4.2 Effective Storytelling Visuals

In a professional setting, the stories of interest will be data stories (with the obvious exceptions, of course). In this section, we discuss how visuals mesh with stories.

4.2.1 Stories and Illustrations

In the seminal Making Comics [35], graphic novelist Scott McCloud makes an interesting point regarding the flexibility of use of visuals in storytelling:

He is specifically talking about comics, but we believe it also applies to more general storytelling formats: any principle is at best a guideline, and it is not too difficult to imagine situations where they can be violated in the interest of getting the message across (see previous section).

That being said, it does help to have a few ideas about how one may incorporate visuals into data presentations: we will discuss, in particular, the combination of words and images to create impactful data stories, and the various visual storytelling choices that can improve the audience’s comprehension of the story.

Words and Images

It is said that a picture is worth a thousand words. How so? “Words bring an unparalleled level of specificity to all kinds of situations: there is no image so vague that words cannot lock it into a desired meaning.

McCloud presents the following example [35]:

What are we seeing here? A coffee cup, certainly. Thrown to the ground by an apathetic city dweller who could not be bothered to place it for a refuse bin? Or perhaps the coffee was too hot and scalded a customer, who dropped it to the ground in agony? Or any of a variety of other scenarios? Notice how the caption below collapses all interpretation to a single one, leaving no doubt as to what it is that the image is conveying.

On the bright side, I got my caffeine. On the not-so-bright side, we got mugged on the way home. [@McCloud_MC]

Figure 3.45: On the bright side, I got my caffeine. On the not-so-bright side, we got mugged on the way home. [35]

Additionally, some specific concepts and names can only be clearly expressed through words. Who could ever correctly guess what is going on in the following image without the caption?

Look! It's Kelly Donovan, twin brother of the guy who played Xander on _Buffy the Vampire Slayer_, plus Humprey Bogart wearing a Freddy Mercury mask, and a robot duplicate of former U.N. Secretary-General Boutros Boutros-Ghali! [@McCloud_MC]

Figure 3.46: Look! It’s Kelly Donovan, twin brother of the guy who played Xander on Buffy the Vampire Slayer, plus Humprey Bogart wearing a Freddy Mercury mask, and a robot duplicate of former U.N. Secretary-General Boutros Boutros-Ghali! [35]

Data story audiences are not always privy to the ins-and-outs of the analytical process that lead to a chart; what may appear obvious to the analysts may be opaque to stakeholders. It is preferable to err on the side of caution and use text wisely, leaving no room for ambiguity.

Visual Storytelling Choices

In The Mechanics of Visual Perception, we introduced the Gestalt principles, whose main objective was to reduce the cognitive load associated with reading (or parsing) charts.

The visual storytelling choices we discuss here are mostly of a different nature: they are concerned with clarity in communication (our treatment is based on an idea from [35]).

Communicating with clarity means that audience comprehension remains the ultimate goal:

  • the choice of moment helps “connects the dots”, which is to say that we show only what matters to the story, as in:

    Population distribution in the Americas - 2020. [@DV_Gapminder]

    Figure 4.20: Population distribution in the Americas - 2020. [87]

  • the choice of frame helps create and direct the audience’s focus, as in:

    Monthly attrition rate - May 2019. [@DV_NK]

    Figure 3.48: Monthly attrition rate - May 2019. [3]

  • the choice of image helps select the right charts for the story, with emphasis on simplicity and an ability to convey the message, as in these the two charts below built from the same data:

    Washington State University; percentage of staff and students by ethnicity, 2004-2013. [author unknown]

    Figure 4.21: Washington State University; percentage of staff and students by ethnicity, 2004-2013. [author unknown]

  • the choice of word (text) helps communicate ideas clearly and persuasively, in seamless combination with the charts, as in:

    Monthly attrition rate - 2019. [@DV_NK]

    Figure 4.22: Monthly attrition rate - 2019. [3]

  • the choice of flow helps guide the audience from one chart to the next, from one page to the next, and creates a transparent and intuitive “reading” experience by arranging the pages in a dashboard, the charts on a page, and the elements within charts intelligently.

