On Representing Aero Data

posted Aug 28, 2009, 6:29 AM by Donald Vescio   [ updated Sep 12, 2011, 8:10 AM ]
We all want to make sure that we're riding the most aero bike possible, but sometimes its difficult to sort through the different claims made by manufacturers regarding the performance of their products.  After working so hard on our training, nutrition, and the other physical and mental aspects of our race preparation, we do not want to undercut our chances for success by relying on less than optimal equipment.

Now, there are lots of problems associated with the aerodynamic data normally available to most of us consumers.  There is the obvious issue of accuracy--can we trust the absolute and/or relative data; but even if the data are accurate, there still is the question as to how it is represented.  How we interpret data is in part based on our reception of its representation, which often can be influences by emotional frames of reference or prior expectations.

Here's a chart that's widely used to discuss the aerodynamic merits of different bicycle framesets.  I edited out identifying information of the author of the chart, as this is not relevant to our current discussion.

Chart One

What Chart One does especially well is highlight the absolute differences in aerodynamic drag among a series of manufacturers' time trial bikes in a range of positive and negative yaw angles.  What this representation suggests is that there are significant performance differences in the series sample.  Note that this presentation of data focuses on a close up view of the drag figures posted by each tested frame. The y-axis values displayed range from 475 to 875 grams.

After a little quick and dirty image editing, I extended the chart's y-axis to run from 0 to 875 grams:

Chart Two

While the absolute values associated with each frame remain the same in Chart Two--and I'm *not* saying that differences in drag among the frames are not significant--the data normally presented now can be regarded in a very different contextual framework that lends itself to an interesting message.  What we now can focus are relative differences between samples, which might encourage us to make a difference frame selection based on collateral criteria, such as return on cost, etc.

Additional Comments/Observations:
  • Extending the y-axis down enables viewers to get a *relative* sense of the difference between frames so that they might not necessarily worry that they can't ever fit properly on a P4, as they might need a frame that's taller and shorter (and which doesn't have numbers as good as Cervelo's).
  • By including a traditional round tube, steel road frame in the graph, one also will get a greater sense of how much faster our current generation of frames are over a very low cost, *standard* frame. If the Centurion, for instance, does not track appreciably higher on the chart, then there are marketing implications, to name one problem (ie, "Does it make sense to spend x thousands of dollars on frame y?"; "What frame offers me the greatest return for my dollars/euros/whatever") If it does track appreciably higher, then a different message will be sent. What's missing in the original is a standard fixed reference.
  • Finally, by extending the y-axis, it makes it easier for the consumer to identify outliers and question whether frame Y is really that much faster than frame Z, or whether frame A is really that much slower then frame B.
  • By cropping the y-axis, differences between frames are exaggerated.
  • Basically, I'm a data person and feel very comfortable consuming information presented in a tabular format; what the two graphs do above, though, is point out how data representation *can* have a very non-rational impact on the consumer(eg, "If I'm not riding a P4, I'm giving up a lot of free speed").