The evidence is not yet something the choice can rest on

The question to ask next

The comparison is agreed. The evidence is missing. Ask the owner: if the comparison came out this way, is that enough to act on, or would you want it checked against something else first? The answer says how much evidence is enough, which is the size of the work.

The evidence exists. What it means, how current and complete it is, or how it was checked is not established. Ask whoever maintains the source: what does this number mean, as of when, and what has it been checked against? Then ask the owner which of those three would change the choice if it were wrong. That one is the check to run first.

Nobody has agreed what evidence would change the choice. Ask the owner, before any analysis: what would you need to see to choose the other option? If nothing would, the decision is made, and the review for question three applies. If something would, write it down; that sentence is the criterion.

I don't know yet. Ask the owner what would make them choose the other option. Then ask whoever maintains the data whether that evidence exists and has been checked. Those are two conversations, and the first has to come first.

The small step

The comparison is agreed. The evidence is missing. Produce the smallest piece of evidence that settles the agreed comparison, with its as-of date and its checks stated, and take it to the owner. Scope out what the comparison does not need. A wider analysis is a later decision, and usually a different one.

The evidence exists. What it means, how current and complete it is, or how it was checked is not established. Run the one check the owner said would change the choice, record what it found, and label the number with what it means, its as-of time, and what it was checked against. A labeled limit is usable. An unlabeled one is a trust failure waiting for the meeting.

Nobody has agreed what evidence would change the choice. Propose a criterion in writing: if this is above that, we do one thing, otherwise the other. Take it to the owner for correction before any analysis runs. A criterion they correct is theirs. One they never saw is yours, and the finding will be read accordingly.

I don't know yet. Write the comparison you think the decision rests on as one sentence and take it to the owner. Take their corrected sentence to the person who maintains the data. Do not start the analysis until both have seen it.

What it costs

Agreeing the criterion and running the checks come before the analysis and delay it, and the owner carries that delay. The checks may show the evidence cannot be relied on for this choice, which narrows what can be recommended. The alternative is a finding that looks usable and is wrong once, which costs more than every check on this list.

What changes the advice

If the criterion is agreed and the evidence checks out, the open question is what the work displaces, which is question five. If the owner cannot name a criterion after two attempts, treat the decision as made and read the review for question three.