Wednesday, December 12, 2012

the functional art

It's a little beat up and has scraps of paper sticking out in various directions marking pages I want to refer back to: it's my copy of the functional art by Alberto Cairo, and it's been everywhere with me over the past month - from LA to DC to Milwaukee, and several places in between (yes, that probably means I'm a slow reader). I finished it on my latest plane ride home. In a word: it was awesome.

As the subtitle declares, the focus of the book is information graphics and visualization. Alberto has a conversational, super accessible writing style. His research is augmented with his extensive experience in data journalism and the book is filled with illustrative stories and examples. the functional art begins with a section on Foundations - what visualization is and does, the importance of building a narrative structure, and introduces a visualization wheel to evaluate competing priorities. It then moves into Cognition - discussion of the eye, the brain, and how people see. The third section focuses on Practice - the infographic creation process and interactive graphics. While I enjoyed the entire book, the final section, Profiles, was my personal favorite - it recaps Alberto's interviews with various practitioners working with infographics. I also enjoyed examining the various examples throughout the book and seeing the progression from sketches to final product.

The book was inspiring, as is Alberto's work in general. I had the pleasure of speaking with him a couple of months ago, when he was getting ready to launch the first massive online course on an Introduction to Infographics and Data Visualization through the Knight Center. This 6 week course filled up quickly, so a second was recently announced that begins in January (details here). Alberto also teaches at the University of Miami and blogs at His passion in this space is clear and from what I've seen, permeates through all that he does.

For those interested in infographics or data visualization in general, I highly recommend the functional art; you can purchase a copy of it here.

Thursday, December 6, 2012

I like [candy] bars better than donuts

I've recently hosted a couple data visualization challenges (here and here), but it's been a while since I've made a contribution to one. That is about to change. Recently, Naomi Robbins announced a makeover challenge in her Forbes blog (details here - open until 12/9 if you want to participate). The challenge included two visuals. My discussion and makeover of each follows.

Makeover #1: I like [candy] bars better than donuts...
The first visual included two donut graphs (the pie chart's even less effective cousin; see here for a related post on donut charts):

It took staring at this for awhile for me to figure out what was going on. Based on the title, it seems we're meant to compare the segments of the donuts across the two charts. This is not an easy comparison - in addition to having to measure angles and areas (something our eyes aren't very good at), the slices are also in different places due to the difference in magnitudes within the donuts, making the comparison even more challenging. 

I also question the takeaway called out - that LinkedIn referral traffic is 16x higher for B2B companies. The proportion it makes up of total is 16x higher, but that's not necessarily the same thing, since total referral volume for B2B vs. B2B could be very different. Unfortunately, we aren't given the context of total referral volume here. I also hesitate to call out an increase like this, since the initial comparison point is 1% - a small number, making the 16x increase not necessarily as impressive as it at first sounds.

To Gavin's dismay, I'm going to go with a bar chart here (but bear with me, no bar charts for the next makeover, I promise!). Yes, I use bar charts frequently. It's because they are so easy to read! Here's where I landed:

In this case, we can easily compare the relative breakdown of referrals by source for B2C vs. B2B. For me, Facebook's dominance across both comes across more clearly here than it did with the pies. (I'm still questioning the use of %s here vs. absolute numbers, but given that's the data we had to work with, I'll let that concern go for now).

Makeover #2: Cole's first slopegraph
The second visual set forth in the contest is a table, accompanied by the challenge to "suggest a graphical representation" of the information displayed. The table shows expenditures for homeowners vs. renters in 1986 vs. 2010 across a number of categories:

In a recent workshop I facilitated, one of the participants asked about using line graphs when you only have two time periods and whether this is advisable. I think if I'd been asked a couple of weeks ago, I would have probably responded no. But I recently finished reading Alberto Cairo's The Functional Art (my book review post is forthcoming), where he features a couple examples of this done really well. I drew an example on the board in the workshop to illustrate when this can work. Here's my first go at creating one for real:

For me, the interesting story here is around which categories increased (depicted in blue above) and which decreased (grey). I think the slopegraph depicts this well. To reduce the business of the visual, I chose not to show all of the categories depicted in the original table. I omitted Entertainment and Other (since they both remained relatively flat) and included a note on this in the footnote. I also chose not to show the subcomponent pieces of the categories included in the table, but rather included some comments on what I found to be the interesting observations from those on the visual directly.

I'd love your feedback on the above visuals. I think in both cases, the story comes across much more clearly in the madeover visuals compared to the originals. If you have other ideas on how to visualize this data, leave a comment with your thoughts (or even better, submit an entry to the challenge!).

