Chapter 9: Storytelling with Data: Narrative Structure for Visual Communication

"Data rarely speak for themselves. More often than not, they need to be helped. They need a guide. And that guide is the data storyteller." — Cole Nussbaumer Knaflic, Storytelling with Data

For eight chapters, we have been preparing for this one. We established that visualization is argument (Chapter 1). We learned the perceptual machinery the eye uses to read charts (Chapter 2) and the color choices that support or undermine that machinery (Chapter 3). We confronted the ethical choices behind every axis range and baseline (Chapter 4) and learned the framework for matching chart type to question (Chapter 5). We decluttered (Chapter 6). We added the words (Chapter 7). We composed multiple charts into coherent figures (Chapter 8). Each chapter added a layer to the craft, and each chapter was necessary — you cannot tell a story with a cluttered, untitled, badly composed chart any more than you can tell a story with ungrammatical sentences.

Now comes the question this book has been building toward: what do you do when one chart, or even one figure, is not enough?

Most real-world data communication is not a chart. It is a sequence of charts — a slide deck, a report, an article, a dashboard, a presentation — that walks a reader through a complete argument. The argument has a beginning, a middle, and an end. It has a main idea and supporting evidence. It has an audience with specific knowledge, specific concerns, and a specific reason to care about the data. The sequence of charts is the narrative, and the quality of the sequence is as important as the quality of any individual chart. This is what "data storytelling" means: the deliberate arrangement of charts, text, and transitions into a coherent narrative that takes a reader from "what is this about?" to "so what should I do?"

This chapter is the capstone of Part II. It does not introduce new chart-design techniques — it uses the techniques from the previous eight chapters to build something larger than any single chart. By the end of the chapter, you will have a framework for structuring data presentations as narratives, a set of practical techniques for guiding a reader through a sequence of charts, and a vocabulary for the ethical and practical decisions that separate effective storytelling from cherry-picking.

A word about the ethical frame. Chapter 4 argued that every chart is an editorial and that the maker's choices shape the reader's interpretation. Storytelling amplifies this responsibility. A single chart makes a single claim; a story makes a whole sequence of claims, building an argument step by step. The chart maker who tells stories has more opportunities to distort and more opportunities to clarify, and the discipline of honest storytelling — which this chapter tries to teach — is what keeps the practice on the right side of that line. We are not teaching you to manipulate readers. We are teaching you to lead them through evidence.

No code in this chapter. Part II remains library-agnostic. The matplotlib and dashboard-building code for building sequences — animated transitions, multi-panel layouts with narrative flow, scrollytelling-style embeds — waits for you in Part III and beyond. This chapter is about the design principles that will guide those implementation choices.

9.1 What Data Storytelling Is (and Isn't)

The Central Claim

Data storytelling is the practice of arranging charts, text, and transitions into a coherent narrative that takes a reader from context, through evidence, to conclusion. It is not just "adding text to charts." It is the structural decision about which charts, in what order, with what framing.

The practice has three elements:

1. The charts themselves, each meeting the standards of Chapters 6 through 8: decluttered, self-explanatory, composed deliberately.

2. The sequence of charts, arranged to walk the reader through a logical argument from beginning to end.

3. The framing around the charts — the titles, subtitles, annotations, captions, transitional text, and overall narrative structure — that ties the individual charts into a single story.

All three matter, and all three fail without the others. A good chart in a bad sequence is still part of a bad story. A good sequence of bad charts is still a bad story. A good sequence of good charts without framing is a presentation that the audience has to piece together themselves. The craft of data storytelling is the integration of all three.

What Storytelling Is Not

It is worth being specific about what data storytelling is not, because the term is sometimes used loosely and the looseness can mislead.

Storytelling is not manipulation. The goal of storytelling is to help the reader understand something true about the data, not to persuade them of something false. The line between storytelling and manipulation is the same line between honest and distorted charts from Chapter 4: whether the narrative matches the evidence or overstates it. A data story that matches the evidence is ethical. A data story that overstates the evidence is manipulation, regardless of how well it is structured.

Storytelling is not fiction. Data stories are about real data — actual numbers from actual measurements. The storyteller does not invent events, exaggerate differences, or fabricate context. The "story" in "data storytelling" refers to the narrative structure, not to fictional content. A good data story is a true story well-told.

Storytelling is not always needed. Some charts do not belong in a story. Reference charts, exploratory charts, dashboards that display live metrics — these are valuable in their own right and do not need to be embedded in a narrative. The chart maker should know when a chart is a story element and when it is a standalone artifact, and should not force everything into narrative form.

Storytelling is not about the storyteller. Good data storytelling puts the data and the audience at the center, not the storyteller. The storyteller's personal style, preferences, and creativity are in service of the audience's understanding, not ends in themselves. A story that draws attention to the storyteller's cleverness is a story that has lost sight of its purpose.

Storytelling is not "just adding text." Writing a longer caption or adding an action title is not, by itself, storytelling. Storytelling is structural: it affects which charts you include, which you exclude, what order they appear in, and how they relate to each other. Text is part of the structure but not the whole of it.

Two Separate Skills: Analysis and Storytelling

The chapter's first practical distinction is between data analysis and data storytelling. These are separate skills, and most practitioners are better at one than the other.

Data analysis is the process of finding the story in the data. You start with a dataset and a question, and through exploration — summary statistics, exploratory visualizations, statistical models — you learn what the data actually shows. The output of analysis is an understanding of the data: its structure, its patterns, its anomalies, its answers to your questions.

