[Intermediate] — the half of the job that isn’t technical, and the reason good analyses get ignored while worse ones drive decisions.

The kid version first

You did the analysis, found the answer, and it’s correct. You’re only halfway done. If you communicate it as a wall of numbers, or bury the finding on slide 14, or answer a question nobody asked, it changes nothing — and an analysis that changes nothing has failed, however correct it was. The best analysts are the ones whose findings actually get acted on, and that’s a communication skill as much as a technical one.

This is the note the whole track has been pointing at: the deliverable is a decision, and communication is how the analysis becomes one.

Why this is half the job

A technically brilliant analysis communicated badly loses to a mediocre analysis communicated well — because decisions are made by people, and people act on what they understand and remember, not on what’s most rigorous. The analyst who can make a finding land — clear, memorable, tied to an action — has more impact than the one with better statistics and worse slides. This isn’t a knock on rigour; it’s that rigour is necessary and not sufficient. The data-scientist track opens with the same lesson, because it’s the one both disciplines most underweight early.

Lead with the answer

The single most important habit: start with the conclusion, not the journey.

Analysts love to walk through their process — the data, the caveats, the method, and then the finding. Stakeholders want the opposite: the answer first, then the support if they want it. This is the “BLUF” principle — Bottom Line Up Front:

   BAD:   "So I pulled the orders table, joined it to users, filtered out test
           accounts, segmented by device, and after checking for seasonality... "
           → the exec has stopped listening

   GOOD:  "Mobile checkout is broken in Germany — it's costing us ~£40k/week,
           and it started with Tuesday's release. Here's the evidence."
           → now they're leaning in

The pyramid principle: answer → key reasons → supporting detail. Anyone can stop reading at any level and still have the thing that matters. Put the “so what” in the first sentence, the subject line, the slide title — not at the end.

The “so what” — the test every finding must pass

Before communicating anything, ask: “so what?” — what should someone do differently because of this?

  • “Conversion is 3%.” → So what? (nothing) → not worth reporting
  • “Conversion dropped to 1% on mobile after Tuesday’s release, costing £40k/week.” → So what: fix the release → worth reporting, and it names the action

A finding without a “so what” is trivia. If you can’t articulate what changes because of it, either dig until you can or don’t send it. This single question filters most of what analysts are tempted to report → actionable over interesting.

Data storytelling

A finding is more memorable and persuasive as a narrative than a table:

  • A story has structure — setup (here’s the situation), tension (here’s what’s wrong / what changed), resolution (here’s what to do). It’s how humans process and remember information
  • One chart, one message — each visual makes a single point, with the point stated in the title (“Mobile conversion halved after the release,” not “Conversion by device over time”). The reader shouldn’t have to interpret; you interpret for them → dashboard design
  • Tailor to the audience — an executive wants the decision and its size; a product manager wants the specific broken step; an engineer wants the exact error. Same finding, different framing. Sending the exec the engineer’s version (or vice versa) fails both
  • Anticipate the next question — “why?”, “how big?”, “are you sure?”, “what do we do?” — and have the answer ready

Honesty and avoiding misleading

The analyst has a duty to be honest, including about their own uncertainty — and it’s tempting not to be, because a confident, clean story is more persuasive than a caveated one:

  • State uncertainty and limitations — the sample size, the assumptions, what you couldn’t determine. A finding presented as more certain than it is will eventually blow up, and it takes your credibility with it
  • Don’t mislead with visuals — truncated y-axes that exaggerate, cherry-picked date ranges, dual axes implying a relationship, 3D pie charts. The same chart can tell the truth or a lie; choose truth even when the lie is more compelling → misleading visualisations
  • Say “associated,” not “caused,” unless you’ve earned it → correlation ≠ causation
  • Present the finding, not the one you (or the stakeholder) wanted — resisting pressure to make the data say what someone hoped is the core of analyst integrity

Credibility is the analyst’s real currency. It’s built slowly through honest, reliable findings and destroyed instantly by one confident answer that was wrong — after which nobody trusts your dashboards.

From insight to action — closing the loop

The finding isn’t the end; the decision and its outcome are:

  • Make a recommendation, don’t just present data. Stakeholders often want the analyst’s judgement — “based on this, I’d recommend X” — not a neutral data dump they have to interpret. (Offer the recommendation and the evidence, so they can disagree with the reasoning.)
  • Follow up — did the decision get made? Did it work? Closing the loop is how you learn whether your analysis was useful, not just correct, and it’s how you build the track record that makes the next finding trusted
  • The relationship matters — an analyst embedded with and trusted by a team drives more decisions than a better analyst treated as a report-vending machine

Key insight

Communication is half the analyst’s job, because a correct finding that isn’t understood or acted on has failed — so lead with the answer (bottom line up front), make every finding pass the “so what” test (what changes because of this?), and tell it as a tailored story where each chart states its own message. Honesty is the constraint: state your uncertainty, never mislead with a truncated axis or an unearned causal claim, and present the finding rather than the one someone wanted — because credibility is the analyst’s currency, built slowly and destroyed by a single confident wrong answer. The loop closes not at the finding but at the decision and its outcome.

Source: [reference] — Sep 2026.

↑ Back to Data Analysis