The costliest error in soil analysis happens in the bucket, not the lab. The rules that make a sample mean something: sample at a consistent time of year (post-harvest to pre-fertilisation); 0–30 cm for field crops, deeper split cores for mineral nitrogen, three depths before planting orchards; one composite sample = 15–25 cores along a zigzag route within a uniform area; avoid field edges, old manure spots and wet depressions.
The catch is knowing what is “uniform”. Satellite management zones show it objectively — so 2–3 well-placed composite samples replace 15 blind ones, and each calibrates the model for the whole field.