Data storytelling is the process of telling the story to an audience. You start with a finding (from analysis) and an audience, and through presentation — explanatory visualizations, narrative structure, transition text — you convey the understanding to people who did not do the analysis themselves. The output of storytelling is an audience that understands what the data shows.

The two skills use different tools and follow different conventions. Analysis values speed and flexibility — many quick exploratory charts, minimal polish, tolerance for messiness. Storytelling values clarity and intentionality — fewer but more polished charts, careful composition, deliberate sequencing. Applying analysis tools to storytelling (showing 47 exploratory charts to an executive audience) fails because the audience cannot follow the sequence. Applying storytelling tools to analysis (spending an hour polishing an exploratory chart) fails because the polish is wasted on a chart you will not show to anyone.

Most practitioners learn analysis first — it is taught in statistics courses, data science programs, and most introductory visualization materials. Storytelling is learned later, if at all. This chapter argues that storytelling is a distinct skill that deserves its own deliberate development, and that the skills of Part II (declutter, typography, composition) are the foundation on which storytelling is built.

Check Your Understanding — Think of a recent project where you analyzed data and then presented findings to an audience. Did you use the same charts in the presentation that you used during the analysis? If yes, how did the audience respond? If no, what changed between the analysis charts and the presentation charts?

9.2 The Narrative Arc Applied to Data

The Three-Act Structure

The standard narrative structure in Western storytelling — going back at least to Aristotle's Poetics — is a three-act form:

• Act 1: Setup. Establishes the context, the characters, the normal state of the world. The audience learns who and what the story is about.
• Act 2: Confrontation (or Rising Action). Introduces a complication, a conflict, a discovery that disturbs the normal state. The story builds tension as the characters confront the situation.
• Act 3: Resolution. The tension resolves. The audience learns what the complication means, what has changed, and what the implications are.

This structure is not arbitrary. It matches how human audiences actually process information: we need context before we can understand a complication, we need tension before we can care about a resolution, and we need resolution before we feel that a story has ended. A story missing any of the three acts feels incomplete.

The same structure applies to data stories:

• Act 1: Context. What is the normal state of the world? What should the reader know about the subject before we start showing evidence? This is where you introduce the data, the baseline, the background trends, the "before picture."
• Act 2: Evidence. What does the data show? This is where the main finding appears — the comparison, the trend, the anomaly, the pattern that is the reason the story exists. The finding is the confrontation; it disturbs or confirms the context set up in Act 1.
• Act 3: Implications. So what? What does the finding mean? What should the reader do about it? What comes next? This is where the story resolves — not by claiming the future is determined, but by telling the reader why the finding matters.

This three-act structure is not a rigid formula. It is a template that most data stories fit, and it gives you a default structure to work with when you are unsure how to arrange your charts. A story that does not fit the template may be better than one that does, but you should know what you are departing from and why.

A Worked Example: The Climate Story

Let us apply the three-act structure to the progressive climate project that this book has been building.

Act 1: Context (1-2 charts). The first chart shows the long history of global temperatures. We use a time series of temperature anomalies from 1880 to 2024. The story in this chart is "here is what temperatures have done historically." The reader sees about a century of relatively stable temperatures followed by a rise in recent decades. The chart is contextual: it does not yet make an argument, but it establishes the baseline and the scope of the story.

Act 2: Evidence (2-3 charts). The second chart shows CO2 concentration over the same time period, revealing that CO2 has risen sharply since the mid-20th century. The third chart, a scatter plot or dual time series, shows the direct relationship between CO2 and temperature — each increase in CO2 corresponds to an increase in temperature. The fourth chart might be a small multiple showing how the warming has varied regionally, establishing that the pattern is global but not uniform. These charts build the central argument: warming is happening, it is correlated with CO2, and the pattern is consistent across the globe.

Act 3: Implications (1-2 charts). The fifth chart projects the implications forward: what do models predict for the next 50 or 100 years under different emissions scenarios? Or: what do current emissions imply for near-term temperature targets (the 1.5°C or 2°C limits)? The final chart closes the story by showing the reader what the evidence means for the decisions that have to be made. It may include a call to action — policy implications, personal actions, the timescale for response.

Five charts total, arranged in a three-act structure. The opening chart establishes the baseline. The middle three charts build the evidence. The final chart delivers the implications. Each chart is decluttered, self-explanatory, and well-composed (Chapters 6 through 8). The sequence is the story.

Notice what happens if we reorder the charts. If we start with the projection chart (Act 3), the reader sees the implications before they understand the evidence, which makes the implications feel unfounded. If we start with the scatter plot of CO2 vs. temperature (Act 2), the reader sees a correlation without the context to know whether it matters. If we end with the historical time series (Act 1), the reader is left in the past instead of moving toward the future. The sequence matters, and the three-act template gives us a principled way to decide the order.

The Big Idea

Cole Nussbaumer Knaflic, in Storytelling with Data, emphasizes the concept of the Big Idea: the single sentence that captures the whole point of the data story. Before you start writing, sketching, or designing anything, you should be able to state the Big Idea in one sentence.

The Big Idea is harder to write than it sounds. For the climate story, a weak Big Idea might be: "Climate change is a problem." This is too vague — it does not tell the reader what specifically the story is showing. A stronger Big Idea: "Global temperatures have risen 1.2 degrees Celsius since pre-industrial times, tracking the rise in atmospheric CO2, and current emissions put us on a path to exceed the 1.5-degree threshold within 20 years." This is a full sentence with specific numbers, a causal claim, and an implication. It is what the story is actually saying.