Decisions having to do with moments, frames, and flow are likely to be made in the dashboard planning stages, while images and words decisions are usually being made right up to the finish line.58 We can:

  • start with a rough sketch of the dashboard (moment, frame, flow), then come up with the narrative (word), and finally populate the dashboard with charts (image);

  • start with a full ‘script’/storyboard (moment, word), then use that to do a rough layout of the dashboard (frame, flow), then populate the dashboard with charts (image), or

  • create a finished chart (moment, frame, image, word) with no idea as to what else will show up on the dashboard until you create another chart (flow), and so on, and so on.

Needless to say, we do not recommend the latter approach; nonetheless, it is probably the most commonly-used; it leads to analysts and stakeholders trying to shoehorn a story to the available charts, the exact opposite of what storytelling with data should be. We understand why that happens in practice: analysts are often overworked and burdened with unrealistic timelines and expectations. Nevertheless, this is not a sustainable strategy and the harm caused by this “fly-by-night” approach has long lasting consequences for organizations that rely on these stories for strategic decision-making.

Visual storytelling, then, sinks or swims based on how text and charts are integrated into the final product:

  • in text-specific combinations, the text provides all that is needed to know while the charts illustrate some aspects of the story that is described, as in:

    Extract from 'COVID-19: The Story of the Impossible Train — Illustrated' [@ImpossibleTrain]

    Figure 4.23: Extract from ‘COVID-19: The Story of the Impossible Train — Illustrated’ [88]

  • in chart-specific combinations, the charts provide all that is needed to know while the text accentuates some aspects of the story that is shown, as in:

    COVID-19 vaccination rates. [@Flourish]

    Figure 4.24: COVID-19 vaccination rates. [49]

  • in duo-specific combinations, both text and charts telling roughly the same story, as in:

    A history of the atom: theories and models. How have our ideas about atoms changed over the years? This graphic looks at atomic models and how they developed. [Compound Interest 2016](www.compoundchem.com)

    Figure 4.25: A history of the atom: theories and models. How have our ideas about atoms changed over the years? This graphic looks at atomic models and how they developed. Compound Interest 2016

  • in intersecting combinations, text and charts work together in some respects but also contribute to the story independently;

  • in interdependent combinations, text and charts combine to convey an aspect of the story that neither could convey alone, and

  • in parallel combinations, text and charts follow seemingly different storylines, without intersecting.

The last three combinations are not used as commonly when it comes to data storytelling;59

In the best data stories, text and charts are like partners in a dance and each one takes turns leading. When both partners try to lead, the competition can subvert the overall goals (e.g., clarity); when each partner knows their roles and supports the other’s strengths (which may differ from one page to the next), data presentation dashboards become fantastic storytelling media. based on an idea from [89]

4.2.2 Data and Stories

There are very few constraints associated with storytelling, in general: all story ideas, stylistic choices, or delivery modes are potentially in play, assuming that the choices are appropriate to the story function (education, entertainment, etc.).

Even when executed flawlessly, such stories are not necessarily good or compelling; they still have to be shared at the right time and the right place to find an appreciative audience, which could prove to be short-lived and/or quite small.60

But as long as a tale is recognized as an honest attempt at a story (using whatever metric the appraiser thinks is appropriate to do so), human being agree that they are dealing with a story.

Not so for data stories: we may only tell stories that are supported by the data. No flight of fancy, no faking data for the sake of the cause, no ignoring contradictory observations because it “makes for a better story”.

As we have discussed above (see Figure 4.8), there are different ways to be “supported by the data”. Data analysts have agency; they select:

  • the question(s) to answer;

  • what data to collect;

  • how to clean that data;

  • which analytical method(s) to use;

  • on what part(s) of the data to focus, etc.

These all have an effect on the stories that can be told with data, relative to the stories that could be told about the situations and events represented by the data.

Data analysts would do well to remember that they are human, that they come to the process with a whole slew of pre-conceived notions and cultural baggage. This does not preclude them having preferences about what the results will show or what the story should be about, of course, but this is another reason why having a tsarina of common sense around as a safety check is a good idea. Here is a nifty story about data stories illustrating the impact data analysis agency can have:

Why Numbers Matter | Episode 2 | [Do Maps Lie?](https://www.youtube.com/watch?v=G0_MBrJnRq0) | by _the Sheffield Methods Institute_

Figure 4.26: Why Numbers Matter | Episode 2 | Do Maps Lie? | by the Sheffield Methods Institute

Scoping vs. Exploration vs. Explanation vs. Persuasion

When working with data, we create visualizations at multiple stages in the process; the degree of polish we expect from such charts, as well as their number, depend on the specific stage.

