The analyst in a commercial argument
Analysis rarely arrives in a neutral room. People have positions, budgets and histories with the question, and a finding lands in the middle of that rather than settling it from outside.
The analyst's contribution depends on holding a particular position: informative, not advocating, and unwilling to be quoted selectively.
Why does advocacy cost the analyst?
Because it converts a source of evidence into a party to the argument. Once an analyst is known to argue for outcomes, every subsequent finding is read as motivated, including the ones that were not.
The value of the role comes from being the person whose numbers both sides accept. That is worth protecting even when a particular argument would benefit from more force.
What does contributing without advocating look like?
Stating the finding, stating what it does and does not support, and stating what would change it. Then answering questions rather than pressing a conclusion.
That is less satisfying in the moment and considerably more effective over a year, because it accumulates rather than spending down.
How do you handle selective quotation?
By anticipating it. A finding with a clear caveat in the same sentence is harder to detach than one with the caveat on a later slide, and a headline written by the analyst is harder to replace than an absent one.
Where a finding has already been quoted selectively, correcting it plainly and once is usually enough. Repeating the correction turns it into a dispute.
What about a decision that contradicts the analysis?
It is legitimate. Decisions involve constraints, judgments and information the analysis does not contain, and a decision-maker overriding a finding is exercising their role rather than ignoring yours.
Record it, monitor what you said you would monitor, and move on. A pattern of overrides that turn out badly is a different conversation and it needs the record to exist.
What should an analyst refuse?
To produce a number they do not believe, to strip a caveat that changes the meaning, and to answer a question the data cannot answer without saying so. Those three are the boundary.
A data function's credibility rests on those refusals being reliable, and the first time one is waived the rest become negotiable.
How does an analyst gain standing?
By being right about things that were checkable, by being clear about uncertainty in advance, and by occasionally delivering a finding that is unwelcome. The last is what establishes that the others were not selection.
What should an analyst do when ignored?
Check whether the work reached the decision in time and in a usable form, because that is the most common cause. Excellent analysis delivered late or in the wrong format is not being ignored; it is arriving after the moment.
Frequently asked questions
Should analysts make recommendations at all?
Yes, clearly labelled as judgment and separated from the finding. Refusing to recommend is its own kind of unhelpfulness, and the labelling is what keeps the finding independent of it.
How do I handle pressure to change a number?
Distinguish between a challenge to the method, which deserves engagement, and a request for a different answer, which does not. The first improves the work and the second should be declined plainly and without drama.
What if my analysis is used to justify a decision already made?
That is common and not always improper, since confirmation is a legitimate use of evidence. What matters is that the caveats travel with it, so the decision is made on the actual strength of the finding.
Should I present alternatives I do not favour?
Yes, and fairly. Presenting the strongest version of the case against your own conclusion is what makes the conclusion credible to a sceptical audience.
How much should I simplify?
Enough to be understood and not enough to change the meaning. The test is whether a simplification would lead a reader to a conclusion the full analysis does not support.
What if two analysts disagree?
Resolve it on method before it reaches the meeting. Two internal numbers presented in a commercial discussion damages both and hands the decision to whoever is most persuasive.
Does independence mean staying out of the business?
No — the opposite. An analyst who does not understand the commercial context produces technically sound work that answers the wrong question. Independence is about conclusions, not about distance.
How do I say no to a senior request?
With a reason and an alternative, in writing where it matters. "That cannot be supported by this data; here is what can be, and here is what we would need for the original question" is a professional answer.
What is the single habit that builds standing fastest?
Stating uncertainty before the result is known, then being right about the uncertainty. It is more persuasive than being right about the point estimate.
How should an analyst handle a hostile challenge?
By separating the method question from the conclusion question and answering the first. A challenge to method deserves a substantive answer; a challenge that is really about the conclusion is a commercial disagreement and should be named as one. Industry context on analytical practice is collected by trade bodies such as FMI.
### What should a brand actually do with this?
State the uncertainty before the result is known, in writing, and then hold to it. An analyst who predicted the range and was right about it accumulates standing faster than one who was right about a point estimate by chance.
Then let the record do the arguing. A file of prior findings with their caveats intact is more persuasive over time than any single presentation.
Working with retail data? See NRS Insights data solutions or contact NRS Insights.