Since the first appearance of GapMinder, thanks to the brilliant and inspiring talks of Hans Rosling, animation has gained substantial and increasing popularity as a way to highlight trends and patterns in complex data. Now people seem to want to show more and more fancy animations à la Gapminder in their own presentations, and an increasing number of new products have been appearing on the market that attempt to follow a similar visual style (see, for instance, MicroStrategy Dashboard and Report Portal as only a few examples).
Of course, data-driven animation in visualization is cool! No doubt. People seem to like it a lot (and I like it a lot too). As a result, powerful animation features have been included in many popular products (ever seen how many animation options MS PowerPoint has?). But the question still remains: does animation really help people perceive interesting patterns?
Several leading researchers from Microsoft Research and Georgia Tech, including George Robertson, Roland Fernandez, Danyel Fisher, Bongshin Lee, and John Stasko, investigated this particular issue in an InfoVis 2008 paper titled "Effectiveness of Animation in Trend Visualization" [PDF, microsoft.com]. In short, the answer is, quite similar to many questions in life, "well, ... it depends".
Robertson and his colleagues compared 3 different data visualizations using UN historical data on country performance indicators (so very similar to what Hans Rosling presented in his famous TED talk). They measured their effectiveness in terms of: accuracy, speed and subjective satisfaction.
Here are the 3 alternatives used:
- Animation: A standard scatter plot with animated bubbles. Meant to reproduce the same settings of GapMinder.
- Small Multiples: A static scatter plot matrix where each plot represents one country and traces show the evolution of bubbles (countries) from their initial to final position (see below on the right).
- Traces: A single standard (static) scatter plot visualization where traces are drawn one on top of the other (see below on the left).
The study interestingly distinguishes between 2 visualization purposes: presentation and analysis, and was specifically designed to compare the performance separately on the two. Accordingly, the answer seems to change according to what is the purpose of the visualization. But let's see the results first.
ResultsHere is a summary of the main findings:
- Small multiples was always more accurate than the others.
- Animation was fastest in Presentation (between 60-70% faster).
- Animation was slowest in Analysis (between 50-80% slower).
- People enjoyed and considered more exciting Animation.
- But when asked which technique they preferred there was no clear winner.
GuidelinesSo, what can you learn out of this? And how can this be translated into practical guidelines? Here is a summary directly reworked from the paper.
- Distinguish between presentation and analysis purposes: the same visualization technique might not be equally effective for analysis and for presentation. Different purposes require different solutions.
- Animation can be appropriate and exciting for presentation purposes: animation, as the results show, is appropriate for presentation. In particular, if you want to create a sense of engagement in your presentation animation might have an empathic effect.
- But don't expect your audience to fully understand the details of it: the study showed that people can get confused very easily and the overall accuracy in reading the data was always quite low with every technique used.
- And you need a good presenter and small data: good communication depends heavily and always on the quality of the speaker. GapMinder works great because Hans is a brilliant speaker first of all. A perfect technique won't cover the deficiencies of a bad speaker. Also, people can get easily overwhelmed by the data. Don't expect this to work with more than a bunch of data points.
Beyond the studyThe problem of animation extends well beyond the scope of this single study. Its real purpose is still highly debated and not well understood yet. Some researchers seem to suggest it has a very limited and narrow application, whereas some others seem to be more optimistic.
One area where its benefits seem to be widely accepted is the use of animation to ease the interpretation of transitions when a visualization changes its configuration. Some people are more interested in animation as a way to convey complex problem solving concepts, like algorithm animation. But the results are more controversial. Finally, the study of animation in environments where dynamic data are visualized (that is, when time passes new data come in) is definitely neglected and would need much more research.
Animation seems to be more widespread and accepted on the web. Great examples are Stamen's Trulia and FlowingData's Growth of Walmart. I don't think designers are too paranoid about animation, they just use it when they feel it is alright, and so in many cases it works. Brilliant results! However, the challenge is to harness its power and make the best out of it. One thing I know for sure: animation has a strong impact on the viewer and we should try to exploit its benefits as much as we can. I wish we will see more studies like this in the future.
And what about you? Any personal experiences with animation? Did it work? Anything more to say?
This blog has been written by Enrico Bertini. He is a researcher in the visualization and data analysis group at the University of Konstanz and he likes to deal with complex data analysis and visualization problems (a.k.a. visual analytics). Enrico regularly posts his ideas, reviews, and experiments in his blog fellinlovewithdata.com, where he tries to bridge the gap between academia and the real world out there.