Alternatives to a Log Scale

Over in our private Academy Slack group, one of our members asked a solid, totally not snarky question about log scales. They’ve been common in visuals about COVID and there’s a fair question out there about how appropriate those are in graphs aimed at public consumption.

Our World In Data uses log scales in their COVID graphs and one of our Academy members spotted this one on their local news:

Which can make it look like Mexico isn’t all that far behind Israel because their bar lengths are SORTA close, until you see, upon further inspection (and most people don’t further inspect), that there’s a log scale taking place and then your mind bends as you try to imagine what this data would actually look like until you give up or spot that Our World in Data has a button for you to change this to a linear scale, which looks like this:

Now the data is on a scale that is a bit easier for most people to digest.

So I was explaining to Academy members that people use log scales to help “equalize” the view of the data so you can just plain SEE some of the categories with smaller values. For better or worse. For the most part, log scales are really hard for non-scientific audiences to interpret.

One of our Academy members, Adria (Twitter, LinkedIn), was like “Hey, yeah, that’s me. I have non-scientific audiences – so what can I do instead?” She shared one of her recent graphs, which used a log scale (on overlapping bars with benchmark lines! I’m impressed!):

Converting this to a linear scale is going to make it difficult to see some of the smaller categories at the bottom of the chart. So I suggested (1) aggregating some of the smaller groups and then (2) pulling them out into a second chart, on its own, more visible, scale.

In our monthly Academy Office Hours webinar, I pulled some of the data on COVID vaccinations and played with a few alternatives, to illustrate these ideas for Adria and everyone in the Academy. I started by adding in some lines to connect the Sum of All Others column to the second graph that expands that column.

This data set is pretty large, and that’s making it a bit difficult to read the labels at this point. And, after seeing some of the other alternatives, Academy members ultimately voted this idea down.

They favored this version with the gray callout much more. In fact, in the original version I showed in Office Hours, the graph on the right had a white background and just the callout was gray. But members suggested that the second graph have a gray background to match my callout shape color. And one of the things I love the most about our Office Hours calls is how we collectively generate better ideas than any one of us could invent alone.

It is where Adria ended up.

Yep that’s the idea!

But in my COVID example, we are still limited by some space constraints, what with the size of this data set.

So one of the members on the call suggested that we rotate the top chart so that it is a bar instead of a column and place the second chart underneath it. Genius idea. Here’s how it would look with yet another formatting option – using color to link the Sum of All Others data to its blown out chart.

In all of these iterations, breaking the data into two charts seemed to be the clearest solution. But maybe this will inspire you to think of yet another version. There’s more than one right answer to this question and I love that the Academy gives us a space to explore all of them.

Once the question was in my ear, I couldn’t help but notice and share other ideas, too. This visual, from Laura Elliot, isn’t a log scale, but shows another option for expanding one section of a graph, into a second graph (here, a sankey diagram growing from a segment in a stacked column chart).

And this one was so inspiring, a member said “This is exactly what I need for this data problem at work – how do I make it?” and the Academy has the answers.

Learn in the Academy!

You can find step-by-step instructions on how to make 60+ awesome visuals in my Evergreen Data Visualization Academy.

Video tutorials, worksheets, templates, fun, and a big-hearted super-supportive community. Learn Excel, Tableau, R or all three. Come join us.

Enrollment opens to a limited number of students only twice a year. Our next enrollment window opens April 1. Get on the wait list for access a week earlier than everyone else!

Master Dataviz with Graph Guides!

Our newest program, Graph Guides, is a custom-built, year-long sprint through 50 Academy tutorials.

When you enroll, we’ll assess your current data viz skill set, build you a customized learning path, and hold your hand as you blaze your way to new talents.

We open enrollment to 12 students at a time and only twice a year. Get on the waitlist for early access to our next enrollment window.

Title Charts Like A Journalist

When you spend the bulk of young adulthood in research-focused academic institutions, like I did, you are steeped in a culture that tells you, explicitly or implicitly, that you can’t ever really make a claim. Taking a position on something can be seen as biased. Claiming an insight can be viewed as naïve – replication studies will be needed before anything approaching an insight is possible. Objectivity rules and subjectivity is foolish.

It is that kind of thinking that leads us to just “show the data” to our audiences and let them figure out their own insights. As if they want to do that work on YOUR data!

That culture of never really stating your point is what trains us to write unhelpful chart titles like “Figure 4. Unemployment claims, United States, 2020.”

Instead, let’s take inspiration from the way journalists title their stories. They include some action. “Job losses soar,” not just “job losses” or “unemployment claims.” They say “U.S. virus cases top world,” not just “U.S. virus cases.”

