Evidence and interview plan

Once the decision question and engagement boundary are agreed, do not begin with vague interviews such as “How is continuity going?” Design evidence collection around assertions whose confirmation or rejection would change the decision. If the firm is considering an alternate warehouse, first-day intake capacity, storage conditions, inventory-data access, and authority to switch matter more than opinions about the plan document. Turn each assertion into a test. Which record, document, observation, or simulation can distinguish it from an impression? Who can explain an exception? What will remain unknown after the test?

Interviews have three different uses. They describe work that a written procedure may miss, explain why a past decision was made, and identify a verifiable record. They are rarely the only performance evidence. Ask the service owner to trace an urgent order from receipt to proof of delivery. Then request an identifiable case, its timestamps, and what was done when a step failed. Ask a night-shift worker the same question. A different answer may expose a shift gap rather than anyone's dishonesty. Record who spoke, when, and about which period or category of orders; a statement about an ordinary day should not become a conclusion about a widespread outage.

Use an evidence matrix with a row for each material question, not a row for each file available. Useful columns are assertion, effect on the choice, primary record, second relatively independent source, data owner, period covered, known bias or omission, verification status, and next step. For example, a written manual-order procedure supports the existence of a method but not its claimed 12-hour throughput. Manual-order logs from an exercise or incident may show transaction count; observing the night shift tests whether trained staff are available. If the log is missing, mark capacity unverified rather than replacing it with a precise-looking verbal estimate.

Two sources are not necessarily independent. A management dashboard may copy a percentage from the same database as a supplier report; agreement between them is not corroboration. A customer-service log contains complaints received, not necessarily customers who could not contact the company. Check data origin, field definitions, exceptions, and time basis. “Order closed” may mean handed to a carrier, not received by a hospital. Measuring delivery from that field moves the endpoint of the metric. Fix the metric definition before debating a strategy based on it.

Write two or three testable explanations for a material disagreement. Here, the bottleneck may have been contact-centre intake, lack of trained staff, or slow backlog processing after a fast platform restore. The separating evidence is call-wait records, shift and training rosters, and stock-allocation timestamps. Do not infer a root cause from coincident events alone. If review finds that only three orders stalled because no authorized allocator was present, the solution may shift from buying another server to changing delegation and shift cover. An evidence plan prevents funding an answer to the wrong question.

Give the sponsor a prioritized evidence plan, not a request for every available file. Start with assertions that separate options and data likely to disappear quickly, such as incident logs or the experience of a departing worker. Protect personal and order data according to organizational authorization and applicable local rules; samples or de-identified records may suffice. If access is denied, record the restriction and its effect on confidence, then propose an authorized alternative test. Close with resolved and unresolved disagreements and a decision on which evidence is sufficient to start impact analysis and which must be checked alongside it. ISO/TS 22317's public introduction links impact analysis to continuity priorities and requirements before strategy selection. The detailed triangulation method taught here is an advisory design for strengthening those inputs, not a formula attributed to the specification.