sampling error
Sampling error is the gap between what your sample says and what the whole material actually is, born from the simple fact that you only grabbed a small piece. It is the same reason a single spoonful from an unstirred pot can taste wrong: the spoon was honest about itself, but it failed to capture the rest of the pot.
Formally, sampling error is the part of the total measurement error that comes from the act of taking the sample, separate from the error in the analysis. Even a flawless laboratory measurement of a non-representative sample carries this error in full — the number is precisely the answer for a portion that doesn't match the bulk.
It matters because it often dominates the whole uncertainty, especially for materials that aren't uniform — ores with rich and barren patches, foods with localized contamination. And it has a humbling property: no amount of care in the lab can reduce it, because it was baked in the moment the sample was taken. The only remedies are better sampling — more increments, larger samples, randomized locations.
Testing one shovelful from the top of a 50-tonne ore pile gives 2% metal, but the whole pile averages 3% — the 1% gap is sampling error, untouchable in the lab.
Error born at the shovel, not at the bench.
The total variance of a result is roughly the sampling variance plus the analytical variance. When sampling variance dominates, buying a fancier instrument barely helps — the smart money goes into taking more and better samples, not measuring the few you have more precisely.