Rules
Part of Can a reporting pack survive contact with its own audience?
Fifteen reporting questions: which ones are worth asking before you build
Reporting requests priced honestly: fifteen questions about the decision, the numbers, the reader and the standing obligation, asked before you build.
Agreeing to build a recurring report commits somebody to work every period for years. The request usually arrives as a small favor, and the commitment is never priced.
These fifteen questions are the ones that price it. Ask them before the first version exists, while the answers are still cheap to hear.
What to take away
- The expensive part of a report is not building it. It is producing it every period, forever, while its readers change and nobody tells you.
- Four of these will be answered with a shrug. That is the finding: a report nobody can scope is a report nobody will use.
- Write the answers down and attach them to the report. Two years later they are the only record of what it was for.
Questions about the decision
1. What decision does this change?
Name the decision and the person who makes it. If the answer is that people want visibility, you have a request for reassurance, which a one-off analysis serves better than a standing obligation.
2. What happens if the number is bad?
There should be an action, and it should differ from the action when the number is good. A report where every outcome leads to the same next step is a record, not a report.
3. What are they doing today instead?
Somebody is already making this decision with something. Find out what. Often the honest answer is that the existing spreadsheet works and the request is really about who maintains it.
4. Would an alert serve better than a report?
If the decision is triggered by a threshold rather than made on a schedule, you want a notification when the threshold is crossed. A periodic report for a threshold decision means the reader either misses the moment or checks constantly.
Questions about the numbers
5. What is the primary number, in one sentence, including its denominator?
If two people write different sentences, stop and settle the definition. Rates hide most of the disagreement, because two people can agree on the ratio and mean different populations underneath it.
6. Which comparison decides whether the number is good?
Prior period, same period last year, plan, or an agreed band. Choose before anyone sees data, because the choice made afterwards is the one that flatters.
7. What counts as a material movement?
A number with no band around it produces commentary on noise. Agree the band, even roughly, and revisit it once you have a few periods of history.
8. When is the data final?
Ask which sources settle late, and by how much. This answer sets the publication date, and it is the single question most often skipped.
9. What will you do when a source fails?
Publish late, publish with a gap marked, or publish an estimate labeled as one. Decide now. In the moment, the pressure is always toward filling the hole quietly.
10. How will restatements be handled?
History gets corrected. Say whether prior periods are restated or the change applies forward only, and where the earlier figure will still be readable. Statistical offices treat this as ordinary practice rather than as an admission, and the published approach to revisions and corrections transfers almost directly into an internal policy.
Questions about the reader
11. Who opens it, and at what moment?
A named person and a trigger: before a Monday meeting, at the start of a shift, when a customer escalates. Vague readership is how a report ends up written for nobody.
12. What does the reader already know that the report should not repeat?
Every period, a report re-explains things its regular readers learned in month one. Cutting that is the fastest way to shorten a pack without losing anything.
13. Who else will see it, and does that change what you can write?
If the pack circulates beyond the operating team, the commentary gets careful. Better to know that now and design two products than to discover it through a softened sentence.
Questions about the obligation
14. Who produces it when the builder is away?
If the answer is nobody, the report has a single point of failure and a hidden manual step somewhere. Find the step now.
15. What would make this unnecessary, and who reviews that?
Set the retirement condition and a review date while the report costs nothing. This is the only moment anyone will agree to one, and a report with no end condition is a permanent tax. A retirement condition is also the first item on the data analysis checklist for any recurring request.
What the answers look like when the report should not exist
Three patterns show up often enough to name.
| Pattern | How it sounds | Better response |
|---|---|---|
| Reassurance | "We just want to keep an eye on it" | A one-off analysis, plus an alert on the one number that matters |
| Displaced argument | Two people give incompatible definitions | Settle the definition first; the report is a symptom |
| Audit trail | "So we have a record" | A stored dataset with a retention rule, not a formatted pack |
None of these is a refusal. Each replaces a permanent obligation with something smaller that meets the real need.
Running this without sounding obstructive
Fifteen questions asked in a row reads as an interrogation. Ask the first two in the original conversation, then send the rest as a short document the requester fills in before a follow-up. Most people complete the easy ten and leave the hard five blank, and the blanks are exactly what the follow-up is for.
Keep the completed sheet with the report. When someone asks two years later why a metric is defined the way it is, the sheet answers it in a sentence.
Related reading on this site
The production discipline that follows a yes is covered in building a report worth keeping. For the equivalent questions asked about a screen rather than a periodic pack, see what to settle before building a page. For questions asked of a number already in front of you in a meeting, see asking about a number in the room.
Definition work sits in writing a metric definition, and the ownership and change control that answers 10 and 14 properly is in data governance.
Common questions
The requester is senior and will not answer fifteen questions.
Ask two: what decision this changes, and what happens when the number is bad. Everything else can be reconstructed later. Those two cannot.
We already have thirty reports that were never scoped. Do we go back?
Not all at once. Run the sheet on the next report that breaks, and on any report whose owner leaves. Both are moments when asking is welcome rather than intrusive.
What if the honest answer to question 1 is that it feeds a board paper?
That is a real decision with a real reader, so it qualifies. Scope it as such: a fixed window, a fixed layout, and a close date, rather than a live page that changes under the person writing the paper.
Nobody will commit to a retirement condition.
Take a review date instead and put it in the report's header. A visible date makes the question askable by anyone, which does most of the work of an agreed condition.
