EDS 240: Lecture 8.1

Data storytelling

Week 8 | February 26th, 2024

Good data visualization design considers:

  • data-ink ratio (less is more, within reason)
  • how to reduce eye movement and improve readability / interpretability (e.g. through alternative legend positions, direct annotations)
  • putting things in context
  • how to draw the main attention to the most important info
  • consistent use of colors, spacing, typefaces, weights
  • typeface / font choices and how they affect both readability and emotions and perceptions
  • using visual hierarchy to guide the reader
  • color choices (incl. palette types, emotions, readability)
  • how to tell an interesting story
  • how to center the people and communities represented in your data
  • accessibility through colorblind-friendly palettes & alt text (see week 2 discussion)

Our job is to turn values and analyses into insights and narratives – doing so truthfully, effectively, and compellingly is all a part of data storytelling.

We won’t be covering anything new, per se



From a technical perspective, you have the tools in your tool kit to build data viz that tell stories.


We’ve seen many examples of data visualizations that are constructed around a narrative (whether or not we explicitly called attention to it).


You have all listened to, read, and told stories (in this class, in other classes, in life).

Why is storytelling important?


“…the very act of telling a story makes people trust you more… Data doesn’t change behavior, emotions do”

-Karen Eber in her TED talk, Why storytelling is more trustworthy than presenting data (2:45-7:07)



“Stories are inherent to us. We’ve been told stories since we were children. Stories are how we learn and how we make decisions.”

-Mike Bugembe in his TED talk, Lies, racism and sexism: The power of data stories (13:00-17:50)

I also recommend watching The Power in Effective Data Storytelling, by Malavica Sridhar and Why Data Storytelling Matters to All of Us, by Jia Hwei Ng

Telling a story involves . . .


adding context

makes it relevant / mean something to your audience


adding narrative

you want stories to be told in a way that everyone can understand


creating visuals

brings your narrative to life / makes it real

Let’s practice storytelling


Don’t advance the slides until I say so!

On the next few slides, we’ll see a series of data visualizations. First basic, then with additional layers added. These are data you’ve seen many times before.


For each version, consider the following questions:

  • who / what is the subject?
  • who might the audience be?
  • what message / narrative pops out, if any?
  • what context is needed to better tell this story?

Tell a story about this plot

03:00

Tell a story about this plot

03:00

Tell a story about this plot

03:00

Recreated based on Cédric Scherer’s visualization as showcased in his talk, ggplot Wizardry: My Favorite Tricks and Secrets for Beautiful Plots in R

Complete code (in case you’re interested)

#..........................load packages.........................
library(tidyverse)
library(palmerpenguins)
library(showtext)

#...........................load fonts...........................
font_add_google(name = "Montserrat", family = "montserrat")
showtext_auto()

#.................get and assemble penguin image.................
# only need to run these first two lines once to download image: 
# url <- "https://raw.githubusercontent.com/allisonhorst/palmerpenguins/master/man/figures/culmen_depth.png"
# utils::download.file(url = url, destfile =  here::here("week8", "images", "penguin.png"))
my_img <- png::readPNG(here::here("week8", "images", "penguin.png"))
my_raster_img <- grid::rasterGrob(my_img)

#...........................build plot...........................
ggplot(penguins, aes(x = bill_length_mm, y = bill_depth_mm, color = body_mass_g)) +
  geom_point(alpha = 0.8, size = 2) +
  ggforce::geom_mark_ellipse(aes(group = species, label = species),
                    color = "black") +
  scale_color_gradientn(colors = c("#F3E5B1", "#6C7B1A", "#044148")) +
  guides(color = guide_colorbar(barwidth = 20, barheight = 0.5, 
                                title.position = "top", title.hjust = 0.5,
                                ticks = FALSE)) +
  scale_x_continuous(breaks = seq(25, 65, 5),
                     limits = c(25, 65),
                     expand = c(0,0)) +
  scale_y_continuous(breaks = seq(12, 24, 2),
                     limits = c(12, 24),
                     expand = c(0,0)) +
  labs(x = "**Bill Length** (mm)",
       y = "**Bill Depth** (mm)",
       color = "Body Mass (g)",
       title = "**Bill Dimensions of Brush-Tailed Penguins (*Phygoscelis*)**",
       caption = "Data: Gorman, Williams, Fraser (2014) PLoS ONE | Illustration: Allison Horst") +
  theme_light() +
  theme(
    plot.title.position = "plot",
    legend.position = "top",
    text = element_text(family = "montserrat"),
    plot.title = ggtext::element_textbox_simple(size = 20,
                                                margin = margin(0.5, 0, 1, 0.25, "lines")),
    axis.title.x = ggtext::element_markdown(size = 15,
                                            margin = margin(1, 0, 0, 0, "lines")),
    axis.title.y = ggtext::element_markdown(size = 15,
                                            margin = margin(0, 1, 0, 0, "lines")),
    axis.text = element_text(size = 10),
    axis.ticks = element_blank(),
    legend.title = element_text(size = 14),
    legend.text = element_text(size = 10),
    plot.caption = element_text(face = "italic",
                                margin = margin(2, 0, 0, 0, "lines")),
    panel.grid.minor = element_blank()
  ) +
  coord_cartesian(clip = "off") +
  annotation_custom(my_raster_img, 
                    xmin = 56.8, xmax = 65.8,
                    ymin = 22, ymax = 30)

