The axes are labeled but I don’t know what the dots represent.

8a7a4e51-c3ca-46ac-9096-fde254e4fdd8-original

John Sukup writes:

I came across a chart recently posted by Boston Consulting Group on LinkedIn and wondered what your take on it was. To me, it seems to fall into the “suspicious” category but thought you may have a different opinion.

I replied that this one baffles me cos I don’t know what the dots represent! This is typical in graphs, the axes are labeled but I don’t know what it is that is being labeled.

Sukup replied:

Indeed. The axes are also labeled strangely and the years used to calculate the CAGR are also not the same. I’d have expected more from BCG—but this type of data visualization error seems pervasive these days!

15 thoughts on “The axes are labeled but I don’t know what the dots represent.

  1. And – where is the “strong” correlation? It is not so strong and it only considers a bivariate relationship. I’ve seen lots of this stuff going around – I had hoped that people would start realizing that 2 dimensional analysis is far from ideal, even if it is easy to draw a graph.

  2. I neither know nor care about “brand advocacy” but:

    (1) this reminds me of a recent quote, “But to see that you’d have to see the talk. I follow recommended practice by keeping the slides [graphs] simple and then saying a lot of words that are not on the slides. I explain the connections to unbiasedness [brand advocacy] in my speech, but there’s no way for someone to see this from the slides.”

    (2) Googling for 10 seconds, there’s more information at https://www.bcgperspectives.com/content/articles/consumer_insight_brand_strategy_brands_need_friends/ .

    This post seems excessively nitpicky. There are a lot of awful graphs out there — I can send you one of my recent favorites of swirling colored lines if you want — and this doesn’t seem very far along on the scale of awfulness. Yes, the labels are odd — the x-axis is presumably various customer company’s scores on this “brand advocacy index,” but so? And the scale of the dot sizes isn’t given, but this isn’t really relevant, since it’s presumably a relative scale for the quantities described at the bottom.

    • Raghuveer:

      I don’t see what your problem is here. All I said was that the axes are labeled but I don’t know what it is that is being labeled. I never said it was a bad graph or a bad presentation. Nor did I at any point say the graph is awful. All that editorial commentary is coming from my correspondent and from you, not from me!

      • Andrew:

        I didn’t understand your question. Each dot represents one firm. I thought.

        And the relative sizes of the dots denote the relative sizes of the firms.

        The y axis seems clear enough. The x axis is admittedly jargon. Some googling reveals BAI = Brand Advocacy Index. Apparently some consulting mumbo jumbo BCG is trying to push here.

        Or am I misconstruing your reason for bafflement?

  3. Why all the focus on the display and not the analysis? It looks like one dataset has 5 points and the other has 8 points – both with 2 dimensions. Yet, they call this “strong” correlation. I guess there is a third dimension since there is a size to the bubbles – but if that means that the data have been aggregated over another dimension, then the aggregate correlations should be fairly “strong” and these don’t look that meaningful to me. It is not the display that concerns me nearly as much as the content and message.

    • I don’t think the dot size matters here, that seems like a parameter extraneous to this particular regression.

      Unless BCG weighted the regression points by firm size. But I don’t think it did.

      Anyways, to an “eyeball test” the conclusion seems OK: Growth seems somewhat correlated to whatever they call BAI.

  4. For a consulting firm graph posted in a place like LinkedIn, this is a wonder of clarity.

    The y axis is growth rate. The x axis is some measure of brand advocacy by customers — without the operational definition, which would be too long to put in a graph caption — such measures aren’t understandable. Because the information may relate to different years, CAGR is used to provide a comparable growth measure.

    The size of the bubble indicates the size of the company, with the industries separated because it would be hard to get a common metric for banks and PCs.

    The main problem here is likely to be selection biases (e.g. there are a lot more banks than this), and other forms of distortion in order to line up with the point BCG is trying to make.

    • +1

      Not even selection bias by accident. I suspect in metrics like this there’s a lot of cherry picking going around. I wouldn’t be surprised if BCG actively tweaked the definition of BAI till they got a good fit.

  5. It’s another nit, but r = 74%, r = 77% are telling bad signs. No good story explains either. Either authors didn’t check carefully and miss that a square sign is needed for r-square, or they think that correlations should be be presented as they were percents, which shouldn’t survive completion of a first course.

  6. Regardless of the labels used in the axes: both linear models do not seem adequate and the residuals will probably tell the story by themselves. Displaying the R^2 as a justification for using the linear model is just another example of rote data analysis.

Leave a Reply to Dale Lehman Cancel reply

Your email address will not be published. Required fields are marked *