Writing an analysis brief that gets answered
A large share of unsatisfying analysis is the result of an unclear request. The analyst answered a question; it was not the question the requester had in mind, and neither party discovered that until the work was presented.
A brief fixes it, and a good one is short.
What does a brief need?
The decision the analysis will inform, the question in one sentence, the population, the period, the measure, and what result would change the decision. Six elements, and the last one is the most useful and most often missing.
If no result would change the decision, the analysis is not needed. Establishing that at the briefing stage saves everybody.
Why does the decision matter more than the question?
Because it disciplines the scope. An analysis framed around a decision has a natural boundary; one framed around a topic expands until somebody stops it, usually at a deadline and usually incomplete.
It also determines the precision required. A decision that holds across a wide range needs less analytical effort than one that hinges on a small difference.
What should be specified about the data?
Which source, which universe, which definitions, and which version. Specifying the source prevents the common outcome where two analyses of the same question use different data and disagree for reasons nobody can explain afterwards.
The version matters where history can be restated. An analysis reproduced six months later against current data may not reproduce.
What about the deadline?
It belongs in the brief and it should be honest, because a deadline determines method. An answer needed tomorrow and an answer needed next month justify different depth, and an analyst who knows which is being asked can propose the right approach.
An unrealistic deadline attached to a demanding question produces a rushed answer presented with unearned confidence.
What should the analyst push back on?
Questions that cannot be answered with the available data, questions where the result would not change anything, and questions whose framing presumes the answer. All three are common and all three are better raised at briefing than at presentation.
That pushback is a service rather than an obstruction, and a team where it is welcome produces better work.
How should the output be specified?
By the form the decision needs: a number, a chart, a recommendation, a list. Agreeing this prevents the mismatch where an analyst delivers a thorough exploration and the requester wanted one figure.
Stating the intended audience helps too, because the same finding is presented differently to a category team and to a board.
What belongs in the record afterwards?
The brief, the method, the result, and the decision taken. That record is what stops the same question being re-asked in six months and what allows an outcome to be reviewed against what was expected.
Frequently asked questions
How long should a brief be?
A page at most, and usually a short one. Length is not a virtue here; the six elements can be stated in a few lines each, and a longer document tends to be read less carefully.
Who should write it?
The requester, with the analyst reviewing and challenging before work starts. A brief written by the analyst on the requester's behalf frequently encodes the analyst's interpretation, which is the problem it was meant to solve.
What if the question is exploratory?
Say so explicitly and set a time box rather than a deliverable. Exploratory work is legitimate and it needs a different contract from a decision-support analysis, because the output cannot be specified in advance.
Should the expected answer be stated?
Stating a prior is useful and stating a required answer is corrupting. "We expect to see X" invites a check; "we need to show X" invites a different kind of analysis entirely.
How do I specify a threshold?
In the units of the measure and tied to the decision: what size of effect would change what you do. Agreeing it in advance removes the negotiation that otherwise happens once the result is known.
What about follow-up questions?
Expect them and treat each as a new brief if it is substantial. The alternative is scope drift, where a one-week piece becomes a quarter of work with no clear end.
Does this apply to recurring reporting?
Yes, and recurring reports benefit most, because they persist long after the original need. A brief with a review date prevents reports being produced for years after anybody stopped reading them.
Who should see the record?
Whoever might ask the question again, which is usually a wider group than the original requester. A searchable record of prior analyses is one of the cheapest productivity improvements a commercial team can make.
What is the single most valuable line in a brief?
What result would change the decision. It converts a request for information into a request for evidence, and it occasionally reveals that no analysis is needed at all.
### What should happen when a brief cannot be met?
Say so at briefing, with what could be answered instead and what the original question would require. That conversation takes minutes and it prevents a project that delivers a disappointing answer weeks later.
Record the exchange. The same question tends to return, and a record of why it could not be answered last time saves repeating the discovery. Industry context on research practice is collected by trade bodies such as FMI.
Working with retail data? See NRS Insights data solutions or contact NRS Insights.