Portfolio management with bubble charts

Champagne bubble

With New Year’s Eve approaching, I checked the typical festivities around the world on Wikipedia. One of the common traditions is to drink champagne (or at least wine with bubbles). What better time of the year to discuss bubble plots (or bubble charts)?  A little research helped me trace back the origins of the bubble chart to William Playfair’s pie-circle chart (dated 1801) and most likely built upon by Minard around 1850 (check out Brinton’s 1939 history of data visualization).

Popularization in the innovation portfolio management domain follows from dr. Robert Cooper’s seminal work in the early 1980’s (as in this paper).

For those of you that haven’t seen a bubble chart yet: it plots projects as circles in a two-dimensional space (with an X and Y coordinate), with a third property determining circle area (bubble size), and a color scheme to separate different types of projects. 

Example bubble plot showing attractiveness (horizontally), value (NPV, vettically), cost (bubble size), and type (color).

Bubble plot showing attractiveness (horizontally), value (NPV, vertically), cost (bubble size), and subportfolio (color).

Now, in this example (and the rest of this post), the bubbles are linked to projects; for the bubble plot to work in portfolio management the bubbles should be linked to the individual decision items (bubbles should represent portfolio items that can be selected more or less independently).

The use of bubble plots for portfolio management

The  main role of the bubble plot in portfolio management  is twofold:

  • to compare projects on three dimensions during project selection:
    the bubble chart is unique in that it allows the comparison of three numerical properties of projects at the same time (X, Y, and size). This helps to focus decision-making on the trade-offs (between X and Y, e.g.g attractiveness and financial value) for the most important decisions first (the large bubbles).
  • to provide a balance view of the portfolio during portfolio review:
    By looking at how projects are positioned in the bubble diagram (combined with their size), the big picture of distribution of value, costs, or resources across the X and Y dimensions become apparent. Using the color codes for different types of projects allows further balance checks, for example balance over geographies, markets, innovation types etc.

A third use for bubble plots is to determine which indicators are helpful for better decision-making. If all the bubbles are clustered in a small area, meaning the X and/or Y values of the projects are very close, the properties we are comparing do not really differentiate the projects. This should lead to the conclusion that these properties are not the right properties to guide project selection for this particular set of projects.

Finally, during portfolio management process definitions, the bubble chart can be useful to settle the debate on which properties (or indicators) should be selected. Usually, the proposed set of indicators is quite large, and determining a meaningful subset by debate or reasoning alone is tricky. A practical way to solve this is to determine all these properties for a representative set of projects, and to check with bubble plots if any pairs of properties correlate well. This shows up in a bubble plot if the bubbles are all on (or close to) a straight line, especially the larger bubbles. In this way, we have found that, quite often, revenue and profit forecasts are strongly correlated, so either one of them can be in the key property subset. The same holds for development costs and resources, or for technology risk and IP newness.

What a good bubble plot tool should support

In this context of innovation portfolio management, what are the requirements for a good bubble chart? I collected requirements from the area of data visualization (and the link to decision support):

  1. The bubble chart should show all 3 numerical dimensions properly (X, Y, and size) :
    For the visualization not to be misleading, the X and Y axes should include the full range of options (not just the range in which the bubbles happen to be). This way the distances between  bubbles are meaningful. Also, the bubble area should reflect the third parameter, not the diameter, so that twice the area is twice the cost (or value).
  2. The bubbles should have a clear link to the projects they represent:
    It has to be clear on the spot which bubble is which project. This means that project names should be in or near the bubbles, at least as an option. As a minimum requirement, if showing all the names would clutter the diagram too much, a pop-up should help to answer the question which project we are looking at (see the example above). Trying to link the unique color of a bubble to a  long list of color-project links in a legend is not good practice (consuming a lot of chart space for the legend in addition to putting a lot of burden on the reader).
  3. Bubble charts should not be overloaded:
    The four dimensions (X, Y, size, color) are already tough for the human mind to juggle with at the same time. Proposals to include more dimensions in the same visualization do not seem to support better trade-of analysis or decision-making. Examples I have seen include 7 or more dimensions in one plot, with the addition of depth (for a Z value in addition to X and Y),  different shapes (not just bubbles but also rectangles, pyramids etc.), cross-hatching patterns, inner circles inside the bubbles, pies inside the bubbles, arrows to show vector-like displacement of bubbles over time, etc.

This last point should not be translated into restricting the analysis to just three dimensions. Portfolio management trade-offs are made across more than three dimensions in practice. It is just more practical to allow switching dimensions easily to explore the portfolio options in more dimensions than to try to picture them in complex visualizations. So what configurations make a good set for meaningful portfolio analysis?

Favourite bubble chart configurations

The following configurations for bubble plots are a good starting point:

  1. Financial versus strategic value, with bubble size related to costs and color related to type
    This is the layout in the above example, it supports trade-off between financial value and strategic contribution (see my earlier blog post here). The ideal project is a represented by a small bubble with high financial and strategic value.
  2. Risk versus reward (or value), with bubble size related costs or resources, and color related to type
    This is the most popular bubble chart according to Cooper’s research. It differentiates high-value high risk (breakthrough) projects from low-risk low-value (incremental) projects, to supports the discussion about risk appetite. In addition, it forces a challenging debate whether the ideal project (high-value low-risk) actually exist, and if so, where its value comes from if the innovation does not imply significant uncertainty (risk).
  3. Expected cost versus actual cost, with bubble size related to value and color to schedule progress
    This plot supports the analysis of progress, by presenting absolute cost and value metrics in a useful combination. The ideal project here is a high-value project ahead of schedule and under budget . Large bubbles get more attention, which is good for this case (larger bubbles represent higher value projects).

In our tool solution FLIGHTMAP, we support exploratory as well as predefined bubble settings (aptly called presets). I would like to offer a collection of useful presets, and share them via this blog as well. So let me know if you have favourite bubble plot configurations and which discussion they support.