Every data story should have a Big Idea that you can state in one sentence. If you cannot state it, the story does not have a coherent argument, and you should go back to the analysis stage and figure out what you are trying to say. The Big Idea becomes the guiding principle for every subsequent design decision: which charts to include (only those that support the Big Idea), which charts to exclude (those that do not), which charts to emphasize (the ones that carry the main claim), and which to demote (the ones that are supporting evidence).

The Big Idea is also the ancestor of the action title from Chapter 7. The action title on an individual chart is a small version of the Big Idea for that chart. The Big Idea for the whole story is a large version of the same principle applied at the story level. Writing a good Big Idea is a disciplined act of interpretation, and the discipline transfers directly to the action titles of the individual charts in the story.

Check Your Understanding — Think of a recent data presentation you made. Write the Big Idea for that presentation as a single declarative sentence. Was this sentence clear in the original presentation, or did the presentation leave the reader to infer the main point?

9.3 Audience Analysis: Knowing Who You Are Telling the Story To

The First Question

The first question of data storytelling is not "what does the data say?" It is "who is the audience?" The same data can be told as ten different stories for ten different audiences, and the choice of audience shapes every subsequent decision: which charts to include, what jargon to use, how much statistical detail to show, how much context to assume, what the call to action should be.

The worst mistake in data storytelling is the "one-size-fits-all" presentation: the same slide deck shown to executives, engineers, and the general public, hoping that each will extract the parts they need. This almost never works. Executives get bogged down in technical detail; engineers feel the context is too shallow; the general public does not understand the jargon. The "one-size" story is actually "no-size" — it serves no audience well.

The discipline of audience analysis is to start every data story by identifying the specific audience and letting the audience shape the design. Who will read this? What do they already know? What do they care about? What decision will they make based on the information? Once you can answer these questions, the design decisions become much clearer.

A Taxonomy of Audiences

Most data stories are told to one of four general audience types:

1. Technical audiences. Scientists, engineers, data analysts, researchers, academics. They understand statistical language, are comfortable with complex charts, and value rigor and precision. They will read detailed methodology notes, examine confidence intervals, and question assumptions. For this audience, you can use technical vocabulary, include more detail, and assume familiarity with visualization conventions. The storyteller's job is to present the evidence rigorously; the audience will do the interpretation.

2. Executive audiences. Business leaders, managers, decision-makers. They have limited time and need the bottom line fast. They care about implications for decisions they have to make. They will not read detailed methodology; they will read the headline finding and ask "what should I do?" For this audience, you need short, high-impact stories: one or two slides with the main finding, the supporting evidence in condensed form, and a clear recommendation. The storyteller's job is to help the audience make a decision, not to present all the evidence.

3. General audiences. Newspaper readers, podcast listeners, people scrolling through social media. They have even less time than executives and less technical background. They need context, plain language, and stories that connect the data to something they already care about. They will not parse statistical charts; they will absorb visual impressions and memorable facts. For this audience, you need accessible charts, clear narrative, and emotional resonance. The storyteller's job is to make the data meaningful to people who did not come looking for it.

4. Mixed audiences. A conference talk with engineers, managers, and journalists in the same room. A blog post that will be read by both experts and novices. These are the hardest audiences because you cannot fully optimize for any single subgroup. The strategy is usually progressive disclosure (covered in the next section): start with content everyone can understand, then layer in detail for readers who want more.

Practical Adjustments by Audience

Once you know your audience, adjust the following dimensions:

1. Vocabulary. Technical vocabulary is efficient for specialists but alienating for general audiences. "The posterior probability of the null hypothesis, given our data, is 0.03" is correct but inaccessible. "There's only a 3% chance this result is random" is accessible but loses some technical precision. Choose based on the audience.

2. Complexity of charts. Violin plots, box plots with notches, and multi-dimensional scatter plots are fine for technical audiences and baffling to general audiences. For general audiences, use simpler chart types (line, bar, scatter) and annotate them heavily. For executives, use charts that communicate a single clear finding quickly. For specialists, you can use more sophisticated chart types without explanation.

3. Amount of context. Technical audiences know the field; you do not need to explain what CO2 is or why climate matters. General audiences may need the background. Executives need just enough context to understand the decision at hand, not the full scientific story.

4. Statistical rigor. Error bars, confidence intervals, p-values, effect sizes — all of these should appear prominently for technical audiences. For general audiences, they should be translated into plain language or omitted entirely, depending on the context. For executives, include enough rigor to defend the finding but not so much that the bottom line gets lost.

5. The "so what" framing. Different audiences care about different implications. A technical audience cares about what the data means for scientific understanding. An executive audience cares about what the data means for business decisions. A general audience cares about what the data means for their lives or their community. Frame the implications to match the audience's concerns.

6. Length. Technical audiences will sit through a 45-minute presentation with 30 charts if the charts are informative. Executives will leave after 10 minutes. General audiences will scroll past in 30 seconds. Match the length to the audience's patience.

A Worked Example: Three Versions of the Same Story

Consider a public health story about a new vaccine that shows 85% efficacy in a clinical trial. Here are three versions for three different audiences:

Technical version (for medical researchers): A detailed presentation with dose-response curves, adverse event analyses, confidence intervals around the efficacy estimate, subgroup analyses by age and comorbidity, and comparisons to other vaccines in the same category. 20-30 slides. The audience will examine the charts carefully and ask detailed questions.

Executive version (for health system administrators): Three slides. Slide 1: "New Vaccine Shows 85% Efficacy" — action title, single chart showing the efficacy comparison to other vaccines. Slide 2: "Cost and Deployment Considerations" — a chart showing vaccine cost per dose, estimated deployment timeline, and staffing requirements. Slide 3: "Recommended Action" — a clear recommendation with the bullet points for the decision.

