Check the frame order before selecting every kth customer

Freeze the unit and the ordered frame

Name the unit the study intends to select. An account, purchase and person are different units even when all three appear in a customer system. Identify the approved eligibility cutoff and create a versioned frame reference. Resolve repeated appearances under the design owner’s rules before drawing. Removing duplicate-looking entries after selection can change the selection process, while retaining several orders for one account can give that account opportunities the study did not intend.

Record exactly how the list is ordered, including tie-breaking fields. A spreadsheet that sorts differently after a refresh may produce a different selected set from the same numerical start. Keep contact information in its approved restricted location; the design review usually needs internal unit references and permitted classification fields, not a copy of everyone’s details. Preserve the order as part of the sampling evidence, rather than treating it as disposable presentation.

Inspect the interaction between order and interval

Statistics Canada describes systematic sampling through a random start and fixed interval, and warns that periodic frame patterns can coincide with that interval.

Ask what repeated structure the operational list contains. A billing export might repeat several product rows within each account, or concatenate regularly sized processing batches. Examine how candidate starts encounter those patterns using synthetic or appropriately restricted frame summaries. Do not select the start that happens to produce the most attractive balance. If a concerning pattern appears, have the analyst revise the ordering or design before the draw, documenting why the change addresses the problem.

Keep this review distinct from checking whether the frame covers the intended population. A carefully spaced draw cannot reach customers absent from the list. Conversely, good coverage does not establish that the list’s internal order is harmless. The specification should retain both limitations without turning this lesson into a claim that a mechanically executed sample is automatically representative.

Specify the draw and its exceptions before running it

Have the analyst document the interval, permitted starting positions and random-start procedure. Record how the plan handles a frame length that does not fit a simple whole-number interval, and whether the intended sample count is fixed. Do not improvise rounding, wraparound or a second pass when the spreadsheet produces an inconvenient number of records. Those choices belong to the design, because they can alter inclusion chances.

Run a small synthetic rehearsal with stable unit references and a visibly ordered list. Check the first selection, interior steps and end boundary against the specification. Preserve enough evidence to reproduce the authorized draw, including the frame version and relevant randomization record. Keep subsequent operational status separate: an unreachable selected unit must not quietly be exchanged for the nearest available record unless the approved design explicitly provides a defensible procedure.

Release the selected set without rewriting its history

Compare the final selected references with the frozen frame and the approved selection output. Check that sorting, deduplication or export filters have not changed the set. Hand fieldwork staff the permitted contact material and the rules for recording noncontact, refusal and eligibility questions. The researcher should be able to distinguish a unit never selected from one selected but never interviewed.

Provide the analyst with the frame order, selection specification and any deviations, not merely the completed responses. If a mistake is discovered, hold the affected release and obtain a documented design decision rather than choosing a replacement that makes the sample look balanced. The accepted result is a reproducible selection path with known limits. Later participation and measurement quality remain separate sources of uncertainty.

Make the reasoning concrete.

A fictional customer study uses an account export assembled in repeating operational batches. A proposed interval lands on the same batch position repeatedly in a synthetic rehearsal.

The team chooses the second route and retains the revised order and draw evidence. It does not claim that this resolves later nonresponse.

Take the questions into your next brief.

Use these prompts in your own notes. Nothing is submitted from this page.

1. Unit

   What appears once on the sampling frame?

2. Order

   Which sort and tie-break rules produced this exact list?

3. Pattern

   What repeated structure could align with the interval?

4. Draw

   How are the start, interval and end boundary specified?

5. Handover

   How are selected-but-uninterviewed units preserved?

Watch for the wrong conclusion.

• Choosing a start after inspecting which sample looks best.
• Rounding or wrapping the interval without a design decision.
• Replacing a selected nonrespondent with a nearby convenient record.

Sources & scope

Workflows, worksheets and fictional examples are original editorial analysis. Sources support only specifically attributed principles. Platform guidance can change; legal requirements depend on jurisdiction and context. No measured results or named human reviewer are claimed.

• Statistics Canada: 3.2.2 Probability sampling →Primary statistical methodology · Supports: Random start, fixed interval and periodicity risk in systematic sampling.