Key Takeaways — Chapter 27: Writing About Data
The summary card. Read this to re-ground before the next chapter or a writing task.
The one idea
A finding is not a conclusion until it answers "so what?" You did the analysis forward—question, method, result—so you write it forward, leading with the apparatus. But the decision-maker needs it backward: the finding first, the method demoted, and above all the recommendation—what to do about the number. An analysis that stops at the number has stopped one sentence too early and handed its hardest interpretive step to the reader least equipped to take it. Cross this threshold and you stop asking "what did I find?" and start asking "what should they do because of what I found?"
Why most data writing fails
The analyst writes in the order they did the work; the reader needs the opposite order. Method-first is right for a peer auditing your model and wrong for everyone else. Three recurring failures: the methodology wall (the reader climbs over the method to reach anything usable), the finding with no "so what?" (a number dropped on the reader's desk), and the uninterpreted exhibit (a chart labeled, not interpreted—Chapter 9's warning).
The data-analysis memo
Four parts, ordered for the reader: the question (in their terms), the findings (lead with the answer, in their units), the recommendation (the part most memos omit), the method (enough to trust + the honest caveats, demoted below the finding and sized to the reader). One page. If it's longer, the method has crept up and swallowed the findings.
Findings-first vs. recommendation-first
- Recommendation-first when the reader's job is to act (approve, choose, ship). Lead with the action; the finding becomes the "why."
- Findings-first when the reader's job is to understand and judge for themselves (a peer, a regulator).
Both beat method-first. The order follows the reader's purpose (Chapter 2's dial). When unsure, findings-first is the safer default—it never feels like you're getting ahead of the evidence.
The "so what?" ladder
Push every finding down three levels: observation (what the data says) → interpretation (what it means) → recommendation (what to do). Only Level 3 is actionable. Quantifying the stakes in the reader's currency (dollars, customers, risk) is often what turns Level 2 into Level 3. The limit: the recommendation must be supported by the finding—if the data shows what but not why, name the gap; never smuggle a hunch past the reader.
Executive summary for data
Lead with the insight, not the methodology. Quantify the stakes, name confidence in a phrase, and point to the method rather than reciting it. The standalone test (Chapter 20): could the reader decide from the summary alone? Method-first summaries fail it every time.
Dashboard text
Interpret in tiny spaces. Titles state the takeaway or the question (not the topic). Labels carry units and a comparison ("vs. target"). Tooltips explain anomalies and define metrics. Wherever a number appears, tell the viewer what it means and whether it's good or bad. And because you don't rewrite a dashboard daily, the interpretation must stay true as the data moves—title for the question, or use a status indicator, never a fixed claim that's false next week.
Themes this chapter surfaced: #2 audience-is-everything (central) · #5 structure-serves-the-reader · #3 clarity-isn't-the-enemy-of-precision.
Threshold concept: A finding is not a conclusion until it answers "so what?"—the number is an observation; the recommendation is the conclusion.
Anchor paid off: Dana Whitfield's churn memo, rebuilt method-first → findings-first → recommendation-first (the device set up in Chapter 2).
Feeds forward to: Ch 28 (the same finding for the public), Ch 32 (diagrams), Ch 37 (the data one-pager / policy brief), Ch 38 (the ethics of overclaiming).
Back to: Chapter 27 · Exercises · Quiz · Further Reading