This is not without reminding one of the general process underlying investigative journalism:

  1. initially, we scope out the area of investigation (objectives, data collection, story); at this stage, data charts usually show up in the form of conceptual doodles and proofs-of-concepts using the most basic of visualization software capabilities (Excel, base R, matplotlib, etc.);

  2. then we explore the situation’s phase space and the data we have collected about it; many more charts are created than will be used in any final deliverable (perhaps at a ratio of 10-to-1?), some of them entirely at random (to provide a baseline and some protection against pre-conceived notions and implicit biases); little effort is made on the aesthetic side of things as chart quantity and diversity are more valuable than chart quality at this stage; automatic and batch visualization capabilities is prefered (base R, matplotlib, etc.);

  3. we may use the outcome of this exploration to to answer the original analytical questions/objectives and explain the situation to our satisfaction; perhaps 1 in 5 exploration charts will be selected for beautification and decluterring via the Gestalt principles guidelines, based on relevance and on the planned data story’s outline and priorities; the tsarina of common sense comes into action at this stage, both in the selection process and as a critique of aesthetic choices (ggplot2, seaborn, etc.);

  4. finally, the most appropriate charts will be incorporated into a data presentation dashboard and woven into a data story used to persuade the audience and stakeholders into a suggested course of action to be taken with respect to the situation; the layout and construction of the final product is often left in the hands of graphic designers who use specialized software (Power BI, Tableau, Photoshop, video editing software, etc.) to create vivid stories.


In [90], Karl Popper differentiated science and pseudo-science by introducing the notion that scientific theories had to be falsifiable: this does not mean that theories had to be false, but that there should be some way to determine if they are false, namely that they should make predictions not just about what would be observed if they were true, but also if they were false.

For instance, “every person on Earth dies at most 500 years after they were born” is a falsifiable statement: all that is needed to contradict the statement is to exhibit a human being that has been alive for more than 500 years ago; “everybody dies at some point”, although almost certainly true, is not a falsifiable statement: any human being exhibited as contradictory evidence might simply not have died yet.

This approach is prescriptive, suggesting that scientists should eschew inductive processes (observation > pattern > hypothesis > theory) in favour of deductive processes (theory > hypothesis > observations > confirmation);

“no matter how many observations are made which confirm a theory there is always the possibility that a future observation could refute it. Induction cannot yield certainty. […] Science progresses when a theory is shown to be wrong and a new theory is introduced which better explains the phenomena. Scientists should attempt to disprove their theory rather than attempt to continually prove it.” [91]

In this view of the discipline, the scientific method is a set of tools that help us get progressively closer to “the truth” (assuming “perfect” experiments and measurements), but which we can never be certain has been reached.

This approach has been criticized by philosophers:

  • the Duhem-Quine thesis argues that since a theory is usually a complex collection of statements, the falsification of a single statement not of primary significance is not sufficient to reject the full theory [92];61

  • Thomas Kuhn argues that far from progressing gradually in small increments, science knows long periods of status quo, shaken rather frequently by seismic paradigm shifts [93].62

Be that as it may, we transport the Popperina approach to data storytelling: it should, in priciple, be possible for analysts and storytellers to imagine some type of data that could falsify the story they are telling. If this cannot be done, then the story and the data are not really connected, and can be told without one another.63

4.2.3 Evolving a Storytelling Chart

While great graphics are not sufficient for great data stories, they are at the very least a requirement. Following the work of [3], we refer to the process that takes raw numbers and analytical results and progressively transforms them into charts worthy of data stories as “evolving a storytelling chart”:

  1. numbers and tables are visualized into …

  2. ugly graphs, which are then decluttered into …

  3. simple graphs, which are further enhanced into …

  4. appealing graphs, which are finally integrated into …

  5. compelling data stories.

We base the following illustrative example on [3]: a charity organization records the number of meals they serve to the needy. From 2010 to 2020, the counts are as in the table below:

Most people will (rightfully) fail to identify anything from these numbers: they are, after all, just numbers. A simple (but franky, downright unappealing) chart shows that there is structure in the table:

We see that the number of meals served increased roughly linearly from 2010 to 2016, and that there was significant drops between 2016 and 2017, and 2019 and 2020, for instance: visualizing the data can provide some ideas for any eventual data story.