The step I’m asking for here can be as little as adding a word or two that takes a generic title to a place that conveys the urgency that the situation requires. Tiny step, bold move.

Academia might request an “objective” title, like “COVID cases by race/ethnicity” but there’s no such thing as objectivity. The data are always talking, it is just a matter of whether we are brave enough to tell its story. “Black Americans bear the brunt as deaths climb” does the data some justice.

We all have the obligation to do our data justice.

Now journalists take advantage of the story’s headline, not the title of the chart itself. I’m encouraging you to work with the chart itself because it is the chart that will get shared on social media and you want that awesome headline to travel with it. Use sentence case. Use action words. Add a period (or an exclamation point!).

It is ok to stake a claim. This is, in fact, why you have been invited to the meeting. Tell people the story.

My Cringe-Worthy Choices

The theme of this blog post can be summed up with one emoji: 🤦

If you aren’t emoji-fluent, that’s the facepalm, a gesture made when you are internally dying of embarrassment over someone’s incompetence. In this case, that someone is me. I have failed to recognize when I was using phrasing and framing that was inappropriate and offensive. I didn’t see it at the time – thought I was just being cheeky. In hindsight, 🤦.

So, shall we walk through some of my more cringe-worthy mistakes?

My Ableist Language

When you are an author, growing is a humbling experience because your thoughts are IN PRINT. In the first AND second edition of one of my books, I referred to something as “schizophrenic.”

Image of the passage from my book where I use the word "schizophrenic" to describe the formatting of a report.

Readers sent emails to me and my editors, linking us to the National Alliance on Mental Illness, which advises that mental disorders not be used as adjectives. Yep! And 🤦.

Mental disorders have not historically been on my radar, largely because I’ve been privileged enough to not have to experience them personally. This has also caused me to review how often I’ve used “crazy” or “bananas” or something similar to discuss a particularly busy chart and how I surely don’t need to be causing anyone who does have a mental disorder to bear any further stress by my thoughtlessness. Here are alternatives to consider. SAGE says they are replacing the word in the next printing of the book.

I didn’t realize how ableist I was until I started hosting people with total blindness in my workshops. The advice I got was “just explain, out loud, everything that you are showing on the screen” which, to be honest, was simply not practical for a data visualization workshop. I don’t have all of the solutions to this one sorted out but I can tell you that I have since become extremely aware of how often I say things like “Notice how in this graph…” and “See how in this graph…?” and “Look at this thing in this graph…” 🤦

My Violent Language

I listened to this beautiful conversation between Ocean Vuong and Krista Tippet, on the On Being podcast, where Ocean talks about how violence is baked into our language. Being successful is often phrased as “slaying it” or “killing it” and that Ocean had become uncomfortable using those words. That resonated with me so hard.

In my workshops I often talk about how if we avoid the use of the red/green color combination, we both handle colorblindness AND the reprinting in black-and-white issue. A couple years ago, I switched from describing this double solution by saying “we can kill two birds with one stone” because it felt meaner than I wanted. Now I say “feed two birds with the same seed.” Sometimes it takes me saying my words in front of big groups to really hear myself. 🤦

PETA got crap for it, but they suggest a bunch of alternatives to some of our violent phrases (and they used red/green 🤦):

Instead of "kill two birds with one stone," say "Feed two birds with one scone." Instead of "Be the Guinea pig," say "Be the test tube."

Whether you agree with PETA or not, their suggested alternatives are more creative and peaceful.

My Culturally Insensitive Language

One more 🤦 example for now. In my Data Visualization Academy, we rate each graph tutorial on a scale that tells you how much effort will be required. When our Graph Guides students learn 50 tutorials in a year, we customize their curriculum to span this scale. Up to this point, I referred to this scale as a “Ninja Level.” I also use this reference in my books and workshops:

a selection from my book that shows a bump chart with an icon of a ninja and a description that this chart is Excel Ninja Level 4.

I thought it was cute at the start. Graph Guide students referred to themselves as ninjas and I even saw someone who dressed up as an Excel Ninja for Halloween. But I saw some criticism that Tableau’s use of “Zen Master” to describe the people they train is appropriative. I agreed. And that made me reflect on my use of Ninja. It isn’t really different.

So we’ve swapped our tutorials inside the Academy. We now assign them Rockstar Levels:

And I’ll be switching to this language in my workshops and book (if/when I write another edition).

