Case Study 1: Dana's Chart Captions, Before and After

A composite, fictional-but-realistic scenario. The company, names, and numbers are illustrative; the writing problem is universal.

The situation

Dana Whitfield, a data scientist, has finished the churn analysis we met in Chapter 2. The finding is solid: new customers who take more than three weeks to reach first value in the product cancel about seven times as often as those who get there within a week. The leak is onboarding, not pricing or features. Now she has to put the result in front of Renée Okafor, the VP of Marketing, who will decide whether to move retention budget upstream.

Dana has built one clean bar chart—three cohort bars, churn rate on the y-axis, onboarding-time buckets on the x-axis (under 7 days, 7–21 days, over 21 days). The chart is fine. The question is what she writes under it. Watch the caption fail, then work.

Draft 1 — the label (what most people ship)

[bar chart: 3 bars — 3%, 9%, 22%]

Figure 3. Churn rate by onboarding-time cohort.

This is where most technical writers stop, and it's the most common data-display mistake there is. The caption names the axes. It tells Renée what kind of picture she's looking at. It does not tell her what to conclude.

Here's why that matters for this reader specifically. Renée is a scanner (Chapter 4) and a non-specialist in churn modeling (Chapter 2). She may read only the figures and their captions in Dana's memo—not a word of the body. Faced with "Churn rate by onboarding-time cohort" and three bars of different heights, she has to do the interpretive work herself: notice the bars climb, register how steeply, infer that onboarding time is the lever, and decide that's where the money should go. That's the analyst's job, and Dana just handed it to the VP. Renée, busy, glances and moves on. The finding evaporates at the last inch.

Draft 2 — the observation (better, not enough)

Figure 3. Churn rises with onboarding time, from 3% for customers
who reach first value within a week to 22% for those who take over
three weeks.

Now the caption states the pattern. A caption-only reader at least learns that churn climbs and roughly how much. This is a real improvement—Dana has moved from what it is to what it shows.

But it still stops at description. It says what happened; it doesn't say what it means or what to do. Renée could read this and still think "interesting—and?" The caption hasn't named the cause (onboarding, not price), and it hasn't pointed to the action (where to spend). It describes the chart accurately and leaves the decision unmade.

Draft 3 — the interpretation (the target)

Figure 3. Onboarding speed, not price, drives churn: customers who
reach first value within a week cancel at 3%, versus 22% for those
who take over three weeks — a 7× difference that points to the
first-week experience as the place to intervene.

Now the caption carries the conclusion. In one sentence it does three jobs:

  • Names the cause: "onboarding speed, not price" — it rules out the alternative Renée was probably assuming (we're too expensive) and names the real driver.
  • Quantifies the effect: "3%, versus 22% — a 7× difference" — the load-bearing number is in words, not just encoded in a bar Renée would have to measure.
  • Points to the action: "the first-week experience as the place to intervene" — it tells the decider what the data implies she should do.

A reader who sees only this caption and the chart leaves with Dana's actual finding and its implication. That is an interpretive caption, and it is the difference between a chart that informs a decision and a chart that gets glanced past.

Run it against the template from §9.5 and you can see the machinery:

[Onboarding speed drives churn], shown by [the gap between fast and slow cohorts]: [3% vs. 22%], which means [retention effort should move to the first week].

The body sentence, too

The caption isn't the only place Dana writes about this data. In the memo's body she first drafted:

❌ "Figure 3 shows the relationship between onboarding time and churn."

—the prose version of a label, wasting the paragraph's topic sentence on an announcement. She fixes it to lead with the finding and cite the figure as evidence:

✅ "Customers who take over three weeks to reach first value churn seven times as often as those who get there within a week (Figure 3) — the strongest signal in the data that onboarding, not price, drives cancellation."

Claim + number + (Figure 3). The sentence now carries the analysis; the figure backs it up.

The same chart for Priya looks different

Worth noting, because it's the chapter meeting Chapter 2. For Priya, Dana's peer data scientist validating the work, the right figure isn't the three-bar explanatory chart at all—it's the richer exploratory graphic (the partial-dependence curve, or the raw scatter of all 4,812 accounts), with a terse, technical caption. Priya has the context to read it and the job of finding the hole. Renée has neither. Same finding; two genuinely different figures and two different captions, chosen by audience. The visual analog of "the finding is not the document."

What Dana learned

A figure does not speak for itself. The chart was identical across all three drafts; only the caption changed, and the caption was the whole difference between a result that landed and one that didn't. The skill isn't making the chart—software made the chart. The skill is writing the sentence that tells the reader what the chart means.

The takeaway for your own writing: For every figure you place, don't ask "what is this a picture of?" Ask "what do I want the reader to conclude from it?" — then make the caption say exactly that. Write the caption you'd want if you were the busy reader who reads only the captions.

Discussion questions

  1. Draft 2 was a genuine improvement over Draft 1 but still "not enough." Pin down precisely what Draft 3 adds that Draft 2 lacks—name the two missing elements.
  2. Dana's body sentence and her caption both state the 7× finding. Is that the "redundant figure" mistake (§9.7)? Why or why not? (Hint: who reads each, and do they serve different readers?)
  3. Renée forwards Dana's memo to the CFO unchanged. Does the Figure 3 caption survive that secondary audience? What in it makes it safe (or not) to forward?
  4. Rewrite Draft 3's caption for a public audience—a customer-facing blog post about why the company is investing in onboarding. What stays, what goes?