The chart is decluttered by:

  • removing the bounding box,

  • increasing the width of the bars,

  • reducing the number of colours,

  • renaming the axes, and

  • eliminating the number labels on the bars.

This leads to a simple and clean chart:

Let us assume that we have chosen to focus on the 2020 drop in number of meals served as the primary data story: we use the Gestalt principle of the focal point to highlight the observation of interest, and add a brief title and description of the insight, which triggers (we hope) long-term memory in the audience.

The appealing chart that results might be sufficient if it only need to appear in an internal memo, for instance. But if it is going to be part of a concerted communication strategy with the organizations that bankroll the charity (or some other stakeholders), some additional aspects of storytelling might help paint a more thorough picture.

While the specifics of the process may change depending on the data, the charts, the audience, and/or the story objectives, the general path is basically consistent.

Evolving a storytelling chart [@DV_NK].

Figure 4.27: Evolving a storytelling chart [3].

Data Storytelling Tropes

The step that transforms simple charts into appealing charts is often accomplished with the help of the Gestalt principles or with the help of data storytelling tropes, which is to say, data visualizations patterns and strategies that have become so familiar as to not need explanation (icons, conventions: see Forms and Structures).

Some examples include:

  • the inclusion of a trend line with any scatter plot, indicating the direction of the correlation between two variables (positive: going up; negative: going down), or the absence of any such correlation;
Based on [Galton's original data](https://www.medicine.mcgill.ca/epidemiology/hanley/galton/).

Figure 4.28: Based on Galton’s original data.

  • using a cluster bar chart with two categories where one is always lower than the other to showcase a “significant” difference between the two categories;
Image taken from [@globepap].

Figure 4.29: Image taken from [94].

  • a point of interest located at the intersection of two curves;
Equilibrium point at the intersection of supply and demand curves.

Figure 4.30: Equilibrium point at the intersection of supply and demand curves.

  • using gauge charts, pie charts, and donut charts in all manners of dashboards;
Chart from Andy Kriebel [@Andy].

Figure 4.31: Chart from Andy Kriebel [95].

  • the use of colour coding to represent political positions in maps and general charts (in the U.S.: red for right and right-leaning, blue for left and left-leaning; in the rest of the world, these are mostly reversed);
Based on [@dougherty2021hands].

Figure 4.32: Based on [96].

Canadian reversal of colours [Mark Gargul].

Figure 4.33: Canadian reversal of colours [Mark Gargul].

  • the use of broken axes to exaggerate the data story effect;
Chart created by Wikipedia user 'Vasyl 10'.

Figure 4.34: Chart created by Wikipedia user ‘Vasyl 10’.

  • the traditional use of the horizontal axis for independent variables and the vertical axis for dependent variables, which can mislead the audience into thinking there is a cause-and-effect relationship between the variables, etc.

There are other data storytelling tropes; we will no doubt continue to identify and construct more of them as data visualization and data stories become commonplace.


In the meals example above, the use of a bar chart for the initial (ugly) visualization informed the look and feel of the final data story. But it is possible that upon first visualization the data into an ugly chart, an intermediate transformation step will be required before we can create an appropriate simple chart.

The following BEFORE/AFTER charts show more examples ugly charts evolving into a compelling data story. Can you think of other ways to present the story?

Evolving a storytelling chart; from [@ricks].

Figure 4.35: Evolving a storytelling chart; from [97].

Evolving a storytelling chart; modified from [@DV_NK].

Figure 4.36: Evolving a storytelling chart; modified from [3].

Evolving a storytelling chart; from [@DV_NK].

Figure 4.37: Evolving a storytelling chart; from [3].

4.2.4 Anatomy of Storytelling Dashboards

Non-data stories have their own presentation conventions, which may, of course depend on the story format and genre, and on the audience expectations.

For data visualization, the exact nature of the presentation depends on the overall context and the amount of time the audience is willing or able to put into “consuming the story”.

Defining the visualization context.

Figure 4.38: Defining the visualization context.

The choice of “product” is prescriptive: we expect the audience to spend at most a few minutes (5, perhaps?) on a simple, non-evolved chart; if we know that our audience only has a few minutes to spare, it would be beneficial to produce a single simple visualization.

The same principle applies to the other types of presentations, each with a different time span:

  • dashboards (at most 12-15 minutes, say),

  • reports (at most 30-45 minutes, say), and

  • data art (no time limit).

