The Design of Information

I keep hearing about Density Design’s RAW tool and had it filed away in the back of my mind to try out sometime. As usual, necessity is the mother of actually doing something. So, I did something, and now I am a happy camper. What a great tool!

I wanted to show a client how an alluvial diagram would do a good job of expressing some relationships in their data and was not thrilled at the prospect of trying to mock it up in Illustrator. In my search for an easier way, I stumbled across RAW again and had a vector sample in minutes. Brilliant.

Here’s a sample output (not the actual client’s data):

Thanks Density Design! Sad I waited so long.

I don’t think I’ve ever found a reasonable use for a coxcomb chart, until now. It seemed to fit the bill for a recent project for The Foundation Center. With funding from the Rockefeller Foundation, they recently published a synthesis of success factors for small-scale coastal fisheries management in developing countries. The Foundation Center hired us to create an interactive visual representation of their findings.

It appears that fisheries have to balance a number (twenty, to be precise) of factors on different levels to be successful. We ended up using a coxcomb chart (or polar area diagram or rose diagram) as the base for an interactive presentation built on raphael.js. The interaction allows you to visualize how different stakeholders might prioritize the twenty factors, as well as drill down to get more detail about each of the factors.

The coxcomb overlays do a particularly nice job (I think) of showing where there might be gaps in perspective that would encourage you to bring another stakeholder to the table. Overall, I think the impression is that there is a lot to keep tabs on if you are managing a fishery.

We developed all the data graphics (about 70 individual graphs and charts) for a report on Gender Diversity in company leadership that was published this week by Fenwick & West. The report compares the top Silicon Valley companies with the S&P 100. According to the survey, Silicon Valley companies overall have less female representation in company leadership than large public companies nationwide, though both seem to be increasing in representation over time.



This bubble chart gives you an overall sense of female representation on boards of different sizes. You can see that the S&P 100 companies tend to have more directors overall, and also more women directors. How long will Silicon Valley companies stay clustered at the bottom?

The Wall Street Journal picked up the story. You can download the full report from Fenwick & West’s website: Gender Diversity in Silicon Valley: A Comparison of Silicon Valley Public Companies and Large Public Companies.

I’m adding a few more graph samples from the Gender Diversity study, pertinent to an interesting discussion of appropriate line graph scales on Alberto Cairo’s The Functional Art blog and a discussion of slopegraphs on Andy Kirk’s Visualising Data blog.

For the Gender Diversity study, we chose to use a 50% maximum for the scale of the line graphs, with the idea that 50% represents parity. I suppose that sends the subtle message that parity is what we are shooting for, so you can visually see how far the line is from parity. For female representation at the largest U.S. companies, those lines are mostly still quite far below the 50% line, which makes it a little difficult to get a good sense of recent change. For some of the data, we included the average number of women in a position over time, using a scale which comes closer to Cleveland’s 45 degree optimum

I think I was first introduced to slopegraphs in Alberto Cairo’s book, The Functional Art (a book I recommend to anyone wanting to do a better job of presenting data). We used them to show the change of women in key positions from the beginning to the end of the survey period. “GC” is for General Counsel – the other abbreviations are probably more familiar.

A data-heavy project I’ve been working intensely on the last week or so was released yesterday. It’s a statistical review of corporate governance practices since Sarbanes-Oxley, done by the law firm Fenwick & West.

I enjoyed wrestling with Excel and Illustrator to create histograms, box and whisker plots and a few original creations. And the client was great to work with – detail-oriented, appreciative of good design, understanding of complexity. You can download the full report here: Corporate Governance Practices and Trends

It’s amazing how much more understanding you get out of a well-designed visualization than a spreadsheet of numbers. We went from something like this:

To these:

Is it strange that I love graphs so much?

Fenwick & West continues to keep me busy creating data graphics for a series of surveys they publish. Here’s one from their Technology and Life Sciences IPO Survey, plotting each deal by number of shares (log scale) and share price at the time of the offering. The bubble size represents the overall deal size. The data visualization software company Tableau made the chart this time – looks like a pretty big deal. We’ll see if these bubbles burst.

