Case Study 1: One Finding, Three Registers — Dana's Churn Result

A composite, anonymized scenario built from the kind of cross-audience communication a data scientist does weekly. Dana Whitfield and her stakeholder, VP of Marketing Renée Okafor, recur across this book. Names and numbers are fictional but realistic.

The situation

Dana Whitfield has finished a piece of analysis: customers who contact support in their first 30 days churn at roughly twice the rate of those who don't. It's a solid finding — the difference held across two quarters of data and survived a controls check. Now she has to communicate it in three places, to three audiences, on the same afternoon:

  1. A methods appendix in a formal analysis report that a data-science peer will scrutinize.
  2. A line in an internal blog post that the broader company — non-technical — will skim.
  3. A Slack message to Renée Okafor, the VP of Marketing, who asked her about it directly.

Same finding. The temptation — and the mistake — is to write it once and paste it three times.

The mistake: one register for all three

Dana drafts the finding in careful, formal, analyst register and reuses it everywhere:

"Analysis of two quarters of customer data indicates a statistically significant association between early support contact (within the first 30 days post-acquisition) and elevated churn propensity, with affected cohorts exhibiting an approximately twofold increase in attrition relative to the control population (p < .01)."

In the methods appendix, this is correct — a peer wants exactly this precision. But pasted into the company blog post, it's a wall: "churn propensity," "attrition relative to the control population," and "p < .01" stop a non-technical reader cold, and the finding everyone could act on dies in jargon. Dropped into the Slack message to Renée, it reads as bizarrely stiff, almost evasive: a VP who asked a quick question gets a paragraph of statistical hedging and has to dig for the one thing she needs ("so… what do I do?"). One register, three audiences, two misfires. The facts were right every time; the register was wrong twice.

The fix: set the register for each audience

Dana writes the finding three times, holding the facts constant and turning the register dial for each reader and genre.

1. Methods appendix (formal):

"Early support contact — defined as any support interaction within 30 days of acquisition — was associated with an approximately twofold increase in 90-day churn (hazard ratio ≈ 2.0, p < .01). The association held across both quarters analyzed and after controlling for plan tier and acquisition channel; the design is observational, so the relationship is correlational, not causal."

Formal register is correct here: technical vocabulary, the writer invisible, and — crucially — the honest hedge preserved ("correlational, not causal"). A peer needs that hedge; stripping it would be dishonest. Note it's formal without being bloated (Chapter 3): no "it is important to note," no nominalized fog.

2. Internal blog post (neutral):

"We found a strong early-warning sign for churn: customers who contact support in their first month are about twice as likely to leave within 90 days. (It's a correlation, not proof that support contact causes people to leave — but it's a useful flag.)"

Neutral register: plain words ("about twice as likely to leave"), a contraction, the writer present but restrained. The jargon is gone, but — this is the skill — the hedge survives in plain language ("a correlation, not proof"). Dana didn't strip the honesty when she stripped the jargon; she translated it. The finding is now something the whole company can grasp and act on.

3. Slack to Renée (informal):

"hey Renée — that churn thing: customers who hit support in their first 30 days leave at ~2x the rate of everyone else. solid across both quarters. it's correlation not causation, but it's a strong flag — worth targeting first-month support experience. want the full writeup?"

Informal register: lowercase, a fragment or two, fast. But notice three things it does not drop: the number (~2x), the hedge (correlation not causation), and — what Renée actually needs — the implication ("worth targeting first-month support experience"). The register is casual; the rigor and the usefulness are intact.

Why the three-register version wins

  • The facts are identical across all three. The twofold rate, the 90-day window, the both-quarters robustness, and the correlation-not-causation caveat appear in every version. Dana adapted the clothing, not the content — which is the whole skill of register.
  • The hedge survives every translation. This is the subtle, important part. The lazy move is to keep the hedge in the formal version and drop it from the casual ones ("it's a correlation" feels too fussy for Slack). Dana keeps it everywhere, because dropping the hedge would change the facts — it would turn an honest "associated with" into a dishonest "causes." Preserving calibrated certainty across registers is an ethical act (Chapter 38), not a stylistic flourish.
  • Each version leads with what its reader needs. The peer needs the method and the caveat; the company needs the plain takeaway; Renée needs the implication and an offer of more. Register and audience-fit (Chapter 2) work together — the dial isn't only formality, it's also what comes first.
  • No version is "dumbed down." The blog and Slack versions are not less precise than the appendix; they're equally precise, in vocabulary their readers share. "About twice as likely to leave" is exactly as true as "hazard ratio ≈ 2.0." Clarity is audience-relative (Chapter 3), and so is register.

The takeaway

Dana wrote one finding three ways in fifteen minutes, and each landed because she treated register as a dial she sets, not a default she pastes. The threshold concept of this chapter — tone is a choice you make, not an accident — is exactly what separated the misfiring "write once, paste thrice" draft from the version that reached all three audiences. And the discipline that made it honest rather than merely fluent was preserving the hedge: the register changed, the certainty didn't.

Try it: take a finding or recommendation from your own work and write it three ways — for a peer (formal), for a general internal audience (neutral), and for a busy decision-maker in chat (informal). Lay them side by side. Confirm two things: the facts are constant, and any honest hedge survives all three. If the casual version quietly became more certain than the formal one, you didn't change the register — you changed the truth.