Chapter 31 Exercises: Communicating Results: Reports, Presentations, and the Art of the Data Story

How to use these exercises: Part A focuses on conceptual understanding — can you distinguish findings from insights and identify audience needs? Part B is applied — you will rewrite, restructure, and critique real communications. Part C involves creating communication artifacts. Part D pushes you toward synthesis and critical evaluation.

Difficulty key: ⭐ Foundational | ⭐⭐ Intermediate | ⭐⭐⭐ Advanced | ⭐⭐⭐⭐ Extension

Part A: Conceptual Understanding ⭐

Exercise 31.1 — Findings vs. insights

For each of the following findings, rewrite it as an insight by adding context, interpretation, and a suggested action.

1. "Customer churn increased 8% in Q3."
2. "The average response time for support tickets is 4.2 hours."
3. "Women are 35% less likely to click on our job ads than men."
4. "The model has an accuracy of 91%."
5. "Sales of Product A peak in November."

Exercise 31.2 — Audience identification ⭐

For each scenario below, identify the primary audience type (technical, managerial, or public/executive), and describe how you would adjust your communication.

1. You are presenting a churn prediction model to the VP of Customer Success.
2. You are writing a blog post about interesting patterns you found in public transit data.
3. You are sharing your analysis notebook with a teammate who will extend your work.
4. You are briefing the school board on student performance trends.
5. You are presenting at a data science meetup about a novel feature engineering technique.

Exercise 31.3 — The Pyramid Principle ⭐

Rearrange the following sentences into Pyramid Principle order (conclusion first, then evidence, then details):

• "We analyzed three years of sales data across 200 stores."
• "Stores that adopted the new checkout system saw 12% higher sales."
• "We recommend rolling out the new checkout system to all locations."
• "The effect was strongest in high-traffic stores (18% increase) and weaker in low-traffic stores (5%)."
• "The analysis controlled for store size, location, and seasonal trends."
• "Customer survey data suggests faster checkout is the primary driver."

Exercise 31.4 — Narrative arc identification ⭐

Read the following data story and label each paragraph as Situation, Complication, Resolution, or Call to Action:

"In 2019, Springfield County achieved its highest-ever childhood vaccination rate: 92%, well above the 85% herd immunity threshold. [] The COVID-19 pandemic disrupted routine health services, and by 2022, the rate had fallen to 76%. Three schools reported outbreaks of preventable diseases. [] Our analysis of clinic-level data reveals that the decline was concentrated in the southern part of the county, where two of three clinics closed during the pandemic and never reopened. The northern district, where clinics remained open, maintained an 89% rate. [] We recommend prioritizing the reopening of at least one clinic in the southern district, with a specific focus on catch-up vaccination drives for children aged 2-5. []"

Exercise 31.5 — Spotting communication mistakes ⭐

Identify the communication mistake in each example:

1. A 45-slide presentation where every slide shows a different chart, with no narrative thread connecting them.
2. "The logistic regression yielded an AUC of 0.847, which, when compared against the baseline model's AUC of 0.723, represents a statistically significant improvement (DeLong test, p = 0.0012)." — presented to a marketing team.
3. A chart showing revenue growth with a y-axis starting at $9.8 million instead of $0, making a 2% increase look like a 40% increase.
4. A report that presents six findings supporting the proposed strategy and omits two findings that complicate it.
5. "Revenue will be $14,782,341.67 next quarter."

Part B: Applied Communication ⭐⭐

Exercise 31.6 — Rewriting for a different audience ⭐⭐

The following paragraph was written for a data science team. Rewrite it for (a) a product manager and (b) a newspaper article.

"We trained a random forest classifier on 18 months of user activity data (n=45,000) to predict 30-day churn. Feature importance analysis revealed that session_frequency (importance=0.31) and days_since_last_login (importance=0.27) were the top predictors. The model achieved an AUC of 0.89 and F1 of 0.82 on the held-out test set. Cross-validation confirmed stability across folds (std=0.02)."

Exercise 31.7 — Writing assertive slide titles ⭐⭐

Convert each topic-based title into an assertion-evidence title:

1. "Q3 Revenue Results"
2. "Customer Satisfaction Survey"
3. "Website Traffic Analysis"
4. "Employee Turnover Data"
5. "Model Performance Comparison"

Exercise 31.8 — Annotating a chart ⭐⭐

You have a line chart showing monthly website traffic for the past 24 months. Traffic was steady at around 50,000 visits/month for the first 12 months, then jumped to 80,000 after a marketing campaign launched in month 13, then gradually declined to 60,000 by month 24.

Write out: 1. An assertive title for this chart 2. Three annotations you would place directly on the chart 3. A one-sentence takeaway you would put below the chart

Exercise 31.9 — Writing an executive summary ⭐⭐

You analyzed employee satisfaction survey data and found: - Overall satisfaction dropped from 7.8/10 to 6.2/10 over two years - The decline was concentrated in the engineering department (7.9 → 4.8) - The top complaint in engineering was "lack of career growth opportunities" - Departments with mentorship programs maintained satisfaction above 7.0 - Engineering has no mentorship program

Write a 150-200 word executive summary for the Chief People Officer.

