What Can’t Be Built with Bricks?

In my recent article for the Visual Business Intelligence Newsletter, titled Building Insight with Bricks, I introduced “bricks” as a new way to display quantitative values geospatially (e.g., on a map), which in theory can be read and compared more quickly and precisely than bubbles. Here’s an example:

Since the publication of the article last week, my concern for a particular limitation of this approach has grown. I mentioned in the article that bricks, unlike bubbles, do not work when they overlap. Even though I recognized this deficiency from the start, I didn’t fully appreciate at the time how much this deficiency limits the usefulness of bricks. Comments from readers, however, have raised my awareness. I especially appreciate the response from Andy Cotgreave, who took the time to mock up examples of bricks vs. bubbles to illustrate the problem, and from Joe Mako, who challenged my assumption that geospatial displays don’t usually involve overlapping values.

While discussing this issue with Joe, I asserted that business uses of geospatial displays don’t typically involve overlapping values. Joe challenged my assertion, inviting me to defend it. As I began to construct my case, it gradually dawned on me that overlapping values are more prevalent than I imagined while designing bricks. This oversight occurred, not because I’m not familiar with broad and common uses of geospatial data visualization, but because I had narrowed my focus to a subset of use cases and concluded too swiftly that this subset was much larger than it actually is. I never stepped back to recognize and sufficiently test my assumption. Even an informal round of peer reviews involving some of the brightest minds in the field didn’t draw my attention to this oversight. I suffered from a blind spot while designing bricks that I never managed to correct.

Bricks are still useful; just not as broadly useful as I imagined and hoped. I failed to add as much value through the invention of bricks as I expected. I’m disappointed, but not discouraged in my effort. I was trying to solve a very real problem, and even though I’ve potentially solved it to a lesser degree than intended, I’ve succeeded more thoroughly perhaps in raising awareness about the problem. I’ll keep working on it, and I hope that you’ll join me. This is how science works. Our limited successes and even our total failures are useful and as such ought to be shared. Others can learn from our mistakes, but only if we make them known. I invite you—challenge you even—to succeed where I failed. If you do, I’ll be content that I contributed through my failure to your eventual success. After all, the benefit that our work delivers to the world is all that ultimately matters.

Take care,

12 Comments on “What Can’t Be Built with Bricks?”


By Michael Thompson. April 29th, 2013 at 4:04 pm

Steve – being a bit hard on yourself. But we wouldn’t expect anything less from you.

Nevertheless: the point on occulsion and overlap was demonstrated well, but in that example (hypothetical) I think it reflects more a matter of a hypothetical failure of the hypothetical designer to regulate the reader’s experience through the right filtering and summary. The examples Andy used (with all due respect, much due) were static and without a clear question or use case at stake. But given both a clear question and a use case, I believe they will work, and I’ll be trying them myself. I would push deeper, rather than retreat.

By Stephen Few. April 29th, 2013 at 7:08 pm

Michael,

I’m not being hard on myself. There’s no shame in partially or even entirely failing in an attempt to do something worthwhile, especially when you share what you learned with others so they can avoid the same mistakes.

By Michael Thompson. April 30th, 2013 at 11:49 am

Fine: to the heart of the matter. I don’t think it failed. I think it is a work in progress. I will be working on it myself.

By Stephen Few. April 30th, 2013 at 12:05 pm

Michael,

Go for it. My point is that it failed in part in that it is more limited in its potential use than I anticipated.

By Steve Wexler. April 30th, 2013 at 1:22 pm

Stephen,

I’m sorry to hear that bricks are not as broadly useful as you first hoped as I think they solve a huge problem in how they help people “see” the differences in area.

I think you are correct with your assertion that our perception of area differences is poor. I’ve conducted informal studies with students where I show circles of various size, asking how much larger one circle is than another. In the case where one circle is twice as large as a second, most students think the ratio is 3:2 and not 2:1. That’s a rather large “you’re off my this much” factor.

I still think there is something very worthwhile here. My deep gratitude for the contribution and to Andy and Joe for showing me where it will and won’t work.

At least where it won’t work for now…

By Stephen Few. April 30th, 2013 at 4:26 pm

Steve,

There is no question that we cannot compare areas well. This has been well established by research for many years. To demonstrate this fact to students, I often ask them to compare the sizes of two circles, the larger of which is exactly 16 times the size of the smaller. In a class of 70 students, judgements usually range from five times bigger to fifty times bigger. After this demonstration, everyone sees from themselves that visual perception does not handle area comparisons well.

By Vicky G. May 2nd, 2013 at 4:16 am

3D visualization is a way to help with this problem. For example, GeoFlow for Excel 2013 provides some nice tools that take the visualization 3D.

By Stephen Few. May 2nd, 2013 at 8:37 am

Vicki,

Actually, 3-D visualization doesn’t address this particular problem at all. The problem has to do with the type of object that is used to encode quantitative values on a map, not with the 3-D vs. 2-D perspective of the map itself. I was asked for feedback about GeoFlow by the team that developed it at Microsoft before it was released, so I’m familiar with its capabilities. While it has its uses, it has little use for viewing quantitative values on a map. GeoFlow’s views are not designed to feature values on top of them, and when we examine geospatially located values, we want the map to be 2D and flat so we can see all of them without occlusion. GeoFlow is useful for virtually exploring actual physical space, much as you might see it when walking or flying around in it. It serves a different purpose from the maps that we use to explore and analyze quantitative value on a map. If you disagree, please explain how GeoFlow is useful for geospatial quantitative analysis.

By Peter. May 4th, 2013 at 9:26 am

Also, in small numbers, like in the example above, they look enough like Tetris blocks to actually produce some visual tension. The eye initially tries to fit them together. Less of a problem once you go beyond 3×3 or so, as you do in one of the maps in your original piece.

By Stephen Few. May 4th, 2013 at 11:29 am

Peter,

Not being a Tetris player, this particular tendency doesn’t exist for me. Even for those that have played Tetris, I suspect that this learned response could be easily unlearned with a little practice.

By Peter. May 6th, 2013 at 12:18 pm

Tetris, Lego, Minesweeper … they just feel to me like they suggest a tighter arrangement. But you’re right, that might be my problem :) and I’m sure it could be unlearned.

By Jeffrey Shaffer. May 17th, 2013 at 7:33 am

This is a very interesting discussion. I’ve always considered the “pie charts on maps” as one of the unsolved data visualization problems. Maybe we should post a list of these separately so people would actively offer solutions to them.

My immediate reaction to this solution was that the “bricks” were simply a variation on Unit Charts, which you’ve cautioned against in the past (i.e. Unit Charts are for Kids). I guess given the choice of evils I might opt for Unit Charts over Pies. It does seem you’ve at least offered what might be a better alternative in some cases. I think the best solution for this problem is having more robust tooltip features. Rather than having the standard text tooltips, I would like to see vendors add full function graphical tooltips. For example, the user should be able to select 2 or more points on a map and pop-up windows would allow for bar chart, line chart, histogram (etc.) comparisons. The closest I’ve seen is a Tableau trick for putting bar charts in a single tooltip, very similar to what Moritz Stefaner did in the OECD Better Life Index. Using this type of solution would allow the pie charts to simply be a quick reference clicking icon.