I know some of you are thinking “Where are the limits of this political correctness?” “What about when ‘Rockstar’ becomes cultural appropriation?” And that, friends, is how you board the slippery slope. Don’t get on. If people are telling you (or – better yet – if you are sensing) that you are using phrasing that is offensive, just change it. Don’t spiral into “what’s next OMG pretty soon you’ll be able to marry your dog!” DO think about how coming from a place of privilege often muddles us and that getting feedback from others is a gift. The act of being open to other people’s lived experience is what makes this so much more than just a swap of phrases.

Maybe one day Rockstar WILL be culturally inappropriate. If that day comes, I’ll keep evolving. Becoming more inclusive is an ongoing process. For everyone. Take counsel from Angelou: “I did then what I knew how to do. Now that I know better, I do better.”

Make Great Graphs in Google Sheets

If you’ve ever been in the audience of one of my workshops, you can tell that I was once a teacher. I use the same classroom management skills when I teach 100 adults how to graph as I did when I was teaching 25 kindergartners how to read.

Back when I was a teacher, I was putting in 80 hour work weeks and I definitely did not have time to learn to code on the side. My school district did not have Tableau (ok, Tableau wasn’t even invented back then but my district would have been too cash-strapped to purchase us all a license anyway). Yet I still needed to make clear, story-telling graphs for my families, my committees, my boss, my boss’s board, and even my students.

People in education know how to maximize the resources available to them, In fact, I’ve seen that resourcefulness in every industry. These days, organizations all the way from schools to Fortune 50s are running in the Google environment. It makes collaboration easy and it is affordable. Awesome! Let’s learn how to use it to master data visualization, too.

Welcome to the Google Sheets Chart Starter Series.

Now, to be honest, Google Sheets is not designed to be a flexible data visualization tool. But it CAN do quite a bit and I’ve pulled together 10 tutorials on how to make the most of its graphing capabilities. Learning the skills I’ll teach you in these 10 tutorials will completely change the way you and your teams are able to use data for decision-making and action-taking.

First, I’ll walk you through my Four Step Process for making amazing visuals. You’ll use this strategy whenever you approach data so that your graphing process is streamlined and a story pops out at the end.

Then I’ll give you an orientation to Google Sheets. If you’ve never used it before, it can be hard to know where to start (so many wordless icons everywhere!). We’ll focus on just the parts needed for graphs and even make a clear and simple graph in the process.

After that, I’ll take you through 8 more tutorials on graph types that speak to the data scenarios I bet you are dealing with at work. They’ll range from easy to WOW and you’ll come out the other side feeling like a total rock star.

You’ll have lifetime access to these 10 tutorials so that if you have to go focus on lesson planning or a return to in-person work for several weeks, we’ll still be here for you when you are ready to dive into your data again.

Maybe, at some point, you’ll even want to step further into data visualization and try out the capabilities of more sophisticated software. Cool. You can join the Academy at a discounted rate at that point.

But my hunch is that many of you just need to know how to make impressive graphs with the tools you’ve got in front of you and then get back to your (extremely full) day job. Don’t let code bros bragging about their D3 skills on social media make you feel inferior or intimidated. EVERYONE can make great graphs, and that includes you.

Proportion Plots

Proportion plots help us compare the share of a population between two metrics. It uses length on the left and right side of the chart and connects the lengths by a band in the middle that swoops a lot if there is disproportionality and stays pretty even if the proportions are the same.

This is the kind of story we often need to tell when we are trying to talk about how our sample compares to the overall population. If your sample is representative, each side of your proportion plot would be equal. Often times, such as with my example below, this is not the case.

We have major disproportionality between the racial makeup of our overall population and the racial makeup of who is killed by police officers.

I made this graph in Excel after seeing the visual produced by Statistic.ly, shown here:

and I’m calling it a Proportion Plot. I can no longer find this image on their website, only on their Twitter account. I pulled my data from their same source, the Washington Post, and you can see how WP visualized this same data here. Here’s how another group visualized it. I think the Proportion Plot is the best.

My clients think so too. I mocked up data on their students in special education, compared to their student population overall and showed this to them in a Zoom meeting.

Responses were “Yes.” “YES!” and “YAAAAAAAAAAAS.”

In working with the development of this graph (rebuilt in Excel primarily by Joe Travers) and experimenting with different data sets, it became clear that when one racial group composes a large proportion of your population, it gets difficult to see how groups that compose a smaller proportion stack up. That can often lead analysts and designers to collapse some of the smaller numbers into one group that they call Other. And that can “other” people and contribute to exclusion. I asked around and learned that “Other races” or “All other races” were phrases that felt more comfortable to people usually clumped into that category, though they still aren’t perfect. Please keep this in mind when you are adapting this chart for your own data. Don’t be like CNN, who offended many indigenous people by using “something else.”

