sub-sampling
Sub-sampling is taking a smaller portion out of a sample you already have — narrowing down from the bucket you collected to the few grams you'll actually weigh. It's the second cut: you've brought home a big bag of mixed nuts, and now you scoop out one small handful, trusting that handful to taste like the whole bag.
In the lab, sub-sampling reduces a bulk sample to a manageable laboratory portion, or splits a laboratory sample into the small test portions an instrument needs. It is done after collection, and each step must keep the smaller portion representative of the larger one — which is why thorough mixing usually comes first, and why techniques like coning-and-quartering or riffle splitting exist to divide a heap fairly.
It matters because the analyte you finally measure may be just a fraction of a gram drawn from kilograms collected — and every reduction is a fresh chance to introduce bias if the material isn't uniform. The honest caveat: the smaller the test portion relative to the particle size and patchiness of the material, the larger the sub-sampling error, so coarse or uneven samples must be ground finer before they can be split small.
A 5 kg field sample of grain is poured into a cone, flattened, and quartered; opposite quarters are discarded and the rest re-mixed and re-quartered until just 50 g remains for the lab.
Coning-and-quartering shrinks a sample while keeping it representative.
Sampling and sub-sampling sit at different scales of the same chain. Sampling takes the original sample from the world (the river, the ore pile); sub-sampling whittles that sample down to test portions in the lab. Bias can creep in at either step, so each needs its own care.