Exercises — Chapter 9: Visuals and Data

Writing is learned by writing. Most of these ask you to produce or revise real text—captions, data sentences, figure descriptions—not pick a letter. Where a task is open-ended, a self-assessment rubric stands in for an answer key. Selected solutions: appendices/answers-to-selected.md.

Difficulty: ⭐ foundational · ⭐⭐ intermediate · ⭐⭐⭐ challenging · ⭐⭐⭐⭐ extension.

Part A — Analyze This ⭐

For each item, identify what works or what's broken. Name the specific principle from the chapter.

A1. A report includes a full bar chart with a single bar of height 312, gridlines, a legend, and the caption "Figure 1. Total tickets resolved." What's wrong, and what should this be instead?

A2. Classify each caption as Level 0 (non-caption), Level 1 (label), Level 2 (observation), or Level 3 (interpretation): - (a) "Figure 4. Sales by region." - (b) "Figure 4." - (c) "Figure 4. The West region drove 60% of sales—triple any other region—making it the obvious place to add headcount." - (d) "Figure 4. Sales were higher in the West than in the other three regions."

A3. A sentence reads: "Figure 7 shows the distribution of response times across our endpoints." Diagnose the weakness using §9.6, and name the slot in the paragraph this sentence is wasting.

A4. A chart uses 3-D bars on a photographic background of the company's office, with each bar a different neon color explained by a legend. List every chartjunk element and say, for each, what information (if any) it adds.

A5. Two presentations of the same data appear in one report: a table of monthly revenue (12 rows) and a line chart of monthly revenue and a sentence stating each month's figure. What mistake is this, and which one(s) would you keep if the point is "revenue is trending up"?

A6. A caption reads: "Figure 9. The line rises sharply from January to March, plateaus through summer, dips in September, then climbs again to a December peak." What's the problem with this caption even though it's accurate? Rewrite the approach (you may invent the meaning).

A7. An analyst pastes a 20-line, 8-series chart from their Jupyter notebook—default colors, cryptic legend (feat_3, feat_7…), no interpretive caption—into a memo for the executive team. Name the failure (§9.4) and the fix.

A8. A bar chart comparing two teams' output has a y-axis that starts at 90 (not 0), making Team B's bar look roughly four times Team A's when the real values are 96 and 99. Name the error, its category of seriousness, and the fix.

Part B — Revise This ⭐⭐

Rewrite the given text. The scenario is provided; the answer is not.

B1. (Fix the caption.) Turn each label caption into a Level-3 interpretive caption. Invent a plausible, specific takeaway where one isn't given. - (a) "Figure 2. Page-load time before and after optimization." (The data: 1,400 ms before, 180 ms after.) - (b) "Table 3. Benchmark results for three databases." (The data: Postgres 14,000 inserts/sec; MySQL 9,000; SQLite 6,500, on write-heavy workloads.) - (c) "Figure 5. Customer satisfaction scores over four quarters." (The data: 7.2 → 7.0 → 6.6 → 6.1.)

B2. (Lead with the finding.) Rewrite each sentence so the finding leads and the figure becomes a parenthetical citation (§9.6). - (a) "Figure 3 illustrates the relationship between class size and test scores." (Data: smaller classes score higher; ~8-point gap.) - (b) "The chart in Figure 6 displays our monthly active users for the past year." (Data: up 40%, steepest after the March redesign.) - (c) "Table 1 presents the results of our A/B test." (Data: variant B converted at 4.1% vs. 3.2% for control.)

B3. (Fix this chart — describe the redesign.) A team wants to show that one of six microservices is responsible for most of the system's error volume. They've made a 3-D pie chart, six slices, six colors, a legend, percentages on each slice, on a gradient background. Describe—in words—the figure you'd build instead, and write its interpretive caption. (Invent that "Service C" causes 70% of errors.)

B4. (Presentation → interpretation.) Rewrite this warehouse-of-exhibits paragraph as an interpreted argument (§9.7). You may cut figures that don't serve the point and reorder the rest. "This report presents our Q3 reliability data. Figure 1 shows total requests. Figure 2 shows error rate by hour. Figure 3 shows error rate correlated with the nightly batch job. Table 1 shows all scheduled jobs. Figure 4 shows CPU usage. These figures show our reliability in Q3." (The finding you're building toward: the nightly batch job is causing the error spikes and should be rescheduled.)