And yes, please adapt this chart for your own data. You can download my Excel file here. It’s a beast, I know. I included some basic instructions if you want to make your chart from scratch. You can also just work from my finished chart in the file. We’ll have step-by-step instructions (written and video) for this chart in my Data Visualization Academy soon – for Excel, Tableau, and R – including how to add or remove more categories. Stay tuned.

Meanwhile, as I watch the news report on yet another unarmed Black person killed by police, it seems to me that one of the many actions I can take is to release this file so that you can use it to tell clearer stories about disproportionality with the urgency those stories deserve.

Follow Up Q & A

I got some great questions about this graph, as one would expect when viewing a new chart type.

Q: Isn’t this just a Sankey diagram?

A: Nope, though they do look visually similar. Sankey diagrams depict flow. So categories on the left break apart and flow into subcategories on the right. A Proportion Plot is not about flow, it is about comparing the proportion of the same categories in two different settings. No subcategories involved. In a Sankey, flows can cross. That wouldn’t happen in a Proportion Plot. Here’s what data similar to Proportion Plot on deaths by police (this is completely fake – I just made these numbers up) would look like in a Sankey. I used SankeyMatic to make this.

Q: Isn’t this just made from a stacked area chart?

A: Yes, I hacked it from an area chart but it is not an area chart. Stacked area charts are typically used to show change over time and proportion plots do not.

Q: But I am reading this as an area chart, as though it is change over time.

A: Yeah, you might have to get out of your own way on this one. Big time data nerds (hey, that includes me) often get our minds fixed on certain ways to read charts and it can take a beat for us to open our minds up to something new. People who are not big time data nerds tend to have less of a problem. Try it! Other people already have given it a go and produced some awesome results. Check out Nick’s proportion plot:

Originally tweeted by NickVsPPT (@nickvsppt) on January 20, 2021.

and this one from Richard:

Decolonizing Data Viz

What is Data Visualization?

A visual representation of
quantitative or qualitative data.

so says Stephanie Evergreen.

The strength of this definition is that it is so broad, lots of things fit under its umbrella. But perhaps that is also its weakness because it leaves what counts as data open to interpretation and this is where the problem lies. Often, people with power and privilege are the ones who get to do the interpreting about what’s in and what’s out. What counts and what doesn’t. The history of data visualization (and often history more generally) is written by white men. And history is upheld by them too, even among the data visualization crowd of today, who cull lists of historically important figures and decide whose stories will be remembered, whose work will become “foundational.”

The accounting of viz history is rooted in Europeans and their wars: Minard’s alluvial diagram of Napoleon’s march,

Nightingale’s coxcomb on soldier deaths in the Crimean War,

Scaife’s History of the Civil War in the United States.

The scope of our field’s history, like so much of the way history has been framed, is Eurocentric in view and rooted in conquest.

The three previous examples – and many more along this same theme – are part a gallery of historical visualizations at Stanford, where the artifacts collected (by a white man) reinforce a position that data viz as a field originated during Enlightenment – so European roots, 17th and 18th century. The curator says as much in his introduction to the exhibit (emphasis added):

Here is another one from that collection, a “new chart of history” made by a European white man. It is hard to see, what with it being dwarfed by the size of the Roman Empire, but America is at the top and its roots barely extend past 1300 Common Era.

Just as this was Priestley’s way of defining what counts as history, the collection of historical data visualization is a way of writing data viz history that omits many people, specifically people in indigenous nations. 

Harvard is preparing to release a new book on the history of data visualization (can you guess who it was written by?) that also claims that the beginnings of graph communication were in the mid-seventeenth century, invented by a white European dude (again, emphasis is mine).

This framing of data visualization as a movement born from white people in Europe 1700-1800 CE falls somewhere in the range between lazy and racist. Of course data visualization existed well outside of these bounds. There is extensive evidence to back this up and I am not the right person to curate that exhibit but I will share just a couple examples.

As early as 1500 CE, the Incans were using knotted cords, called khipus, as a form of data visualization.

The color of the cord, the location and size of the knot, and the way the cords were tied to the primary cord all encoded data.

From Gary Urton’s chapter in Their Way of Writing. Urton, the primary researcher of khipus, is a white man who worked as a professor at Harvard until multiple women came forward with stories of sexual harassment.

Khipus were used quantitatively, to record census data including clan, social rank, occupation, and tax payment. They were also used qualitatively to record stories.