General audience version (for a news article): A single chart or a scrolling narrative with 4-5 charts, written in plain language. The charts show the efficacy finding, the comparison to past vaccines, the real-world impact (how many infections prevented), and the practical question of when the vaccine will be available. Jargon is minimized; the story connects the data to the reader's experience.

All three versions are based on the same underlying data. They differ in which charts are shown, how much detail is included, what vocabulary is used, and what framing is applied. Each is right for its audience; none would work well for the other two.

Check Your Understanding — Think of a data story you might want to tell. Identify three different audiences who might care about it, and describe how the story would differ for each one.

9.4 Progressive Disclosure: Shneiderman's Mantra

The Principle

Ben Shneiderman, a pioneer of information visualization, formulated a mantra that has become foundational to interactive data design:

"Overview first, zoom and filter, then details on demand."

The principle, called progressive disclosure, says that good data communication starts with a big-picture view that everyone can understand, then lets readers drill into specific aspects they care about, and only provides the most detailed information when they explicitly ask for it. The reader who wants only the headline gets the headline. The reader who wants more can get more. The reader who wants everything can get everything.

Progressive disclosure is native to interactive visualizations — dashboards, scrolling narratives, interactive charts with hover tooltips — where the reader can literally request more detail by clicking or scrolling. But the principle also applies to static data stories, with some adaptation:

Overview first. The opening chart or paragraph gives the reader the Big Idea. If the reader reads nothing else, they should still get the main finding. The opening is the 5-second version of the story.

Zoom and filter. The middle of the story shows specific evidence: individual charts, detailed comparisons, subgroup breakdowns. Different readers will focus on different parts based on their interests. An executive might scan the charts for the key numbers; a technical reader might study the methodology notes; a general reader might follow the narrative prose.

Details on demand. The technical details — the exact methodology, the full dataset, the supplementary charts, the assumptions and limitations — are available for readers who want them, but they are not on the main page. They are in an appendix, a footnote, a linked document, or a clickable expansion.

The structure respects the reader's attention budget. Most readers will only read the overview. Some will read the middle. Very few will read the details. The structure is designed so that all three levels of engagement get something valuable, rather than forcing everyone to read the same level of detail.

Progressive Disclosure in Different Media

The implementation of progressive disclosure depends on the medium:

Static documents (reports, articles, papers). The overview goes in the headline, the abstract, or the executive summary. The zoom-and-filter middle goes in the body of the document. The details on demand go in footnotes, appendices, and methodology sections. A well-structured report can be read at three different depths: headline only, main text only, or full text including appendices.

Slide decks. The overview is the opening slide with the main finding. The zoom-and-filter middle is the body of the deck with supporting evidence. The details on demand are in the backup slides — slides at the end of the deck that are not presented but are available if someone asks a question. A well-structured deck has a clear primary flow plus a reservoir of detail for Q&A.

Dashboards. The overview is the top-level metrics visible when the dashboard loads. The zoom-and-filter middle is the ability to click on a metric to see its trend or breakdown. The details on demand are the raw data tables, the methodology documentation, and the underlying filters. A well-designed dashboard starts simple and reveals complexity on demand.

Articles and scrolling narratives. The overview is the headline and the lead paragraph. The zoom-and-filter middle is the body of the article with embedded charts. The details on demand are linked to data sources, companion articles, and technical supplements. The New York Times pandemic coverage is a good example: the main article was accessible to general readers, but links led to increasingly detailed data pages for readers who wanted more.

Practical Techniques for Progressive Disclosure

1. The hero chart. The first chart in a story carries the Big Idea. It should be the most prominent, most self-explanatory, most polished chart in the whole document. If the reader looks at only one chart, this is the one. Subsequent charts provide the supporting evidence and the nuance.

2. The grayed-out strategy. In a sequence of charts showing progressively more detail, grayed-out elements show context while emphasized elements carry the focus. The first chart might gray out everything except the hero metric; the second chart might highlight a different subset; the third chart might show the full complexity with annotations. The grayed-out strategy lets you reveal detail progressively without starting over each time.

3. The layered annotation. A chart can carry different levels of detail through layered annotation. The first glance reveals the title, the action title, and the basic pattern. A longer look reveals the annotations. An even longer look reveals the footnote references. The reader who wants more detail can get it, but the glance-level information is also complete.

4. The footnote layer. For static documents, footnotes are the textual version of "details on demand." The main text reads cleanly without footnotes; readers who want to know the specific source, the methodology, or the caveat can drop into the footnote. Use footnotes as a safety valve for detail that would clutter the main text.

5. The "click to expand" idiom. In digital documents, the <details> HTML element or a JavaScript expand-on-click widget lets the reader choose to see more. The default view is clean; the detail is available to readers who want it. This is a direct implementation of progressive disclosure in the interactive context.

When Progressive Disclosure Fails

Progressive disclosure is not always right. Some failure modes:

1. Hiding essential information. If the "detail" is something the reader actually needs to understand the main finding, hiding it in a footnote or behind a click is a mistake. The question is whether the detail is essential context or optional enrichment. Essential context belongs in the main view; optional enrichment belongs in the progressive layer.

2. Over-disclosure. Some stories do not need layered disclosure because the audience wants the full picture at once. A technical audience reading a research paper is usually better served by a comprehensive chart with all the details visible than by a simplified chart with details hidden.