B5. (Under-captioned table.) A table titled only "Table 2. Survey responses" has columns: Department, Respondents, % Satisfied, % Would Recommend. The striking fact is that Engineering's "% Would Recommend" (38%) is far below every other department (all above 70%). Rewrite the caption and write the one sentence of body prose that should accompany the table.

B6. (Choose the tool.) For each, decide figure / table / prose and rewrite the delivery accordingly (write the sentence, describe the figure, or sketch the table): - (a) Your service's uptime last month was 99.95%. - (b) Daily active users for each of the last 30 days, where the point is "usage is climbing steadily." - (c) The exact per-endpoint latency (p50, p95, p99) for 14 API endpoints, which a later section will reference by number.

B7. (Strip the geometry.) Rewrite this prose to describe meaning, not the chart's shape: "As you can see in Figure 8, the blue line starts low, crosses the orange line around week 6, continues upward while the orange line flattens, and ends well above it by week 12." (Meaning: the new method overtook the old one around week 6 and pulled ahead.)

Part C — Write This ⭐⭐–⭐⭐⭐

A scenario → produce the artifact.

C1. (Caption from raw data.) You have these cohort results: customers contacted by support within 24 hours renew at 88%; those waiting 2–3 days renew at 71%; those waiting 4+ days renew at 52%. (a) Decide figure/table/prose. (b) If a figure, describe it with full alt-text (no chartjunk). (c) Write its Level-3 interpretive caption using the template [finding], shown by [comparison]: [values], which means [implication].

C2. (Describe a figure with alt-text.) Since this book is text, practice the accessibility skill directly. Take any chart you've seen recently (or imagine a scatterplot of study-hours vs. exam-score showing a positive relationship with a few outliers) and write a complete figure description with alt-text that conveys both what it shows and what it means, to the standard used in this chapter's "Figure 9.1b (described)."

C3. (A data paragraph.) Write a 4–6 sentence paragraph for a manager that interprets the following and leads with the finding: monthly churn rose from 2.1% (Jan) to 4.0% (Dec); the cohort breakdown shows slow-onboarding customers (>3 weeks to first value) churn at 22% vs. 3% for fast ones; only 41% of new customers complete onboarding. Recommend an action. Cite figures parenthetically (you may invent figure numbers).

C4. (Two captions, two audiences.) Dana must caption the same cohort chart for Renée (VP of Marketing, non-technical, deciding budget) and for Priya (peer data scientist, validating the work). Write both captions. State in one sentence what differs and why (tie to Chapter 2 audience analysis).

C5. (Fix a real-feeling figure + its prose.) You're handed this draft results passage:

"Figure 1 shows our experimental results. As the figure demonstrates, there are differences between the groups. Table 1 also shows the numbers. Overall the data shows that our intervention had an effect." The actual result: the treatment group improved 18% vs. 4% for control (p small, effect held across subgroups). Rewrite the passage so it leads with the finding, interprets, and cites the figure/table properly—and write a Level-3 caption for Figure 1.

Part D — Synthesis & Critical Thinking ⭐⭐⭐

D1. (Translate for three audiences.) Take one finding—"our nightly batch job causes an 8 p.m. performance spike; rescheduling it would fix it"—and write the figure choice + caption you'd use for each: (a) the on-call engineers (peers), (b) the VP of Engineering (decider), (c) a public status-page postmortem (general/customer). What changes across the three, and what stays constant?

D2. (Find the flaw.) A colleague's slide shows a dual-axis chart: "new signups" (left axis, 0–500) and "marketing spend" (right axis, 0–$50k), the two lines scaled to overlap almost perfectly, captioned "Figure 1. Marketing spend drives signups." Identify at least two problems (one design, one inferential) and explain why the caption overclaims.

D3. (The Challenger redesign, in words.) Using only the verifiable facts from §9.8 and Case Study 2, describe the single clean figure you would build to make the temperature-vs-O-ring-damage relationship unmissable to a time-pressured decision-maker. Specify the axes, what each point represents, the sort order, and how you'd mark the forecast launch temperature. Then write its Level-3 caption. (Keep strictly to the historical record; invent no quotes, dates, or casualty figures.)

D4. (When the convention fights the chapter.) A journal in your field requires "descriptive" captions (state what the figure is, no interpretation) and reserves interpretation for the Results text. Reconcile this with the chapter's "captions interpret" rule: where does the interpretation go, how do you still serve a caption-only reader, and what does this tell you about rules vs. conventions (preview Ch 35)?

