By Isaac Muthui, CEO, Tamalaki Business Network Ltd. Originally published at factoraerp.com.

If you build or integrate systems for coffee processors in Kenya, the EU Deforestation Regulation (EUDR) is a lineage problem. From 30 December 2026, the EU operator that first places coffee on the EU market (usually the importer) must file a due diligence statement that includes the geolocation of every plot the coffee came from, plus the production period and supplier details. Micro and small operators follow on 30 June 2027.

The field data (plot points and polygons) is increasingly there. What usually breaks is the join between that data and the lots a factory, mill and warehouse actually move. Four design decisions matter:

  1. Plots as first-class entities, not grower attributes. A grower can farm more than one plot. Store plot with its own ID, a foreign key to grower, a geometry, area in hectares and a mapping date.
  2. Geometry rules enforced at write time. The guidance summarised by the European Coffee Federation: a GPS point with six decimal places is enough for plots up to 4 ha; plots above 4 ha need a polygon. Data must be exportable as GeoJSON.
  3. Lot composition as a recorded edge, not a reconstruction. When deliveries are pooled into a factory lot, or lots are bulked into a grade, write the edges (input, output, weight) at that moment. Lineage becomes a graph walk.
  4. Production period on the lot. The statement needs a date or time range of production (for coffee, the harvest season). Derive it from the min and max delivery dates and store it with the lot.

A minimal schema

CREATE TABLE grower ( grower_id TEXT PRIMARY KEY, -- same ID used in delivery, payment and mapping systems coop_id TEXT NOT NULL, name TEXT NOT NULL ); CREATE TABLE plot ( plot_id TEXT PRIMARY KEY, grower_id TEXT NOT NULL REFERENCES grower(grower_id), area_ha NUMERIC(8,3) NOT NULL, geom_geojson JSONB NOT NULL, -- Point or Polygon, WGS84 mapped_on DATE NOT NULL, mapped_by TEXT ); CREATE TABLE delivery ( delivery_id TEXT PRIMARY KEY, grower_id TEXT NOT NULL REFERENCES grower(grower_id), plot_id TEXT REFERENCES plot(plot_id), -- NULL = unknown plot: treat as a gap delivered_on DATE NOT NULL, kg_cherry NUMERIC(10,2) NOT NULL ); CREATE TABLE lot ( lot_id TEXT PRIMARY KEY, stage TEXT NOT NULL, -- 'factory', 'milled', 'grade', 'bulk' season TEXT, produced_from DATE, produced_to DATE ); -- Edges: deliveries into factory lots, and lots into downstream lots CREATE TABLE lot_input ( lot_id TEXT NOT NULL REFERENCES lot(lot_id), delivery_id TEXT REFERENCES delivery(delivery_id), input_lot_id TEXT REFERENCES lot(lot_id), kg NUMERIC(10,2) NOT NULL, CHECK ((delivery_id IS NULL) <> (input_lot_id IS NULL)) ); 

Validating plot geometry

def validate_plot(geom: dict, area_ha: float) -> list[str]: errs = [] t = geom.get("type") def six_dp(x): return len(repr(float(x)).split(".")[-1]) >= 6 if area_ha > 4 and t != "Polygon": errs.append("plot > 4 ha must be a Polygon") if t == "Point": lon, lat = geom["coordinates"] if not (six_dp(lon) and six_dp(lat)): errs.append("point needs 6 decimal places") elif t == "Polygon": ring = geom["coordinates"][0] if len(ring) < 4 or ring[0] != ring[-1]: errs.append("polygon ring must be closed with >= 4 positions") else: errs.append(f"unsupported geometry {t}") return errs 

Add overlap checks between plots (in PostGIS, ST_Overlaps on a geometry column) and flag implausible areas against the recorded area_ha.

Walking the lineage

Given a bag's grade lot, a recursive query collects every delivery, and so every plot, upstream:

WITH RECURSIVE up(lot_id) AS ( SELECT 'GRADE-AB-2026-0141' UNION SELECT li.input_lot_id FROM lot_input li JOIN up ON li.lot_id = up.lot_id WHERE li.input_lot_id IS NOT NULL ) SELECT d.delivery_id, d.grower_id, d.plot_id, d.delivered_on, li.kg FROM up JOIN lot_input li ON li.lot_id = up.lot_id AND li.delivery_id IS NOT NULL JOIN delivery d ON d.delivery_id = li.delivery_id; 

Any row with plot_id IS NULL is a traceability gap you must resolve before the lot goes into an EU contract. Turn the distinct plots into a GeoJSON FeatureCollection for the buyer, with MIN(delivered_on) and MAX(delivered_on) as the production period.

Things that bite in practice

  • ID mismatch. Cooperative grower numbers, payment IDs and mapping-tool IDs often differ. Build the crosswalk once and enforce one ID.
  • Multi-plot growers. If the delivery only records the grower, all of that grower's plots end up in the lot's plot list. Capture plot at intake where you can.
  • Mixed lots. Decide whether unmapped deliveries are excluded from EU-bound lots, or tracked as a share. A lot with unknown inputs is hard to sell into the EU after the application date.
  • Bulking and blending at the mill. New references must keep their input edges. This is where most paper systems lose lineage.
  • Retention. Keep what you sent, when and to whom, and the buyer's statement reference, for at least five years.

Context and sources

Kenya is classified as low risk under the EU country benchmarking, which allows the operator simplified due diligence (no risk assessment or mitigation unless there are signals of risk). It does not remove the need for plot geolocation. In July 2026, the Kenya Coffee Platform told the EUDR Learning Community for Africa that about 90 percent of active coffee farmers had been mapped, and that the main remaining gap is traceability from cooperative level down to individual plots. That gap is a data model.

The full 12-point operations checklist for cooperatives, mills and exporters, with an FAQ, is on the FactoraERP blog: https://factoraerp.com/en/blog/eudr-traceability-checklist-coffee-processors-kenya.