Next decision

Act on mixed or inconclusive findings

Choose a focused mitigation, retest, limited pilot, or decision to stay instead of forcing a false winner.

Audience: Decision owners resolving trade-offs after an assessment.
Question: What should the team do when the candidate improves some measures and worsens others?

Rank findings by decision impact

A blocking policy failure matters more than a small average improvement. Classify each finding as blocking, material, useful context, or unresolved before choosing the next action.

Choose the smallest action that can change the answer

Do not repeat the entire assessment when one risk drives the uncertainty. Change the relevant guardrail, case, configuration, or repetition plan and keep the rest of the method stable.

01
Mitigate a reproducible failure when the cause and control are clear.
02
Gather more evidence when variation or sample size drives the uncertainty.
03
Run a bounded pilot only when the assessed risk is contained and reversible.
04
Stay with the baseline when the candidate’s upside does not justify the remaining risk.

Write the condition for changing course

Record the evidence that would reverse the current recommendation. This makes a stay or investigate decision useful instead of indefinite.

Worked example

Example: mitigate, then retest

The candidate is cheaper and usually clearer, but fails a credit-authorisation rule. Add an exact guardrail, rerun that case with more repetitions, and permit a low-risk pilot only if the blocking failure reaches zero in the agreed retest.