D5. (Redundancy audit.) Take a report or paper you have (yours or a published one) and audit it for redundant figures: any figure that says the same thing as the prose or another table/figure. For each, decide keep-the-figure-and-cut-the-prose, keep-the-prose-and-cut-the-figure, or keep-both-because-they-serve-different-readers. Justify each call with theme 6 (every element earns its place).

Part M — Mixed Practice (Interleaved) ⭐⭐–⭐⭐⭐

These mix this chapter with earlier ones. Decide which principle applies before you apply it.

M1. (Ch 9 + Ch 4.) You've written a results section. Run a reverse outline (Ch 4) on it, reading only the first sentence of each paragraph. Several first sentences are "Figure N shows…". What does the reverse outline reveal about your structure, and how does fixing the data sentences (Ch 9 §9.6) also fix the reverse outline?

M2. (Ch 9 + Ch 3.) Apply the "so what?" test (Ch 3) to this caption: "Figure 2. Error rate over time." It passes the sentence-clarity bar (it's clear) but fails another test. Which, and what does the failure tell you about the difference between clear and interpretive?

M3. (Ch 9 + Ch 2.) A finding will be shown to (i) a database engineer and (ii) the CFO. The data is per-endpoint p99 latency for 14 endpoints. For each reader, choose figure/table/prose and a caption style, and justify using K-R-A-C. (Hint: the engineer may want the table; the CFO wants the one number that matters.)

M4. (Ch 9 + Ch 8.) Here are three data sentences in a paragraph that don't flow: "Churn rose to 22% for slow-onboarding customers (Figure 3). Onboarding completion is only 41% (Figure 4). Churn was 2.1% in January and 4.0% in December (Figure 2)." Reorder them using given-new (Ch 8) and the interpret-don't-present logic (Ch 9 §9.7) so they build an argument. Add the transition words and the topic sentence.

M5. (Ch 9 + Ch 4 + Ch 2.) The Challenger case has now appeared in Ch 2 (audience), Ch 4 (structure), and Ch 9 (data display). In a short paragraph, distinguish the three failures the case illustrates—an audience failure, a structural failure, and a data-display failure—and argue which (if any) was the "root" failure, or whether they were one failure seen from three angles.

M6. (Ch 9 + Ch 3 + Ch 4.) Take the warehouse paragraph from B4 (or your B4 answer) and quantify the improvement the way the book does: word count before vs. after, number of figures cited before vs. after, and one sentence on what the reader can now do that they couldn't before.

Part E — Extension ⭐⭐⭐⭐ (optional, Deep Dive track)

E1. (Small multiples.) Research Tufte's idea of small multiples (a grid of small charts sharing axes, one per category, for comparison). Describe a data situation where a single cluttered chart should become small multiples instead, sketch the layout in words, and write the shared caption. When do small multiples beat one combined chart, and when do they just multiply clutter?

E2. (The ethics of the axis.) Build a short "gallery of honest vs. misleading" pairs from public examples or your own: a truncated-axis bar chart vs. its zero-based version; a dual-axis correlation vs. two honest separate charts; a 3-D pie vs. a sorted bar. For each pair, write the one sentence a critic would say and the one-line fix. Then state the principle that unifies all three (preview Ch 38: clear visuals carry ethical weight).

E3. (Reverse-engineer a great figure.) Find a figure you find genuinely excellent (from a paper, The Economist, FiveThirtyEight, a textbook). Reverse-engineer why it works against this chapter's criteria: data-ink ratio, chart-type fit, caption level, whether it interprets, accessibility. Write a one-paragraph critique as if you were the senior editor in this book's voice. Then find one thing you'd still change.

Self-Assessment Rubric (for open-ended items)

For any caption you wrote (B1, C1, C4, C5, D3), score it:

7–8 = excellent; 5–6 = solid, push the so-what; below 5 = re-read §9.5 and rewrite.

For any "fix this chart" item (A4, B3, D2, E2), check: Did you remove all chartjunk? Choose a chart type that fits the relationship? Start bars at zero? Label directly? Make the caption interpret? Ensure color isn't the only channel? Five or more yes = solid.

Go deeper: Case Study 1 — Dana's captions · Case Study 2 — the Challenger display · Quiz