Data stories are usually presented as single (evolved) charts, infographics, or storytelling dashboards.64

In the last section, we discussed how to evolve data into storytelling charts and infographics; in this section, we briefly discuss the composition of a (storytelling) dashboard.

The latter takes into account various elements:

  • the audience and its expectations;

  • the storytelling goals and the available data and analysis results;

  • the dashboard’s narrative approach, and

  • how that narrative is presented (its logic).

Practically speaking, the layout is informed by visual memory considerations (see Visualization and Memory).

Exploration, Situational Awareness, and Storybook Dashboards

Dashboards come in many flavours, although the distinctions are sometimes lost on users. In short,

  • exploratory dashboard focus on understanding the DATA,

  • situational awareness dashboards, on keeping tabs on a SITUATION of interest, and

  • storybook dashboards on communicating a STORY.

Exploration dashboards use visualizations as a tool to explore the data. They come with a high level of interactivity (filters, sliders, drill-down options, etc.) and offer high levels of detail: all aspects of data should be represented (tables, columns, calculations, etc.), not just summaries. Since such dashboards are usually prepared for small internal audiences with an in-depth understanding of the context from which the data arises, annotations (or explanations) are neither required nor useful, for the most part. Exploration dashboards capture as much interesting information about the data as possible, in as short a time as possible: the focus is on automatic chart generation, not on aesthetic considerations.

A dashboard exploring financial data for a fictitious organization.

Figure 4.39: A dashboard exploring financial data for a fictitious organization.

Situational awareness dashboards use visualizations as a tool to provide a “real-time” snapshot of the data. They are implemented with a medium level of interactivity, allowing for the focus to be temporarily directed to different departments or processes. These dashboards are not “scripted” as the focus is determining whether certain key performance indicators (KPI) are staying or trending above (or below) some pre-determined “warning” or “emergency” thresholds. As the dashboards may be regularly updated with new data, it is crucial that they be well organized and decluttered so as not to hinder the organization’s reaction speed, when needed. Situational awareness dashboards contain data summaries tables and charts and may feature anomaly detection.

A dashboard providing situational awareness of financial data for a fictitious organization.

Figure 4.40: A dashboard providing situational awareness of financial data for a fictitious organization.

Storybook dashboards use visualizations as a tool to explain the data. They are accompanied by a low level of interactivity, in part to ensure the audience does not trip over itself by modifying the charts and losing the story in the process (if some interactivity is provided, a reset option should also be present). Storybooks are characterized by low levels of detail: typically, numbers are rounded or altogether removed from the presentation, and only the key aspects of the data and analytical results are represented. As the target audience may have little data and domain expertise, annotations used to drive the story and to ensure that there is no ambiguity about what is being communicated. Storybooks present the results of quantitative data analysis as qualitative insights.

Storybook for stars and cultural constellations. [@visualcinnamon]

Figure 4.41: Storybook for stars and cultural constellations. [52]

Audiences (Reprise)

We have already discussed the importance of the audience to storytelling, and that it is preferable to avoid preparing a data story that will be all things to all people all the time.

Instead, we suggest addressing lines of business, such as Finance, Engineering, Marketing, HR, your immediate supervisors and team members, the minister’s office, etc. Doing so will allow the storyteller to identify the decision-makers and the role played by various audience members, which will, in turn, inform the storytelling process: if a minister is reputed to put a lot of faith in raw numbers when making decisions, it could be useful for a few such figures to appear on the dashboard.

This is not to say that analysts and storytellers must always cave in to their audience’s wishes. While we say that we want to make data-driven decisions and that we are data-friendly, the reality is that this is typically only the case as long as the data keeps supporting our beliefs and positions. It is good practice, then, to throw the audience a bone and help soften any blows which could lead to audience push-back and and reject conclusions that are not to their liking. This is best achieved, however, when the audience is known by the data storytellers.

But knowing who the audience is is not the same as knowing the audience, and what what relationship analysts and dashboard designers have with them. In particular, it is crucial to figure out how they perceive data and data analysis, and how trust and credibility can be established with them (heeding the lessons from the boy who cried wolf, presumably).

Another thing that might change from audience to audience is what they need the data storytelling for, what they need it to do for them. To get a sense for this, we need to know how the results will be used and what actions the decision-makers are likely to take:

  • what decisions are people going to make from the analysis?

  • how often are they going to be looking at the data?