3. Bad defaults. If the "overview" version of a chart gives a misleading impression that is only corrected by the "detail" version, the progressive disclosure is doing harm. The overview must be accurate on its own terms; the detail should refine the overview, not correct it.

4. Disclosure theater. Adding interactivity for its own sake, when the reader has no reason to click or scroll, wastes effort and can confuse readers who do not know the expand-on-click idiom. Progressive disclosure works only when the layers of detail are genuinely valuable.

Check Your Understanding — Think of a complex data document you have read recently (a long report, a dashboard, a detailed article). Did it use progressive disclosure effectively? What was in the overview, what was in the middle, and what was in the details-on-demand layer?

9.5 Visual Emphasis: Guiding the Reader's Eye

The Problem

A chart that shows everything at once asks the reader to figure out what matters. A chart that visually emphasizes the important elements tells the reader where to look. For most explanatory charts, the second approach is much more effective, because it reduces the reader's work and makes the main finding unmissable.

Visual emphasis is the set of techniques for making some elements of a chart more prominent than others. The goal is to create a visual hierarchy that matches the semantic hierarchy of the story: the most important elements should be the most visually prominent, and the supporting context should be visible but less prominent.

This chapter has already touched on several emphasis techniques in earlier sections and chapters. The action title from Chapter 7 is emphasis applied to text. The hero panel from Chapter 8 is emphasis applied to composition. Now we bring these techniques together into a systematic practice.

Techniques for Visual Emphasis

1. Color emphasis. The most common technique: the important element is drawn in a bright accent color, and everything else is drawn in muted gray. A chart of 50 state trajectories might show one state (the focus of the story) in red and the other 49 in pale gray. The reader's eye goes immediately to the red line, and the gray lines provide context without competing for attention. Color emphasis is powerful and widely applicable; the key is to use it sparingly (one or two highlighted elements per chart) so that the emphasis remains effective.

2. Size emphasis. Important elements are drawn larger or thicker than supporting elements. A trend line that represents the main finding might be drawn with a thicker stroke than a reference line or a secondary trend. The larger element pops out pre-attentively. Size emphasis combines well with color emphasis — a thick red line against a background of thin gray lines is a powerful hierarchy signal.

3. Annotation emphasis. A callout, an arrow, or a text label placed on the important element draws the reader's attention explicitly. Annotations say "look here" in the most direct way possible. One annotation on the key data point, paired with the rest of the chart in quiet context, is often more effective than a chart with multiple annotations competing for attention.

4. Spatial emphasis. The most important element is given more visual space, more whitespace around it, or a more prominent position in the layout. The hero panel in a multi-panel figure gets more space than the supporting panels. A key data point is separated from its neighbors with additional whitespace. Spatial emphasis creates hierarchy without adding any ink to the chart.

5. Contrast emphasis. The important element has higher contrast against the background than the supporting elements. A dark line against a light background stands out; the same dark line against a dark background disappears. The luminance contrast principle from Chapter 3 applies here: the important element should have the highest contrast against the background.

The Grayed-Out Strategy

The most powerful single technique for visual emphasis is the grayed-out strategy: draw the context in muted gray, and draw the focus in a bright color. This technique works because the reader's eye automatically prioritizes high-contrast elements over low-contrast ones. The gray context provides the full picture, but the reader's attention is directed to the colored focus.

The grayed-out strategy is the default approach at most data journalism outlets because it scales from single charts to complex figures. In a single chart, gray out the background data and color the focus series. In a small multiple, gray out the non-focus panels and color the focus panel. In a full dashboard, gray out the routine metrics and color the ones that need attention. The same principle applies at every scale.

The grayed-out strategy is also ethically clean in most cases. The context is still visible; the reader can still see the full data. The emphasis directs attention without hiding information. This is different from cherry-picking, which removes inconvenient data entirely. With the grayed-out strategy, the reader who wants to see the full context can look past the emphasized element; with cherry-picking, there is nothing to look at.

Emphasis Across a Sequence

In a multi-chart story, visual emphasis works across the sequence, not just within each chart. Each chart in the sequence can emphasize a different aspect:

• Chart 1 emphasizes the context. The full dataset is shown, with no particular element highlighted. The reader absorbs the baseline.
• Chart 2 emphasizes the main finding. The context remains visible in gray, and the finding is colored prominently. The reader sees the claim against the background of the context.
• Chart 3 emphasizes a different aspect (a subgroup, a comparison, an extreme point). The context is still there, but the focus has shifted.
• Chart 4 emphasizes the implication. The same data is shown with the key takeaway annotated directly.

Each chart uses the same base data but directs the reader's attention to a different part of the story. This is how visual emphasis supports narrative: the sequence of emphasized elements walks the reader through the argument, revealing one step at a time.

The "Big Idea Highlight" Technique

A specific implementation of visual emphasis that deserves explicit attention: take the single chart that carries the Big Idea, and make it the most visually prominent element in the entire story. Larger than the other charts. More color. More annotation. More whitespace. This chart is the single most important thing the reader should remember, and its visual prominence should match its importance.

In a report, the Big Idea chart might be a full-page figure while the others are half-page. In a slide deck, it might get a dedicated slide with a large action title while the other slides are less prominent. On a dashboard, it might be the largest tile in the layout. In an article, it might be the opening figure, the one the reader sees first.

The technique is simple but underused. Most reports treat every chart equally, and the reader cannot tell which one carries the main point. The Big Idea highlight forces a hierarchy onto the document: this chart is the one the reader should remember, and everything else is supporting evidence. The hierarchy makes the document easier to read and the main message easier to retain.