Threestory Studio’s second data visualization project with Silicon Valley law firm Fenwick & West was released to the public this week. This report looks at trends in venture-funded deals in the life sciences.

Though not as extensive or complex statistically as the first one (Corporate Governance Practices and Trends), this one presented some interesting challenges in presenting data clearly, accurately and concisely. I’m happy with the results.

With the presidential election fast approaching, interest in the predicted outcome is high. I’m impressed by detailed data graphics on Nate Silver’s FiveThirtyEight blog for the New York Times, not only for their clarity but for his thoroughness in examining the data.

I guess we’ll see about the accuracy of all these predictions after November 6.

Trust the Olympics to inspire some innovative data visualization. Thanks to friend Peter F. for the tip on a nice series of visualizations from The New York Times that compare today’s winning sprinters, swimmers and jumpers with past medalists.

Interesting to see the steady march forward over the years in swimming and sprinting.

Curious that there’s not a similar progression in jumping. Is that because the long jump is a more complicated venture than a pure sprint? Or just because there was only one Bob Beamon?

I was thinking about the visual display of uncertainty today and came across this nice example from a weather site in Norway. It shows the probable range of future temperature and precipitation levels for the city of Oslo. This is a good solution for something I’ve puzzled about for a while: When I hear that there’s a 30% chance of rain, I’m always asking myself “a 30% chance of how much rain?” A 30% chance of a light sprinkle is a much different forecast than a 30% chance of a deluge.

It would be interesting to know what factors go into the variability of the forecast. I imagine that the further out in time the forecast is, the more uncertain it would be, but there are obviously other factors that affect probability as well.

I also like their “detailed meteogram” with an hour-by-hour view of precipitation, temperature, and pressure, enhanced by an elegant indication of wind speed and direction, and topped off with an artful visualization of cloud cover.

Makes me proud to be 1/8 Norwegian.

I recently completed an online class, Data Visualization and Infographics with D3, taught by two great teachers: Alberto Cairo and Scott Murray. I have worked on a few D3 projects from the design side before, but this was my first real foray into doing the code myself. For class exercises, I picked a dataset to work with that I cared about: youth suicides.

You can see the interactive graphs I created here.

Clusters of suicides among young people in our community have, understandably, caused much concern. The school my children attend is highly competitive and full of students motivated to do well. One huge concern in the community is that school pressures are a major contributor to these tragic deaths. This has led to many discussions about homework, high expectations, class schedules, parental pressure, and more, with a strong undercurrent among parents and educators of a desperate need to change something.

The message I get from my son (a junior), is that the school is not the problem and the system shouldn’t be changed as drastically as some propose. He and many classmates feel like the proposed changes diminish the educational experience and are senseless.

All of this made me want to know if there really was an alarming trend here or not. How does our community compare to others? What does the data say?

I was initially relieved to see that our state and county were below the national average. Suicide data is not reported on the city level, so I tried extrapolating from what was available anecdotally for Palo Alto (a collection of publicly known cases). I was relieved again, until I realized I was extrapolating against the population of the whole city instead of the age-specific population that relates to the data. A more accurate estimate suggests that we are definitely on the high side.

With the small sample size of a single city (or even some of the smaller states), the data gets jittery. Pretty soon you are looking at individual lives – probably helpful if you really want to understand causation, though less helpful for seeing trends. I may need to take a look at three year moving averages to smooth out some of the jitter and see if that clarifies any trends.

My son saw me working on this and encouraged me to pull in the comparisons to national and county data for a clearer picture. When he saw the graph, he said “You have to share this!” – the power of accurate data displayed clearly.

Part of the challenge for the community discussion here is that one suicide is too many, so talking about comparative data can feel cold and dehumanizing. What wouldn’t we do to save even one life? The potential problem comes when you change whole systems based on a handful of tragic cases, and then later realize that you damaged the system and didn’t solve the problem you thought you were solving. I hear echoes of this challenge in what I have been reading in Daniel Kahneman’s book Thinking Fast and Slow regarding loss aversion and the way humans respond to risk.

As with many things, it’s complicated.

This has been a valuable, if painful, discussion in our community, causing us to examine what we really value and how that gets reflected in our education system. I hope some clear data can contribute positively to the conversation.