When we talked about these khipu and other forms of indigenous data visualization in a recent panel (with January O’Connor (Tlingit, Kake, Alaska), Mark Parman (Cherokee), & Nicky Bowman (Lunaape/Mohican)), someone in the audience commented, “It made me reflect on traditional Hmong clothing and how my ancestors have embroidered certain motifs on traditional clothing to communicate one’s clanship, what dialect of Hmong one spoke, marital status, everyday life, etc.” And this is one reason why it is so critically important to decolonize data visualization. When white men decide what counts (and doesn’t count) in terms of data, and what counts (and doesn’t count) as data visualization, and what counts (and doesn’t count) as data visualization history, they are actively gaslighting Black and Brown people about their legacy as data visualizers. When we shine a light on indigenous data visualization, we are intentionally saying the circle is much much wider and, as Nicky Bowman said, “There’s room for everyone in the lodge.”

Here’s another example by Wendy Red Star:

Wendy Red Star annotated a photo of Peelatchiwaaxpáash to highlight the data being encoded all over his hair and regalia. These annotations (a popular tool in data visualization) point out data related to wars that took place in the 1800s.

Why is this not a part of the data visualization collection pulled together at Harvard or Stanford? It is because you aren’t familiar with this form of data. It is because you don’t know how to read it. And what could just sound like a You Problem is actually an Us Problem. White people actively write about what counts and what doesn’t and we do so in a way that paints a history and paves a future that looks a lot like us. Sometimes I think that the formalization of data visualization as a field is so recent that people in it like to think it is somehow post-racial but in actuality, data visualization has repeated and reinforced the same pattern of white privilege that is present everywhere.

Indigenous Data Visualization is not something that somehow stopped being a thing when Charles Minard was born. Though I’m specifically drawing a comparison to the history being published about data visualization today, if you look at sources outside of the white males with the loudest voices in data viz right now, you’ll find plenty of examples of more modern indigenous and non-westernized data viz. It is time for white people to pay attention to and amplify those voices. And if that’s you, and you need a platform to share your knowledge and skills, let’s talk. I can host you here, connect you to other platforms, or help you build your own.

“Decolonizing data viz” is explicitly intended to honor the work of Abigail Echo-Hawk, at the Urban Indian Health Institute, who coined “decolonizing data” as a way to talk about the movement of the center of data collection and interpretation from white people to indigenous communities. Data visualization needs the same movement. This is a great read from Bhakti Shringarpure (via Vidhya Shanker‘s Twitter feed) on the popularity of “decolonize” movements and how to do more than make it a hashtag.

How to Avoid Making THAT MAP (and Other Mistakes Graphing Census Data)

My hot take: Let’s pair a new Census with coronavirus vaccine distribution. A door-to-door campaign. Because Census workers are usually local and trusted and skilled at finding and counting people without doors, too.

No doubt, despite the likely incomplete data from the 2020 Census, we will start seeing updated graphs that talk about the demographics of our country and estimates of its future. So I’m begging extremely hard that we avoid making stuff like this:

Do you, uh, spot any issues with the way the data are presented in this map? Such as the suggestion that all Asians moved to Canada in 2010 and by 2060 they’ll occupy a sliver of Maine and only Maine? Only White people live in Alaska and Hawaii?? This is so bad it is laughable.

And the danger is in how credibly this was shown to the world. This graphic is sourced to Pew Research Center – a very reputable organization. And was presented on Nightly News with Brian Williams, which, at the time, was very reputable.

Here’s what happened: Pew Research Center had prepared this data as a stacked area graph. Now, I think a stacked area graph is still a weak choice. For example, it can make it look like the Black population is decreasing (the blue stripe in the graphic above) even though the numbers are going up, because of how the White population (the red section in the graph above) is placed at the bottom of the chart.

But then the Nightly News team thought they needed to spice up the Pew data a little more for prime time. So they used a mask feature in a graphic design software to place the cutout map of the US on top of the area graph, obscuring some of the data and adding geographic-racial ties that are inappropriate and hilarious.

Saurage Market Research flipped the layers in the visual to show the stacked area on top of the map cutout so you can kinda see what was going on:

What can we learn from this grand mistake so that we do not find ourselves in similar situations as we make 2020 Census data viz?

Hemingway said, “The most essential gift for a good writer is a built-in, shockproof, shit detector.” and I’d like us to extend that notion to dataviz as well. See, I’m not trying to blame some overworked and underpaid graphic designer, though there was surely one who actually pushed the buttons in Adobe Illustrator to make this visual. There was also an entire production team who saw this visual at some point and let it continue to move forward.