  • how often do they expect the data to be refreshed?

We also need to know what the audience needs to know:

  • about data availability,

  • data cleanliness,

  • data governance and accessibility, and

  • whether the data it is being “massaged” or used to paint a rosy picture?

Finally, the audience’s need/wishes to interact with the charts will inform the type of dashboards that is produced: will they be passive receptacles for the dashboard? Are they expecting to be able to slice, dice, and filter the data and the charts?

All of this is part of the legwork (soft-skills) required before the story can be told, no matter what format that will take: finding the answers requires questions to be asked and answers to be gotten, which is definitely much easier to do when the audience is known.

Identifying and Gathering Presentation Requirements

The requirements for a dashboard, report, or presentation are driven by the primary consumers, the stakeholders that will primarily be getting “value” from using the product.

It is a (sadly) very common mistake to cast the net too wide and to build something for too many consumer types at once, if only because the answers to the previous section’s questions might then be all over the place and attempts to exhibit a uniformly coherent visualization deliverable will be stymied by the competing audience priorities.

Once the group of primary consumers has been identified, we suggest following a formal process to gather the visualization requirements as accurately as possible; these can be obtain by surveying the stakeholders and obtaining answers to questions such as (but not necessarily limited to):

  • what is the proposed name of the product?

  • who are the target data consumer(s)?

  • what are the product high-level objectives?

  • when does it need to be published/made available?

  • with what frequency is the data updated?

  • what kind of business decisions will be made by the target consumer group?

  • what are the sources of data?

  • is the data/information duplicated anywhere else (e.g., by a 3rd party)?

  • what is the sensitivity level of the source data?

  • what is the sensitivity level of the final product?

  • how is the source data gathered?

  • what quality assurance is performed on it?

  • etc.

This might seem like unnecessary work to add onto already busy analysts and dashboard designers, who mostly want to focus on the presentation and its contents, but removing ambiguities and ensuring that the visualization product is aligned with stakeholders’ needs will reduce the risk that the deliverables fail to align with stakeholders needs.65


Once we have a set of well defined requirements we are in a position to perform a storyboarding exercise. Storyboarding is a way to summarize the flow of information into a coherent whole, before we start working with data and software proper.

This helps the design team determine how many pages/elements per page might be needed to create an impactful data story. Note that this is NOT the same as designing the dashboard layout: storyboarding is used to define the story and the dashboard’s content.

Example of high-level storyboarding for a fictitious Government of Canada department.

Figure 4.42: Example of high-level storyboarding for a fictitious Government of Canada department.

In the age of multiple available software options for every aspect of the data analysis pipeline ranging from data collection to publication, it may seem counter-intuitive (if not downright retrograde) to eschew digital tools; however, storyboarding is a pen-and-paper exercise that may require a heavy time commitment in order to be conducted properly – this is another occasion where the external eye of a tsarnina of common sense can provide support.

Dashboard Narratives

We have briefly discussed narrative structures as they apply to general stories in Forms and Structures; what does that look like in dashboard terms?

There are a number of ways of constructing a narrative, including:

  • all events being presented in chronological order;

  • the most/least important event is presented first/last;

  • the first/last occurring event is presented first/last;

  • the most/least successful event is prepared first/last;

  • the worst/best news are presented first/last, etc.

When its applicable, we try to tell data stories in a number of different ways so as to maximize buy-in and minimize push-back.

A dashboard’s logic determines its format:

  • a horizontal logic dashboard is akin to a saga or an epic (such as Lord of the Rings, Babylon 5, or the Anne of Green Gables series, for instance), in which the chapter/episode/book titles, when laid out in succession, provide a summary of the story – in this analogy, the chapters/episodes/books correspond to the dashboard’s pages; the horizontal logic can be reinforced with an executive summary or report placed at the presentation’s beginning;

  • a vertical logic dashboards is closer to an anthology on a theme (such as Black Mirror, Dropped Threads, Billions and Billions, for instance), in which individual episodes/stories/essays are self-contained and logically linked to one another – in this analogy, the theme represents the data, and the various pages, different aspects of the data stories.

A hallmark of horizontal logic dashboard is that the order of its pages is crucial and must be respected when the story is presented: a missed page (or a page read out of order) changes the story entirely and makes it altogether unintelligible; the order of the vertical dashboards, on the other hand, is not: pages may be missed ommitted without “crashing” the overall dashboard story since each page stands as a coherent story in its own right.