Exercise 31.10 — Narrative arc construction ⭐⭐

Using the following facts, construct a data story with clear Situation, Complication, Resolution, and Call to Action. Write 4-6 sentences for each section.

Facts: - A city's public library system serves 200,000 cardholders - Digital book checkouts have increased 300% since 2019 - Physical book checkouts have declined 45% in the same period - Three branch libraries are at risk of closure due to budget cuts - Analysis shows that branches in low-income neighborhoods have the highest physical checkout rates - Digital checkout rates are lowest in these same neighborhoods (limited internet access)

Exercise 31.11 — Dashboard design critique ⭐⭐

A colleague shows you a dashboard with the following elements on a single screen: - 12 different charts (bar, line, pie, scatter, area, donut, treemap, bubble, stacked bar, waterfall, radar, and gauge) - No filters or interactive elements - Each chart uses a different color scheme - Charts have descriptive titles like "Sales" and "Revenue" but no annotations - There is no clear visual hierarchy — all charts are the same size - Data source and date are not shown

List at least six specific improvements you would recommend, explaining why each matters.

Exercise 31.12 — Translating numbers ⭐⭐

Rewrite each technical statement in plain language that a non-technical stakeholder would understand:

1. "The model's precision is 0.78 and recall is 0.65."
2. "We used a 70/30 train-test split with 5-fold cross-validation."
3. "The p-value was 0.003, so we reject the null hypothesis."
4. "The R-squared is 0.42."
5. "We applied L2 regularization with lambda=0.01 to prevent overfitting."

Part C: Creating Communication Artifacts ⭐⭐⭐

Exercise 31.13 — Building a narrative notebook ⭐⭐⭐

Using any dataset you have worked with in this course (or the vaccination project data), create a Jupyter notebook that follows the narrative notebook principles from Section 31.6. Your notebook should:

1. Open with a Markdown cell explaining the analysis question and data source
2. Include at least 5 Markdown cells that narrate the analysis (not just code comments)
3. Produce at least 3 annotated visualizations with assertive titles
4. End with a conclusions cell that summarizes findings and limitations
5. Be readable top-to-bottom by someone who has never seen the data

Evaluation criteria: Could a classmate read this notebook without you present and understand what you found and why it matters?

Exercise 31.14 — Designing a slide deck outline ⭐⭐⭐

Create a 10-slide outline for a presentation about your vaccination analysis project, targeted at a county health board (non-technical audience). For each slide, write:

1. The assertion-evidence title (a full sentence)
2. What visual evidence would appear on the slide (chart type, key data shown)
3. What you would say verbally (2-3 sentences of speaker notes)

Exercise 31.15 — Annotated visualization ⭐⭐⭐

Using matplotlib, create a bar chart comparing a metric across 5-8 categories. Apply the following communication principles:

1. Use an assertive title that states the insight
2. Add a reference line showing a benchmark or target
3. Annotate the highest and lowest bars with their values
4. Add a callout text box highlighting the key takeaway
5. Remove chartjunk (unnecessary gridlines, borders, decorations)

import matplotlib.pyplot as plt
import numpy as np

# Example: Vaccination rates by district
districts = ['North', 'East', 'South', 'West', 'Central',
             'Northeast', 'Southeast']
rates = [88, 82, 67, 79, 91, 75, 71]
target = 85

fig, ax = plt.subplots(figsize=(10, 6))

colors = ['#2ecc71' if r >= target else '#e74c3c' for r in rates]
bars = ax.barh(districts, rates, color=colors)

# Reference line
ax.axvline(x=target, color='#2c3e50', linestyle='--', linewidth=1.5)
ax.text(target + 0.5, 6.5, f'Target: {target}%', fontsize=10,
        color='#2c3e50')

# Annotate highest and lowest
max_idx = np.argmax(rates)
min_idx = np.argmin(rates)
ax.text(rates[max_idx] + 1, max_idx, f'{rates[max_idx]}%',
        va='center', fontweight='bold', color='#2ecc71')
ax.text(rates[min_idx] + 1, min_idx, f'{rates[min_idx]}%',
        va='center', fontweight='bold', color='#e74c3c')

# Assertive title
ax.set_title('Three districts fall below the 85% vaccination target,\n'
             'with South district at a critical 67%',
             fontsize=13, fontweight='bold', loc='left')

ax.set_xlabel('Vaccination Rate (%)')
ax.set_xlim(50, 100)
ax.spines[['top', 'right']].set_visible(False)

plt.tight_layout()
plt.show()

Customize this for your own data and insight.

Part D: Synthesis and Critical Evaluation ⭐⭐⭐–⭐⭐⭐⭐

Exercise 31.16 — Communication audit ⭐⭐⭐

Find a real data visualization in a news article, blog post, or social media post. Using the communication checklist from Section 31.12, evaluate it on:

1. Does it have a clear insight (not just a finding)?
2. Is it appropriate for its audience?
3. Are axes labeled and honest?
4. Are key data points annotated?
5. Is uncertainty communicated?
6. Could it be misinterpreted? How?