Now, I’ve been fortunate enough to have production teams in my work and I can attest (and my readers let me know) that even when there are many eyes on the same piece of work, mistakes still make it into print. Perhaps, as is often the case for me, everyone was just moving too fast and not slowing down long enough to comb carefully for errors.

Perhaps the people involved did not have sufficiently well-honed data literacy skills to identify a problem. But I’m guessing that somewhere, at least one person had a built-in, shockproof, shit detector and spoke up about the introduction of errors in this graphic. And their objections were dismissed.

Whichever the reason (likely one beyond what I’ve suggested here that is either way more exciting or way more boring), checklists can help us catch mistakes. Checklists request that we pause and review. Checklists request that we obey their authority, bringing power to a situation where some people might feel powerless. “The checklist says you shouldn’t have extraneous decoration in your chart” can spur more corrective action, like it or not, than “I think the map isn’t helpful.”

I’ve combed the research to pull together the Data Visualization Checklist. It has been put to formal tests and used by thousands of people to prevent them from making the mistakes that were made at Nightly News. That map only scored 40.9%.

You can view the breakdown of that map’s score here. Run your own graphs through the Checklist and get help on any points where you don’t score well.

Beyond the visual, let’s also keep in mind who the Census counts and doesn’t count. If you take your Census data timeline back far enough, your dataset will exclude many people. For example, this map conflates population with Census data, absolutely excluding Indigenous people, Black people who were enslaved, and white women. No one in these groups was counted as a person, according to the Census, for quite some time.

Communicating Data is About Handling Egos and Emotions

Behind every furrowed brow and annoying question about your slide is someone’s ego getting dinged. People don’t like their egos dinged.

This tiny bit of emotional intelligence will give you massive insight into how to handle tough data communication scenarios we all find ourselves in at one point or another. I used to take disruptive people personally, like I had failed, until I got some confidence that my visuals and my content were, in fact, super helpful. That confidence (that comes with experience) is what helped me see disruptive people in their own light.

I had one the other day. Someone in my workshop was being a bit difficult and I handled it with some humor (more on that below) and put the question out to Twitter to see how others handle those situations. Soti’s compassionate way of handling someone disruptive revealed – BINGO – there were some bruised egos and emotions brewing.

Let’s walk through some common data communication scenarios, take a peek at what’s really going on with people’s egos and emotions (assuming your data, graphs, and slides totally rock), and think through an appropriate response. I’m going to offer multiple responses so you can pick what feels comfortable to you, and knowing that the appropriate response will vary depending on the power dynamics of the situation.

The Behavior Exhibited

Someone doesn’t like your graph and is dismissing you by questioning how you collected and analyzed the data.

The Emotional Ego Underneath

They don’t like the numbers the graph is displaying because they aren’t what this person hoped. This person has a personal stake in how those numbers look and if performance isn’t good, there may be consequences. Will I get in trouble for these numbers? How will I explain this to my boss? Could I get fired? Who would hire me during a pandemic? (People go down the emotional rabbit hole real fast.)

The Insightful Response

You can be prepared with a detailed description of your methods and analysis and sometimes that works. But it is trying to logic your way through people’s emotions. If your documentation isn’t cutting it, that’s how you know you have an emotional issue. Test the waters quickly with “I’m happy to send you the detailed documentation that back up how solid our methods are, given the budget and parameters we have.” If the frown is still there, try to distribute the burden with “We all want these numbers to look better than they do.” Then present it as an opportunity: “Thankfully, we have caught this issue now, so that we have the chance to turn this around before it gets worse. Let’s problem solve together.” Empathy is usually the best response to try, the first time someone is disruptive.

The Behavior Exhibited

Someone doesn’t like your graph and is making that clear by loudly proclaiming “I don’t understand this!” and “I can’t even read this!” and, even after some patient explanation, “I just don’t get it.”

The Emotional Ego Underneath

They don’t like your graph because they aren’t familiar with the graph type. They don’t know how to make that chart themselves. They don’t feel they will be able to learn to make something so intimidatingly cool. Their skill set is becoming obsolete. Can they make it to retirement without having to go back to school?

The Insightful Response

Help them feel like they are not alone, left out to dry: “Many people who are accustomed to the small selection of default charts in Excel don’t like this one at first glance…” then add in something that addresses the underlying issue: “… and that is often because they don’t know yet how to make it.” Then some reassurance: “I made this one right inside Excel and it is surprisingly easy. I’ll show you and once you know how to make it, you’ll warm up to it more.”

The Behavior Exhibited

Someone doesn’t like your chart type because “my boss won’t let me get away with that.”

The Emotional Ego Underneath

If I upset my boss, I could get fired. (So much of our ego response is based in threats to our underlying sense of security. Can you see that in these emotional ego thought patterns?)

