Multi-Outlet Operational Checks: Define the Sample

Define the population before selecting anything

Write down the target population in plain terms: what kind of record, from which outlets, over which period. Then state the eligible record unit, meaning what counts as one item to select, and confirm there are no duplicates. List the exclusions and the reason for each, and fix them before selection so they cannot be adjusted once results start to appear.

Freeze a complete frame: a list or register reference that contains every eligible record at a stated time. Keep the excluded records on their own list with their reasons. If the frame is incomplete, say so on the form; a selection drawn from part of a register describes only that part.

Agree the selection rule before inspecting content

Write the rule that chooses records and have it agreed before anyone opens them. For this simple deterministic check, someone else applying the same rule to the frozen list should pick the same records. Preserve the ordering and original selected IDs as soon as they are chosen, and note whether each one was retrieved.

Sampling guidance treats a sample as elements chosen from a defined population and warns about bias, imprecision and inconsistency. Fixing a deterministic rule before inspection reduces discretionary picking after results are seen, but the rule can still favour particular kinds of records. It is not random and does not make the selection statistically representative. How many records a design needs depends on that design; this guide does not recommend a number.

Fix the criterion and the evidence

State the check criterion in one sentence, name the procedure revision it comes from and name the verifier role. For each selected record, write what was observed against the criterion and reference the evidence, using protected record references rather than copies of personal details. Record a fact, not a grade.

Keep the criterion narrow enough that two verifiers would record the same observation. A broad criterion about general completeness invites judgement calls; naming the specific field and the procedure revision keeps the check consistent across outlets.

This example checks only whether an owner role is named. It does not establish that the role is appropriate, that someone accepted the responsibility or that the requisition was approved. Those would be separately defined checks with their own evidence.

Worked example: CHECK-023 purchase requisitions

In this fictional check, the population is internal purchase requisitions dated 1 to 7 November 2026 at the North, South and East outlets. The record unit is one unique requisition. Cancelled drafts were excluded by a rule fixed before selection, and the excluded list and reasons were retained. The frozen register held 8 eligible requisitions per outlet, 24 in total.

Within each outlet, records were sorted by creation timestamp and then by unique ID, and positions 2 and 5 were selected before any content was opened, giving six records. Two per outlet is part of the fiction, not a recommended size. The criterion, from procedure revision 3, is that the requisition names an accountable owner role.

Reading the result

All six selected records stay in the result: four observed meeting the criterion, one observed not meeting it and one not inspected. Report those three counts as they are. Turning them into a rate by dropping S05 would describe a different, smaller selection, and the worksheet holds no percentages for that reason.

Missing evidence is not a positive result. S05 is unknown, not compliant. If it is retrieved later, record that retrieval separately, linked to S05 and dated, and keep the original not-inspected entry. Do not replace S05 with the easiest available South record, and do not replace S02 because it did not meet the criterion.

The result covers the six selected records, of which only five were inspected. They do not establish why S02 was blank, and they say nothing about the remaining 18 requisitions. Estimating a rate for the whole register would need a suitably designed check with specialist input; a simple purposive check like this one establishes observations about the records inspected.

Follow-up without moving the denominator

If someone wants to look at more South requisitions, record that as a separate targeted follow-up with its own references, outside the original sample. Name the owner for each follow-up, and link any corrective action or investigation to a separate action follow-up case rather than recording it here.

If the procedure changes, for example a new revision of the requisition form, start a new check with its own population, frame, rule and criterion. Comparing it with CHECK-023 is reasonable only when the differences between the two designs are written down.

Where the principle comes from

ASQ's page on sampling describes sampling as selecting elements from a defined target population, names bias, lack of precision and lack of consistency as sources of sampling error, and explains that the required sample size depends on the design and its specified conditions. Only those general points are used here. The page's broader claims about randomised samples, its acceptance-sampling standards and any universal sample size are not adopted, and no confidence intervals or population rates are produced.