Check Your Understanding — Look at a slide deck or report you made recently. Can you identify, within three seconds, which chart carries the Big Idea? If not, how would you apply visual emphasis to make that chart more prominent?

9.6 Storyboarding: Planning the Sequence Before Designing the Charts

Why Storyboarding

The mistake most data storytellers make is to design the charts first and then try to arrange them into a story. This order is backwards. The charts should serve the story, not the other way around. You should plan the story first — what the Big Idea is, what evidence supports it, what sequence of charts will walk the reader through it — and only then design the specific charts to match the plan.

This is what storyboarding is for. Before you open matplotlib or your plotting tool of choice, sketch the story on paper. One chart per sheet, or one frame per whiteboard square. Write the Big Idea at the top. Sketch the opening chart (the context). Sketch the main chart (the finding). Sketch the closing chart (the implications). Look at the sketches side by side. Does the sequence make sense? Does one chart lead naturally to the next? Is the argument clear?

Storyboarding is low-fidelity by design. You are not trying to produce a beautiful sketch; you are trying to make the structural decisions before you spend hours on chart polish. Sticky notes, a whiteboard, a stack of index cards — any medium that lets you rearrange easily works. The point is to find the right sequence before committing to any specific chart design.

The Sticky Note Method

Cole Knaflic popularized a specific storyboarding technique using sticky notes. The method:

1. Write the Big Idea at the top of a wall or a large piece of paper.
2. For each chart you think belongs in the story, write the chart's action title on a sticky note and place it on the wall.
3. Rearrange the sticky notes until the sequence feels right — setup, then evidence, then implications.
4. Look for gaps (where is the sequence unclear?) and redundancies (where are two charts making the same point?).
5. Remove redundant sticky notes, add new ones to fill gaps, rearrange until the sequence tells the complete story.
6. Only then do you sit down to design the actual charts.

The method is deliberately analog because the physical act of moving sticky notes forces you to commit to decisions. On a screen, it is too easy to have "draft" arrangements that never get resolved. On a wall, the sticky notes are either there or they are not, and the current arrangement is whatever you see. The constraint of physical space turns the decision into an action.

The sticky note method also surfaces problems that screen-based planning hides. If you have 30 sticky notes on the wall and they will not fit into a coherent sequence, you have too many charts or an unclear argument. If you have 3 sticky notes and the sequence feels thin, you have not yet thought through the evidence. The physical sticky notes make the problem visible in a way that a mental plan does not.

Storyboard Levels

A storyboard can operate at multiple levels, from very rough to quite detailed:

Level 1: The Big Idea and the rough sequence. Just the Big Idea and a list of chart topics in order. No sketches yet, no specific chart types. This is the 10-minute version that answers "does this story work?"

Level 2: Rough sketches of each chart. Pencil sketches of the individual charts, showing the general chart type, the axes, and the rough shape of the data. No specific numbers, no polish. This is the 30-minute version that answers "is the sequence of visual forms coherent?"

Level 3: Detailed sketches with action titles. Each chart sketched with its specific chart type, axis labels, and action title. This is the 1-2 hour version that answers "do the individual charts work and do they connect to each other?"

Level 4: Digital prototypes. First-draft versions of the charts in your plotting tool, using real data but with rough styling. This is the 4-6 hour version that turns the storyboard into actual visualizations. At this point, the sequence should already be decided; you are implementing it, not designing it.

Different projects require different levels. A short executive presentation might stop at Level 1 or 2. A major report might require Level 3 before any digital work begins. A newspaper article might go from Level 2 directly to the final charts. Choose the level that matches the stakes and the complexity.

The Revision Discipline

A good storyboard is revised many times. You sketch the sequence, look at it, find a problem, rearrange, look again. The revision cycle is the point — you are iterating on the narrative structure at a low cost, before committing to expensive chart-design work.

Common revisions:

• Cutting charts that do not support the Big Idea. Every chart should advance the argument; if it does not, remove it. Most storyboards end up with fewer charts than they started with, and the cuts usually improve the story.
• Adding charts that the Big Idea requires. Sometimes the sequence is missing a piece — a context chart that the reader needs, a comparison chart that supports the finding, an implications chart that closes the story. Add it.
• Reordering to match the reader's natural flow. The order that feels right to you as the analyst is often not the order that works for the reader. Test the order by explaining the story to a colleague and watching where they get confused.
• Consolidating related charts. Sometimes two charts that were separate can be combined into a small multiple or a panel. Sometimes one chart that was trying to do too much should be split into two.

Revision is not failure. Revision is the normal working process of storytelling. A first draft of a storyboard is rarely the right sequence, and the revision process is where the sequence gets good.

9.7 Storytelling and Ethics: The Line Between Narrative and Manipulation

The Tension

Every data story has a point of view. The storyteller chose which charts to include, what order to present them in, which audience to write for, which annotations to emphasize, and which implications to highlight. These choices are editorial — they shape the reader's understanding in specific ways. In this sense, data storytelling is unavoidably a form of persuasion, not just a neutral display of facts.

The ethical question is whether the persuasion is honest. This is the same question Chapter 4 raised about individual charts, extended to the full sequence of a story. A story that walks the reader through a true finding in a clear and accurate way is ethical persuasion. A story that walks the reader through a distorted finding, or that hides inconvenient evidence, or that overstates the implications, is manipulation.

The line between these two is not always clear, and the chapter has to be honest that the temptation to slide from storytelling to manipulation is real. A well-told story is memorable, persuasive, and influential — and if the story is not quite true, it is persuasive in the wrong direction. The chart maker has to develop the discipline of telling true stories in compelling ways, which is harder than telling compelling stories that are loose with the truth.