The Insightful Response

In my case, I can usually reassure them that their boss, in fact, hired me to come in as the consultant and make these changes, so the green light has been lit. If there is still some hesitancy, I ask: “Even if you couldn’t pull off this entire visual revolution overnight, can you get away with implementing some small aspect of it? Maybe just the great title or a switch in colors? If you can make tiny changes every few months, in a year you’ll get there. Sometimes it is an evolution, not a revolution.” Make it feel do-able.

The Behavior Exhibited

Someone doesn’t like your graph and is making that clear by proclaiming “That just isn’t the way we do it. People won’t understand this change.”

The Emotional Ego Underneath

What they are really saying here is that the proposed change deviates from the norms. And disruption is bad, people will get upset about change and I’ll have a lot of upset people on my hands. (Now the original person is bringing in the emotions and egos of other people too, see how this works?)

The Insightful Response

This person needs to see that change is not necessarily bad just because it is change. Pitch a response that is contained in a halo: “The hero of every story becomes the hero because they made a change to the status quo.” Nice, right? Address concerns: “Some people might drag their feet on this but other people will be by your side and we’ll grow as a team through any discomfort.” Then appeal to some other emotions: “Taking people through change is what leaders do.”

The Behavior Exhibited

Someone doesn’t like your graph and is making that clear by picking at some minor aspect of your slide, distracting the conversation from your main point.

The Emotional Ego Underneath

They actually really like your slides and are so impressed by the design they are scrutinizing how you made and didn’t even realize that the conversation was elsewhere. You made good design look so easy they think they can do better than you so they are taking mental (and verbal) notes about what they’d tweak.

The Insightful Response

Humor works best here (at least for me). Laughter is the expression of emotion, so make a joke like “That’s a fantastic observation! I’ll have you make all of my slides next time.”

Another big piece of protecting egos and emotions is how we deliver the insightful responses. For some people, they’ll need to be addressed privately, pulled to the side during a break in the agenda or debriefed after the fact, like Soti did. For other people, the more gregarious or intentionally incompetent crowd, humor and confidence go a long way.

When I first became a parent, someone gave me some sage advice. When your fresh baby is screaming and red-faced, it can look like anger but it is probably one of a small number of underlying conditions: hunger, tired, lonely, or sick. What’s showing up on the surface is rarely the full story.

The more we can train ourselves to hear the emotional need behind the “difficulty,” the better we will be at communicating with data AND bringing everyone with us.

So What?

In the introduction to our dataviz workshop at a Fortune 50, the Chief Operating Officer told the room of his employees that he was looking forward to seeing their work improve as a result of what they learned with us. Because, he said, what they wanted to see on the slides in their decision-making meetings was “your claim and the evidence that supports your claim.” Can I get an amen?

He wants to see something like this:

I don’t think he is unique at all. In fact, I am of the belief that most of us are filtering what information we choose to consume by how well it answers “So What?”. What’s your point? Many, perhaps the majority, of data visualizations are not answering “So What?” for our intended audiences and are under-used as a result.

Let’s walk through some examples, improving as we go.

Example 1: That Table

You’ve seen tables like this throughout your life, I’m sure. Tables presented in this typical, research-y way are a struggle for many folks because there is no evident point, no clear answer to “So What?”. The title doesn’t provide insight. There many be interesting points contained in the table, but it takes insider knowledge of the study, of how research is conducted, and of the possible implications of the data presented here to be able to answer “So What?” and that, my friends, is a tall order.

Example 2: Tesla’s Range

We are improving in that now we have a visualization of the data. The chart itself is a fairly familiar type – line charts are ubiquitous (even though this isn’t change over time, meaning a line might not be the right option).

Can you answer “So What?” when you look at this? All lines are going down as we move to the right, so you might answer “So What?” with something like “As speed increases, miles decrease.” But there are 4 lines, so maybe the answer to “So What?” is in the fact that the purple line is lowest? Or the red line is the highest? I don’t know. It’s a guessing game.

The real kicker here is that the text preceding the chart acts as if data literacy is commonplace and that the point will be evident if you just LOOK at the chart. But the issue is that there are so many take-aways in this chart that even the data literate are stuck asking “So What?”. The title needs to tell people the answer.

Example 3: 5G vs. Corona Scatterplot

Check out that crystal-clear title: 5G doesn’t spread the coronavirus. Perfect! That answers “So What?” instantly. Excellent! But….