Write a 200-word evaluation.

Exercise 31.17 — Three versions of one story ⭐⭐⭐

Take a single analytical finding from your project work and write three different communications:

1. A two-sentence Slack message to a data science colleague
2. A one-paragraph email to a product manager
3. A three-paragraph executive brief for the CEO

Each version should communicate the same core finding but adapt language, detail, and framing for the audience.

Exercise 31.18 — Ethical communication dilemma ⭐⭐⭐

You have analyzed employee productivity data and found that remote workers are 15% less productive than in-office workers (measured by tickets resolved per day). However, you also noticed that remote workers handle more complex tickets on average, and when you control for ticket complexity, the gap disappears entirely.

The CEO has asked you to present "the remote work productivity numbers" at next week's board meeting. She has not asked about ticket complexity.

1. What are the ethical considerations in how you present this data?
2. Write two versions of a slide title: one that is technically accurate but misleading, and one that is honest and complete.
3. What would you say if the CEO asked you to present only the headline number?

Exercise 31.19 — From notebook to narrative ⭐⭐⭐

Take an existing analysis notebook from a previous chapter's exercise (or create a simple one). It likely reads as a "lab notebook" — cells run in order, but the narrative is missing.

Transform it into a narrative notebook by: 1. Adding a title and introduction cell 2. Adding Markdown narration before and after each analysis step 3. Cleaning up all code (removing dead code, improving variable names) 4. Ensuring all visualizations have assertive titles and annotations 5. Adding a conclusions cell

Compare the two versions. Which would a colleague rather read?

Exercise 31.20 — The anti-dashboard ⭐⭐⭐⭐

Design a dashboard that is intentionally terrible — one that violates as many dashboard design principles as possible while remaining technically "accurate." Then, for each violation, explain what is wrong and how to fix it.

Include at least 8 deliberate violations (from Section 31.7 and your own judgment).

Exercise 31.21 — Stakeholder simulation ⭐⭐⭐⭐

Work with a partner (or imagine one). One person plays the data scientist presenting vaccination analysis findings. The other plays a skeptical county commissioner who: - Does not understand statistics - Is worried about budget cuts - Has heard "data people always have an agenda" - Needs to vote on clinic funding next month

The data scientist must present three key findings and a recommendation in five minutes. The commissioner asks challenging questions like "Why should I trust this data?" and "Can you guarantee this will work?"

Write out both sides of a realistic dialogue (at least 10 exchanges). Focus on how the data scientist navigates skepticism, translates technical concepts, and maintains honesty about uncertainty.

Exercise 31.22 — Cross-format communication plan ⭐⭐⭐⭐

You have completed a major analysis on student performance in an urban school district. Your findings show that class size has a small effect on test scores, but teacher experience has a large effect. Small classes with inexperienced teachers perform worse than large classes with experienced teachers.

Create a communication plan that includes: 1. A one-sentence headline for each of three audiences (school board, teachers' union, parents) 2. The key chart you would show each audience (describe it) 3. The call to action for each audience 4. One thing you would not say to each audience, and why

Exercise 31.23 — Presentation rehearsal and feedback ⭐⭐⭐

Prepare a 5-minute presentation of any analysis from this course. Record yourself (audio or video). Then watch/listen and evaluate yourself on:

1. Did you lead with the key finding, or build up to it?
2. Did you use jargon that your audience would not understand?
3. Did you read from slides or speak naturally?
4. Did your slides support your words, or duplicate them?
5. Were there any "um," "uh," or filler words that could be replaced by pauses?
6. Did you end with a clear call to action?

Write a one-page self-evaluation with specific plans to improve.

Exercise 31.24 — Communication failure post-mortem ⭐⭐⭐⭐

Research a real-world case where poor data communication led to a bad decision or public misunderstanding. (Examples: the Challenger disaster, the 2016 election forecast misunderstanding, early COVID-19 data visualization problems.)

Write a 300-word analysis that addresses: 1. What was the communication failure? 2. Who was the audience, and what did they need? 3. What could the data communicators have done differently? 4. What general lesson does this teach about data communication?

Exercise 31.25 — The one-page portfolio piece ⭐⭐⭐⭐

Take your best analysis from this course and create a one-page visual summary — a single page that someone could read in 2 minutes and understand your question, approach, key finding, and recommendation. Include:

1. A compelling title
2. A one-paragraph introduction (3-4 sentences)
3. One or two well-annotated charts
4. Three bullet-point key findings
5. A clear recommendation
6. A data source note

This is the format used for policy briefs, conference posters (in miniature), and portfolio pieces. It forces you to distill an entire analysis to its essence.

Reflection

After completing these exercises, consider: Which type of data communication feels most natural to you? Writing? Presenting? Visualizing? Designing dashboards? And which feels most difficult? The communication skills that feel hardest are usually the ones worth practicing most, because they represent your biggest growth opportunity as a data scientist.

Remember: the last mile of data science is communication. An analysis that nobody understands might as well not exist.