Note: The positioning / size of text, penguin image, etc. was chosen based on how it looked as rendered in the slides. You may need to modify for viewing on your own device.

Scrollable data stories help to weave longer narratives

Take some time to explore these with your learning partners and be prepared to share out some thoughts.

08:00

We’ve shown that static viz can also tell stories


The next ten slides (14 - 23) each have a different static visualization. With your learning partner(s), discuss the following:


  • what story(ies) is the viz being used to tell?
  • what design elements (e.g. text, color, images) do they authors use to help tell the story?
  • how do the authors provide context (may be related to your above answer)?
  • what is the purpose of this viz (e.g. to answer a question, spark additional questions, etc.)?
  • what do you find effective? what don’t you find effective?


Note: You don’t need to get to every viz! There are lots of options so you can explore those that peak your interest most.

08:00

The Jan. 6 inquiry, by the numbers (New York Times)

China’s population continues to shrink as deaths outnumber births (New York Times)

Wildfires in Canada burned this much land so far this year (Reuters)

‘Catastrophic’ California oil spill kills fish, damaages wetlands (Reuters)

Access to clean fuels and technologies for cooking (Reuters)

Almost half of U.S. cacti are vulnerable to extinction (Reuters)

Lionel Messi: all 672 goals for Barcelona by season (Reuters)

Refugees resettled in the United States (Reuters)

How employed people spend their day (New York Times)

“Hockey Stick” graph, by Mann et al. (1999)

Anytime we present data, we have the opportunity to tell a story


oral talks


academic journal articles


blog posts


dashboards


infographics / static visualizations

A few last important reminders before you jump back into storytelling. . .

Identify your audience


To tell a clear story, you need to know your audience

  • Which story is interesting for them?
  • What are relevant details to include?
  • Which variables are meaningful to them?
  • How will they encounter the visualization?
  • Do I need a visualization at all?


Data viz can look wildly different depending on how you answer the above questions!

From Cédric Scherer’s talk, Effective Data Visualization and Graphic Design with ggplot2 workshop

Consider the audience


Seven endangered species that could (almost) fit in a single train carriage, by Mona Chalabi (two of seven shown here)

03:00

Consider the audience


Fig. 3, by Ingeman et al. 2022: Glimmers of hope and critical cases. Find figure caption online.

03:00

Consider the audience



Plants in Danger, by Cédric Scherer (code)

03:00

Understand your data & be accurate

This has less to do with the physical act of creating data viz / telling a story and everything to do with your own critical examination of your data and domain expertise. Two important reminders:

  1. Our data is never a perfect reflection of the real world – our data is always a sample

From Cédric Scherer’s talk, Effective Data Visualization and Graphic Design with ggplot2 workshop

Understand your data & be accurate

This has less to do with the physical act of creating data viz / telling a story and everything to do with your own critical examination of your data and domain expertise. Two important reminders:

  1. Our data is never a perfect reflection of the real world – our data is always a sample

From Cédric Scherer’s talk, Effective Data Visualization and Graphic Design with ggplot2 workshop

Understand your data & be accurate

This has less to do with the physical act of creating data viz / telling a story and everything to do with your own critical examination of your data and domain expertise. Two important reminders:

  1. Our data is never a perfect reflection of the real world – our data is always a sample


  1. The best use of data is to teach us what isn’t true – you should be challenging yourself to find out something with your data, not to show that your assumptions about what you’ll find are true

Avoiding Data Pitfalls, by Ben Jones

From Cédric Scherer’s talk, Effective Data Visualization and Graphic Design with ggplot2 workshop

Take a Break

~ This is the end of Lesson 1 (of 2) ~

05:00