It still takes a LOT of data literacy skill to see how (or whether) that point is made evident in the chart itself. Scatterplots can be complicated for many viewers. We could actually take this as an opportunity to teach data literacy by adding some elements to the chart that would help readers interpret it, such as a second, comparison chart that would indicate what a strong correlation would need to look like. So, in this example, the title is on point but the graph needs a tweak in order to support that title.

Example 4: Britain’s Coal

This chart from the Guardian nails it. The title is a short, succinct point that quickly answers “So What?”. The graph shows the evidence to support that point by popping out the days at 0% coal with an eye-catching green. It works.

It works even though the chart type is novel. Instance charts are a fairly recent evolution, used to mark the extent of something at regular occurrences. Despite needing to learn some data literacy in how to read the chart if you’ve never seen it before, all of the elements (including the legend) in this visual are working together to answer “So What?”. And because this point is clear in the title, it makes learning how to read the chart easier. It teaches some data literacy.

Shaky Data Viz Advice

The biggest a-ha moment that came from my dissertation was discovering what shaky ground we stand on in data visualization.

When my friends heard I was going to study data visualization, they filled my desk with books from Edward Tufte, Stephen Few, and even Garr Reynolds. I was thirsty for resources and references because this was back in 2009, my young reader, before data visualization was so firmly established as a field. So I went down the reference rabbit hole. Oh, you know what I mean. You get one reference and you look up their references, read those articles, look up THEIR references, and so on until you know everything there is to know on the topic. Let me tell you: I didn’t get far.

Tufte, Few, Reynolds, and others who could be considered the elders of our field don’t have much to support their declarations and assertions – the ones we hear echoed, almost taken as “common sense,” still today. They cite incredibly little actual research-based evidence to back up their claims. Tufte has 5 in his famous The Visual Display of Quantitative Information. The sources they do cite are shared – meaning, they cite the same people. Few cites Tufte. It’s cute until you realize the implications.

You might think that they are excused because we didn’t have any data visualization research back then. But we did. Sure, less of it, but it was there. And in places it wasn’t, perhaps they could cool down the aggressiveness of their “standards.”

And they could make reasonable assumptions based on research from related fields. That’s what I had to do for my dissertation. I combed through literature in survey design, user interface testing, graphic design, typography, presentations, and more.

The fact is our heroes and history stand on shaky ground. Even modern books on data visualization lack research-based references.

One of my mentors is infamous for saying that not everything needs a randomized controlled trial to prove it is true or good. He said, some things are so obviously good or true, they have “interocular validity” – the kind that hits you right between the eyeballs.

But my mentor has something in common with Tufte, Few, and Reynolds. He was a privileged white guy, so his view of what looks good to his eyeballs is not necessarily shared.

The problem with interocular validity is that it means: If I think it’s right, it’s right. This puts us into difficult situations, where we are left tossing opinions. And opinions conflict. The Branding team at Nike may have had the opinion that this chart is good…

… and you might think they have a green obsession. It is no longer just a difference of opinion when the deep research on color by Dr. Cynthia Brewer gives us evidence to the contrary. Research gives you a backbone.

Communications may have thought this Tesla chart looks right…

…even if you are the internal data person trying to argue that bar charts should start at 0%, especially in this case, where a truncated axis makes them look safer than reality. When you have research to back up your recommendations, it turns your shaky ground a little more solid.

Now, research conflicts, too. But at least actual data gives us something to stand on (and this is why all Academy tutorials come backed by research).

Let’s knock down a few other pillars of viz history.

Ever seen this classic backbone as to why we should visualize?

The oft-cited 60,000x justification sounds sexy but has no basis in research. The reference rabbit hole on this one leads to a memo produced by 3M’s Visual Systems Division. Convenient that the Visual Systems Division would make such a conclusion about visuals, huh?

And this classic viz, produced by John Snow, is heralded as an example of the power is visualization. You probably know the story – during the cholera epidemic people thought the disease was spread by air. Snow marked cholera deaths on a map – one black line per person – and noticed they circulated around a water pump on Broad Street so they just removed the handle from the pump and the epidemic was over. Hero.

In reality, the notion that cholera spread through the air was accepted thinking until after Snow’s death. This viz may have been one drop of evidence but it didn’t have the heroic impact we like to think.

Let us question the foundation of our work because what may look legit to your eyeballs could be full of myth and conjecture.

Evidence, even if it is still evolving, informs decision-making. Heck, that’s why many of us are interested in data visualization in the first place, right? So let us not accept anything less in who we refer to for our data visualization guidance.

This is why my books have references at the end of every chapter. I am not messing around. My books are a good starting point. To nerd out more, investigate IEEE and Multiple Views.

From the blog