Why sampling matters as much as the test

The laboratory only sees what it receives

A truck, warehouse lot or field is larger and more variable than the portion placed in a vial. Sampling is the process used to make that portion informative about the defined lot. If the sample is biased or the lot is poorly defined, a perfectly executed test can still support the wrong conclusion.

This is why representative does not mean convenient. Scooping from the cleanest top bag or accepting a sample prepared by an interested seller may exclude the very variation the decision is supposed to assess.

Analytical accuracy cannot compensate for a sample that never represented the lot.

First define the decision and the lot

A sampling plan begins with purpose: regulatory compliance, supplier approval, process control, investigation or surveillance. It then defines the lot—the quantity considered to share relevant production or handling conditions. A very broad lot may hide important variation; a tiny lot may make the programme impractical.

The parameter matters too. Microbial hazards, pesticide residues, moisture and foreign matter may have different patterns of distribution and require different approaches.

Increments build a more credible picture

Plans may draw multiple increments from defined positions or randomly selected units and combine them into a composite where appropriate. The number and location of increments, sample mass and method of mixing should be written down. Heterogeneous bulk materials deserve particular care.

Composite sampling can estimate an average condition efficiently, but it may dilute a localised problem and remove the ability to identify which unit failed. Individual-unit sampling preserves that information at higher cost. The plan must match the risk and decision.

Sampling plans manage two kinds of error

A finite sample creates the possibility of accepting a lot that does not meet the requirement or rejecting one that does. Codex sampling guidance frames plans around fitness for purpose and the risks to consumers and producers. More samples are not automatically better unless the decision rule is coherent.

The plan should state the acceptance criterion before results are known. Otherwise, a business can unconsciously redesign the question around the answer it prefers.

People and protection remain part of the method

Sampler training, clean equipment, suitable containers, seals, transport conditions and chain of custody protect the work after selection. FSSAI’s Food Safety Officer manual describes representative sampling and practical sample handling because these details affect reliability.

Conflicts of interest also matter. Independent or jointly witnessed sampling can be appropriate when the commercial stakes are high. The identity of the sampler should be visible on the record.

What a shopper should ask

A brand need not publish a statistical textbook beside every report, but it should be able to explain which batch was sampled, who sampled it, whether the plan was representative and what decision the result supports. “One random sample” is not a plan unless random has an actual selection method.

Once sampling is visible, a lab report becomes less like a magic certificate and more like a carefully framed photograph. You know what is inside the frame—and what may still sit outside it.