The Cherry-Picking Temptation

The most common failure mode in data storytelling is cherry-picking: selecting only the evidence that supports the story, and omitting the evidence that complicates it. A story about "our new product line is succeeding" might show charts of the three best-performing products while omitting the two that are struggling. The story is not technically wrong — the three products really are succeeding — but it leaves the reader with a false impression of the overall line.

Cherry-picking is tempting because it produces cleaner stories. A narrative with complications is harder to tell than one with a clear arc. The discipline is to tell the complicated story anyway, because the audience deserves to see the full picture. A good storyteller can handle complications: "Three of our five products are performing well; two are struggling for specific reasons that we are addressing." This is a longer sentence than "Our products are performing well," but it is honest, and the honesty builds trust that the shorter sentence does not.

The chart-level rule from Chapter 4 — show the full dataset when the pattern is the point — extends to the story level. If your story is about a trend, show the data that contradicts the trend as well as the data that supports it. If your story is about a comparison, show the comparisons that work against your finding as well as the ones that work for it. The transparency does not usually weaken the story; it usually strengthens it by preempting the reader's skepticism.

The Overstatement Temptation

The second common failure mode is overstatement: pushing the implications of the data beyond what the data actually supports. A chart showing a 5% improvement in customer satisfaction becomes a story about "dramatic improvement in customer loyalty." A chart showing a correlation becomes a story about a causal relationship. A chart showing a recent trend becomes a story about a long-term transformation.

Overstatement is easy to do unintentionally, because storytelling rewards vivid language and strong claims. "Customer satisfaction improved slightly" is a weaker headline than "Customer satisfaction surged." The storyteller reaches for the stronger phrasing without realizing that the stronger phrasing makes a claim the data does not support. Over time, this drift produces stories that systematically overstate findings.

The discipline is to match the language to the evidence. If the data supports "improved slightly," the story says "improved slightly." If the data supports "surged," the story says "surged." The discipline requires the storyteller to be honest with themselves about what the data actually shows, which is harder than it sounds. A good practice is to write the draft story, then go back through every claim and ask: "Does the chart really support this phrasing, or am I reaching?"

The Framing Temptation

The third common failure mode is framing manipulation: choosing a frame that makes the finding seem more impressive or more alarming than a different, equally valid frame would. A 5% decrease in emissions can be framed as "we're making progress" or "at this rate, we won't meet our targets for 50 years." Both are true; they imply very different judgments.

The ethical response is to choose the frame that best serves the reader's understanding, not the frame that best serves the storyteller's preferred conclusion. This sometimes means resisting the most persuasive framing in favor of a more balanced one. It sometimes means presenting multiple frames and letting the reader decide. It always means being aware of the framing choice and willing to defend it.

The Storytelling Ethic

A practical ethical rule for data storytelling: could you defend every claim in your story to a skeptical reader who has access to the same data? If yes, the story is probably honest. If no, something in the story is overstated, cherry-picked, or framed in a way you cannot defend — and you should fix it before publishing.

This is the adversary test from Chapter 4, applied to the full story rather than a single chart. Imagine the most skeptical reader in your audience — the one who disagrees with your conclusion, who will scrutinize every chart, who will challenge every claim. Would your story survive their scrutiny? If not, what would you change to make it defensible?

The adversary test does not mean you cannot tell strong stories. It means the strong stories you tell should be ones you can defend. A story that cannot survive scrutiny is either too weak to tell or too overstated to tell honestly. The middle ground — strong stories that are honest — is the goal.

9.8 The Climate Story: Bringing Everything Together

By the end of Part II, the progressive climate project has become a complete data story. The evolution across chapters:

• Chapter 1: A single ugly default chart of temperature anomalies.
• Chapter 6: The same chart, decluttered.
• Chapter 7: The decluttered chart with an action title, subtitle, annotations, and source attribution.
• Chapter 8: A three-panel small multiple showing temperature, CO2, and sea level on a shared time axis.
• Chapter 9: A five-chart narrative sequence with a clear beginning, middle, and end.

The full narrative structure of the climate story:

Big Idea: "Global temperatures have risen 1.2 degrees Celsius since pre-industrial times, tracking the rise in atmospheric CO2, and current emissions put us on a path to exceed the 1.5-degree target within roughly 20 years."

Act 1: Context (1 chart).

Chart 1 — "Global Temperatures, 1880–2024." A single time series of temperature anomalies from 1880 to 2024, with a horizontal line at 0 (the 1951–1980 baseline). The chart establishes the scope: 144 years of data, a long period of relative stability, and a clear rise in recent decades. The reader sees that the current warming is real and significant relative to the historical baseline.

Act 2: Evidence (3 charts).

Chart 2 — "CO2 Concentration Has Risen Sharply Since 1950." A parallel time series of atmospheric CO2 concentration from 1880 to 2024, with the pre-industrial baseline of ~280 ppm marked as a reference line. The chart shows CO2 rising from about 280 ppm to about 420 ppm, with most of the rise happening in the second half of the 20th century.

Chart 3 — "CO2 and Temperature Rise Together." A direct comparison — either a dual time series (with appropriate disclaimers) or a scatter plot with year as the color — showing the temporal correlation between CO2 and temperature. The causal story is implied: CO2 and temperature move together.

Chart 4 — "The Warming Is Global." A small multiple or choropleth showing the spatial distribution of warming across major regions. Every region is warming, though at different rates. The pattern is not localized.

Act 3: Implications (1 chart).