Saturday, December 1, 2012

and the winner is...

Thanks to those who submitted visuals in response to the recent data viz challenge (and thanks for your patience in waiting for this recap!). The goal was to turn a selection of the stats included in a recent Pew Research Center article on how teens research into visual form. There were four entries, which I'll recap here, in addition to announcing the highly anticipated winner!

Submission 1: Joe Mako
Joe took the item "Research tools teachers say their students are most likely to use..." and created a Tableau visual, commenting "I like diverging stacked bar charts for plotting the results of Likert and other rating scales. They enable you to simply see the overall direction and the detail at the same time."

I really like this visual: it's well labeled and organized, with thoughtful attention to design. For example, very light grey shading helps your eye read across horizontally from the label to the data it describes without being obtrusive, sorting in increasing order of unfavorable/decreasing order of favorable provides a nice construct within which to interpret the data, and your attention is drawn clearly due to strength of color to the tails - not at all likely and very likely - the most interesting pieces of data. Subtly labeling the data bars allows for easy comparison across the different research tools (which would otherwise be difficult for the tails due to the lack of consistent baseline).

In terms of feedback, there are three (relatively minor) things I would potentially change in this visual:
  1. Swap the formatting on the title and the takeaway to emphasize the main takeaway, that the internet is much more popular than resources at the library.
  2. Move the legend to the top of the graph, so the audience encounters what the bars mean before they get to the bars themselves.
  3. Correct typo: I don't think I would have caught this, but reader Rupert Stechmann did - there's a typo in the takeaway at the top where 'is' is repeated twice. Attention to detail is critical, and perceived lack of it can call the entire analytical process used into question.
The direct link to Joe's image can be found here.

Submission 2: Sam Feldman
Sam created his data viz on how students use their cell phones in class:

I give Sam high points for creativity - using color within the cell phone image to demonstrate the percentages. But it took me a little time looking at the visual to realize that's what's happening. In this case, I'd recommend a few modifications to make this visualization higher impact:
  1. Reverse the sort ordering so that you start with the biggest segment at the top, and work downward to the smaller segments. Then your takeaway could become: the majority of students use cell phones in class to look up information or take a photo/video for an assignment.
  2. Make the data in the cell phone stand out more: take away the sort of marbled shading within the different colors (it's distracting and doesn't add anything) and play with making the lines on the cell phone more subtle (try making them white or grey) - basically, I want the colored segments to stand out much more than the cell phone itself, but still preserve the ability to recognize it as a cell phone.
  3. Right justify the labels so each label is directly next to the portion of the cell phone it describes. Omit the boxes that tie the label to the given portion of the bar; instead make the label itself the same color as the bar it describes. There's plenty of space, so I'd consider making the labels larger as well.

Submission 3: Hrvoje Smolic
Hrvoje picked a pie chart visualization from the article and turned it into a slopegraph:

Hrvoje's blog post on this makeover can be found here. I agree that the relative increases and decreases are much clearer here than in the original pies. What I crave in this case are more words to make what we're looking at clear: let's state the main takeaway (what's interesting or noteworthy), add a graph title, a more descriptive y-axis label, and the data source. Now that we can see the data in a more straightforward fashion, let's think about what story we want to tell and draw our audience's attention to the relevant parts.

Submission 4: Jane Pong
Jane visualized how different groups of teachers perceive the impact that school policies have on their teaching. She converted a table included in the article into a more visual form (full size available here):

Jane's approach is similar to Joe's, plotting the Likert scale in horizontal bars anchored at zero. I think this is a good approach, but a little more labeling would help the audience to more quickly interpret what they're looking at here. It took me a bit of time to understand that the item text is what's shown at the left, and the breakdown at the right is income (mostly below poverty level vs. mostly middle/upper class) and size of area (large metro vs. small town). I'm not sure what the different colors represent, so we should make that clear.

This visual shows a lot of different comparisons - this is one of those cases where clearly identifying a single story or two that we want to focus on could be helpful for determining what data to show (perhaps we don't need all of this) and making it easier for our audience to consume. 

...and the winner is...
While each entry has it's merits, the winner of this data viz challenge in my opinion is Joe's diverging stacked bar chart. It clearly tells a story, both through words and through the accompanying visual, which is utopia for me when it comes to storytelling with data. Joe, I'll be reaching out with the promised offer to pen a guest blog post.

A great big thanks to everyone who submitted entries. I really appreciate the time you took and the work you shared!