Blind-tested against the lab: 85% accuracy, from space
Eurofins took geo-referenced wheat samples from the field. Messium predicted the nitrogen concentration and biomass at those exact coordinates - blind, using only satellite data. The verdict: 85%+ accuracy against certified lab results. The results were shared openly with farmers and over 15 corporate partners.
The test
Eurofins, the global testing and laboratory services company, collected geo-referenced tissue samples from real wheat fields across multiple sites. Each sample was tested in their certified laboratories for nitrogen concentration and biomass - the standard against which Messium had to compete.
Messium was given the GPS coordinates only. Using hyperspectral satellite data from passes within 10 days of the sampling, Messium returned a predicted nitrogen value and biomass estimate for each location. No lab values, no ground truth until after the predictions were submitted.
Why blind matters
Accuracy is only meaningful if you cannot cheat it. A blind test removes every possibility of overfitting, cherry-picking, or tuning to expected results. The 85%+ number is what Messium's models produced when they genuinely did not know the answer in advance.
A standard lab tissue test costs £8–12 per sample and takes 5–10 working days to process. Messium returns a comparable result for every hectare in a field, every week, with no sampling required.
Shared transparently
The full results - including the cases where Messium's prediction was outside the acceptable range - were shared with farmers and with more than 15 corporate partners across agronomy, finance, and food supply chains. This was not a press release. It was a technical validation document, with the data attached.
Validation work like this is how Messium continues to build the evidence base - and why 85%+ is a number backed by a certified third party, not a marketing claim.