The following comments are in response to Stephen Few’s recent newsletter entitled “Information Visualization Research as Pseudo-Science” in which he critiqued an academic paper by Borkin et al entitled “Beyond Memorability: Visualization Recognition and Recall“. I’m not an academic researcher, so I will leave it to others in the field to respond to Few’s specific criticisms of the paper’s methods. My goal in this article is to respond to opinions Few voiced about memorability in data visualization.

I’d like to start by asking a few questions:

  • Does it matter whether a data visualization is memorable or not?
  • Should we, as data visualization practitioners, care about memorability?
  • Should we design our visualizations so that those who view them are more likely to remember them at a later point in time?
  • Is memorability a worthwhile area of study for those studying data visualization in academia?

In my opinion, and in my experience, the answer to each of these questions is ‘Yes’.

In Stephen Few’s recent newsletter entitled “Information Visualization Research as Pseudo-Science”, though, he put forward a differing opinion:

“Visualizations don’t need to be designed for memorability— they need to be designed for comprehension. For most visualizations, the comprehension that they provide need only last until the decision that it informs is made. Usually, that is only a matter of seconds.” – Stephen Few (emphasis his)

This statement helped me understand why Few and I disagree about memorability: we disagree about how data visualizations are used by groups of people. Simply put, I don’t believe data visualizations are “usually” followed by decisions “only a matter of seconds” later. That may be how a robot or a computer algorithm would approach decision-making, but it’s just not how groups of humans in organizations go about it.

How do groups of humans usually work with data visualizations, then? Well, analysts prepare dense packets for pre-reading materials, directors and VPs attend review meetings where they look at lots and LOTS of data and charts, sometimes they take copious notes, sometimes they zone out and check their smart phones, then they break for lunch, check their email, reconvene and consider different topics, only to have the final decision made at a totally different planning meeting or off-site weeks later.

Sound familiar? That’s a whole lot messier than question -> visualization -> decision in seconds. And that’s only one reason why memorability matters.

In my experience, the memorability of the overall message (of which the visualizations are a critical element) matters most when:

  1. Decisions won’t be made immediately
  2. The audience doesn’t care deeply about the topic
  3. The environment is already saturated in data and visualizations

To illustrate these three conditions, let me relate a personal story from my experience working with data and groups of decision makers. The specific details of the account have been altered to protect the innocent.

A Practitioner Wins Thanks to Memorability

One time I had the unenviable task of presenting the results of the launch of a product that was, shall we say, less than “top-of-mind” to the executives at a Fortune 500 company. Think “razor” of the razor – razor blade model. Sales should just be a pull-through, so they didn’t pay much attention to it at all.

But what we were finding was that the relative neglect of this high-touch product was causing a lot of dissatisfaction, and our lack of attention to the details of the product offering was causing us to lose customers.

In preparing for the presentation, I created plenty of nice, Tufte-compliant charts and graphs, like this one (a generalized mock up), to show how the recently-launched product was doing in the marketplace:

ChartNoPic

A comprehensible but not particularly memorable chart

Do you notice the problem in the chart? That’s right, we didn’t launch a green SKU in Configuration B.

Why not? Tooling investment.

Who cares? Customers did. A lot of them. The nature of the product was such that customers couldn’t select between A & B. There were factors that pre-determined that for them.

Now I was scheduled to be the fifth presenter in a very long review meeting where many other topics would be discussed, and as I mentioned, this product just didn’t matter to the executives. My charts were going to get glossed over. If the executives gave me 10 seconds of attention on each chart, I’d have considered myself lucky. The way the situation was shaping up, I felt pretty sure that this product line’s issues weren’t going to be addressed as a result of my presentation.

So instead, I showed charts like this, with actual photographs of actual customers and their actual quotes:

ChartWithPic

The same chart made more memorable by the addition of a human’s face and their own words

The result was palpable.

They leaned in. They looked at the faces in the pictures. Actual customers. People that looked like their sons, their daughters, their mothers. They chuckled at the funny social media handles. They cared. For the first time in a long time, they actually cared about the razor. And they cared about the fact that customers just weren’t loving it.

A few weeks later, I received an email that the go-ahead had been given to resolve a number of problems with this product line, including the missing green SKU in Configuration B. The VP thanked me for showing the “human side” of the data in my presentation.

When the time came to make the decision, they opted to fund a product they didn’t used to care about, thanks to charts they couldn’t forget.

Memorable or Comprehensible, or Both?

Stephen Few made the statement that comprehensibility matters, but memorability doesn’t when it comes to designing data visualizations. Well the original charts in my real-life example above were definitely comprehensible. I changed them because they weren’t particularly memorable.

My original charts were in the bottom right quadrant of the 4-blocker below, and all I did was push them up to the top-right. Sure, sometimes, it’s not necessary to do so. Sometimes, though, it’s make-or-break:

memorablecomprehensible

Note that for scenarios where the audience members already deeply care about the data, comprehensibility itself will result in memorability. Adding photos of beautiful, smiling faces just isn’t necessary.

But let’s be honest. Having an audience of 100% of the key decision makers that wait with bated breath for our next bland chart that results in a blank check being given right there on the spot just isn’t normal. It would be nice, sure, but how many times have you actually been in that situation? So many times you absolutely need them to remember your message. Having charts that draw them in and stay in their brains just isn’t a bad idea.

Sometimes There’s Just No Decision

So far I’ve written about data visualizations in the context of human decision-making. But many data visualizations don’t inform decisions at all. Decision support is but one of many possible purposes. Data visualizations can be created to merely inform, to educate, and yes, even to entertain. In those cases, design for memorability can be the difference between having someone share your work with others, and having them forget they ever saw it.

Few made the following comment about adding images to visualizations:

If I incorporate an image of a kitten into a data visualization, I can guarantee that a test subject would remember seeing that kitten if it is shown to her again a few minutes later. But how is that useful? Unless the visualization’s message is that kittens are cute and fun, nothing of consequence has been achieved. – Stephen Few

He answers the question himself quite well: images are useful if the visualization’s message is enhanced by the presence of the images.

Take my Edgar Allen Poe timeline for example:

Does the image of Poe add any value at all? How about the image of his signature? Are these components nothing more than “chartjunk” (Few mentioned to me in an email that he would not call the image of Poe “chartjunk” based on his 2011 writings on the subject), or do they actually perform a function?

I submit that they perform a vital function. The visualization shows the life works of one man as blocks stacked together in the years they were written. Works that were written on ink and paper by his own hand.

There’s no decision here. The visualization is simply intended to educate you. And it’s my opinion that your education takes on a whole different meaning – a whole different feeling – when you see Poe’s face and an artifact of his own penmanship.

And let’s be honest, the following version is pretty damn boring, you’d probable ignore it if you saw it in your twitter feed, and it’s not nearly as memorable, is it?

I’d like to conclude by quoting from Stephen Few’s critique one final time:

The greatest tragedy of this research is that what makes a visualization memorable is actually of no consequence. – Stephen Few

I hope I’ve made it clear in this blog post why I think that memorability can actually be of great consequence in data visualization. But did you notice that in my comments above I used phrases like “in my experience” quite often, and that all I really did was relate an anecdote and state my opinion? My opinion does not amount to codified knowledge, and my experiences do not amount to rigorous research.

And that’s exactly why I would appreciate further attempts by academics to study what makes charts more or less memorable. I’m sure this task isn’t easy. Visualizations are but one piece of an overall message that can be delivered in myriad ways to a variety of audiences. For those who are studying this topic, do know that there are practitioners out there who are hoping that the insight you glean into this topic can help us all.

Thanks,
Ben