two closing points. First, start now. A geolocation request propagates slowly (from you to an exporter to a cooperative to hundreds of farmers), and each hop adds a season, not a week. The data is far easier to gather ahead of a deadline than in a scramble against one, and nothing you collect early is wasted if a date moves.

Second, collect it with care for the people behind it. Farm coordinates are information about real farmers and their land. Be clear with suppliers about what the data is for, gather only what the regulation needs, and store it securely. Doing this well is also good supplier relations: you are asking cooperatives for real effort, and it helps to explain why and to reuse what they already hold rather than duplicating it.

This is a large part of what we do. We take your origin data in whatever shape it arrives (a co-op’s spreadsheet, a certification export, a folder of phone readings), clean and validate the geometry, flag the office pins and the swaps and the gaps, run each plot through a type-aware deforestation screen that won’t mistake agroforestry for clearing, and assemble the evidence file behind a defensible conclusion. Then, if you want, file the statement in TRACES under your operator identity. What we do not do (what no service honestly can) is take on the legal responsibility, which stays with you as the operator. The full picture is in the complete importer guide; the method is on how it works.