Chart 5 — "At Current Emissions, We Exceed 1.5°C by 2045." A forward-looking projection showing the temperature trajectory under current emissions scenarios vs. a scenario consistent with the 1.5°C target. The gap between the two trajectories is the call to action — the difference between the current path and the target path.

The sequence is the argument. The reader starts with context (temperatures have been stable until recently), moves through evidence (CO2 is rising, CO2 and temperature are correlated, the warming is global), and ends with implications (current emissions will exceed the target within a human lifetime). Each chart is designed to the standards of Chapters 6–8 — decluttered, with action titles, properly composed. The sequence is designed to the standards of this chapter — three-act structure, visual emphasis, progressive disclosure from overview to detail.

A reader who has never seen any of these charts can walk through the five-chart sequence in five to ten minutes and come away understanding the climate story at a level that neither a single chart nor a data table could convey. This is what data storytelling does at its best: it transforms evidence into understanding, and it does so without distorting the evidence or manipulating the reader. The story is true, it is compelling, and it is honest.

Chapter Summary

This chapter argued that data storytelling — the deliberate arrangement of charts, text, and transitions into a coherent narrative — is a distinct skill from data analysis and chart design. The skills of Chapters 6 through 8 (declutter, typography, composition) produce individual charts and multi-chart figures. Storytelling produces sequences of charts that walk the reader through a complete argument.

The central structural device is the three-act narrative arc: setup (context), confrontation (evidence), and resolution (implications). Most data stories fit this template, and it gives you a default structure when you are unsure how to order your charts. The Big Idea — the single sentence that captures the whole point of the story — should be articulated before you design any specific charts and should guide every subsequent decision.

Audience analysis is the first discipline. The same data produces different stories for different audiences: technical, executive, general, or mixed. The chart maker adjusts vocabulary, complexity, context, rigor, and length based on who is reading. One-size-fits-all presentations usually serve no audience well; audience-specific stories serve their specific audiences effectively.

Progressive disclosure (Shneiderman's mantra: "Overview first, zoom and filter, then details on demand") applies to both interactive and static data stories. The overview carries the Big Idea; the middle provides specific evidence; the details are available to readers who want them. Good stories respect the reader's attention budget by layering detail rather than forcing everyone to read the same level.

Visual emphasis — through color, size, annotation, spatial treatment, and the grayed-out strategy — directs the reader's eye to the important elements without hiding the context. The grayed-out strategy is the most powerful single technique and scales from individual charts to full dashboards.

Storyboarding is the practice of planning the narrative sequence before designing the charts. The sticky note method — write each chart's action title on a sticky note and rearrange physically until the sequence works — is a useful low-fidelity way to iterate on the structure before committing to expensive chart design work.

The ethical dimension is central. Data storytelling amplifies the responsibility of the chart maker, because a well-told story is more persuasive than a single chart. The three temptations — cherry-picking, overstatement, and framing manipulation — are real, and the discipline of honest storytelling requires constant vigilance. The practical rule: could you defend every claim in your story to a skeptical reader who has access to the same data? If not, fix it.

The threshold concept is that the order in which you present charts is itself an argument. Reordering the same charts changes the story. There is no neutral sequence, and there is no way to tell a data story without making sequencing choices. The chart maker who pretends the sequence is arbitrary is hiding an editorial decision. The chart maker who embraces the sequence as deliberate is doing the work of narrative.

This chapter is the capstone of Part II. The craft of data visualization, as we have been building it across eight chapters, culminates in the ability to tell a coherent, honest, compelling story with data. The skills of Part II are not just for making individual charts — they are for walking an audience through a complete argument, one chart at a time, from context to conclusion.

Next, in Part III, we move from the craft of design to the tools of implementation. Chapter 10 begins the matplotlib sequence — the library you will use to build every chart you have been imagining across the first nine chapters. Part III answers the question "how do I actually make these charts in Python?" The answers are specific, technical, and detailed. But the principles of Part I and Part II — the perceptual foundations, the ethical discipline, the design craft, the narrative structure — will guide every matplotlib call you make from this point forward.

Spaced Review: Concepts from Chapters 1-8

These questions reinforce ideas from earlier chapters. If any feel unfamiliar, revisit the relevant chapter before proceeding.

1. Chapter 1: The "visualization as argument" framework says every explanatory chart makes a claim. How does storytelling extend this framework? Is a data story a single argument or a sequence of arguments?

2. Chapter 4: Chapter 4 identified the four zones of honesty: transparent, careless, negligent, and deliberate. How do these zones apply to data stories? Can a story be transparent in each individual chart but still be dishonest as a whole?

3. Chapter 5: The chart selection matrix from Chapter 5 tells you which chart type to use for which question. How does the matrix interact with storytelling? Does a story with five charts need five different chart types, or can they all be the same type?

4. Chapter 6: The declutter discipline produces clean individual charts. Does storytelling require even more decluttering than a standalone chart, or less? (Hint: in a sequence, the reader is looking at each chart briefly before moving on.)

5. Chapter 7: Action titles state the finding for individual charts. How do action titles across a sequence of charts fit together into a story? Should the action titles be different for every chart, or should they reinforce a single Big Idea?

6. Chapter 8: Small multiples enforce comparison through consistent encoding. In a storytelling context, where does a small multiple fit in the three-act structure? Is a small multiple more often a context chart, an evidence chart, or an implications chart?

7. Chapter 8 (continued): The chapter on layout argued that reading order follows the Z-pattern. How does this interact with the sequential reading of a narrative document? Is the Z-pattern relevant within each chart, across a multi-chart figure, or across the full sequence of a story?