In practice, dashboards may combine both horizontal and vertical logic: for instance, there could be three vertical segments, each of which consisting of a horizontal sequence of two pages each, say.

Putting it All Together

Pragmatically, all of these come together to suggest a storytelling dashboard’s guidelines:

  • an executive summary page, which provides the story outline and contents and addresses issues related to long-term memory retention by combining text and visuals;

  • at most 7 or 8 (and quite often much fewer than that) regular pages, arranged according to context and dashboard logic considerations;

  • each of which containing at most 4-5 decluttered, evolved, and annotated charts, employing pre-attentive attributes to attract the eye (iconic memory), a small number of charts (due to short-term memory limits), and text to drill the visuals into long-term memory.

A good general principle for storybook dashboards is that we should cut from them as much as is necessary to convey the story, and once that is done, to go re-visit and cut again as there is likely still too many elements on the dashboard. Less is more.

Here are two examples (which are way too busy to constitute anything other than a first pass at a storybook).

Based on an idea by [@DV_NK].

Figure 4.43: Based on an idea by [3].

Based on the 2012 _Gapminder Health and Wealth of Nations_ chart [@DV_Gapminder].

Figure 4.44: Based on the 2012 Gapminder Health and Wealth of Nations chart [87].


C. Nussbaumer Knaflic, Storytelling with Data. Wiley, 2015.
S. McCloud, Making Comics: Storytelling Secrets of Comics, Manga and Graphic Novels. Harper, 2006.
H. Rosling, The Health and Wealth of Nations. Gapminder Foundation, 2012.
S. McCloud, Understanding Comics: The Invisible Art. Harper, 1994.
K. R. Popper, The Logic of Scientific Discovery. University Press, 1959.
S. A. McLeod, “Karl Popper: Theory of falsification,” Simply Psychology, May 2020.
W. Van Orman Quine, Two Dogmas of Empiricism. Longmans, Green, 1951.
T. S. Kuhn, The Structure of Scientific Revolutions. University of Chicago Press, 1969.
Clayton High School, Crossing the line: Student sexual harassment & assault,” CHS Globe, Dec. 2011.
J. Dougherty and I. Ilyankou, Hands-On Data Visualization. O’Reilly, 2021.
E. Ricks, Declutter! (And question default settings),” Storytelling With Data Blog, May 2019.

  1. “The best-laid plans of mice and men oft go astray” definitely holds for data stories.↩︎

  2. Analysts/storytellers of the future: take note!↩︎

  3. It is far from a perfect science: J.K. Toole’s modern classic A Confereracy of Dunces only found a publisher posthumously, 11 years after his early death at the age of 31; cubism was widely derided as a movement when it first came to prominence, but it is firmly established as “real art” in the current era; achromatic music is still finding its footing and feels like it might never really outgrow its experimental/novelty status.↩︎

  4. Common descent (all living organisms on Earth descend from a common ancestor, E. Darwin) is often packaged with the theory of evolution by natural selection; common descent implies evolution by natural selection, but it is secondary to it. The former could be disproven by finding life that does not use DNA and RNA for information storage and retrieval (or by finding extra-terrestrial life, see Rational Wiki); while this would reduce the plausibility of the latter, it would not challenge its validity.↩︎

  5. Going from Newtonian mechanics to quantum/relativity theory requires a giant leap, not a series of small ones (fixed maximum velocity \(c\), the quantum realm running on stochastic processes, etc.); moving from Lamarckian biology to Darwinian natural selection requires abandoning the former model entirely, etc.↩︎

  6. In the Bill Nye-Ken Ham debate around “Is Creation a Viable Model of Origins?” held on February 4, 2014, the debaters were asked what would make them changed their minds on creationism: Ham said that he was committed to his views, being a Christian, whereas Nye’s reponse was that a single piece of evidence to support creationism (the universe is not expanding, rock layers can form in just 4,000 years) would cause him to change his mind immediately. One has the potential to be a data storyteller, the other, emphatically does not. We leave you to connect the dots and determine which is which.↩︎

  7. Some reports can take on a distinctly lyrical quality, and could be perceived as data stories, if it was not for their prohibitive lengths.↩︎

  8. Thankfully, identifying and gathering presentation requirements can be conducted simultaneously with the get-to-know-your-audience fact finding from the previous section.↩︎