I have just stumbled across this blog while trying to find out how to make a multiple pie chart exactly like the one described above. Funnily enough the reason I want to use a multiple pie charge with each pie scaled to represent relative size is because it is a complex and confusing way to show data. When we start advising clients they often have multiple investment portfolios spread across a range of different asset classes. This makes it hard to see exactly where their assets are invested. I want to show a chart that illustrates the chaos of the current situation and then compare this to the simplicity of a coherent investment portfolio (using a simple bar chart).

I note that you say this can be done. How would I go about designing such a table? Also, if it turns out that this doesn’t work as well as I envisage, how do I create the 2nd Pareto chart you show above with a lightly shaded area to represent the total portfolio and colored bars inside this to represent the individual asset classes.

There are trade-offs. The advantage of a common baseline for comparisons (as well as space for labeling) should not *always* be enough to tip the decision to favor simple or cluster bar charts.

Sometimes, the advantage in compactness and in obviousness of an underlying part-to-whole relationship should tip the decision to favor stacked bar charts.

You could survey and chart {how many of your readers use pie charts, and how many of your readers use dot plots} in the same cluster of a bar chart. It wouldn’t be kosher to show those numbers in the same stack (or the same pie) because these categories overlap rather than add up towards your whole readership. Thus the stack immediately, even subconsciously, communicates more information than a cluster to an aware audience.

The lack of a common baseline reduces the effectiveness of the stacked bar chart. If the comparison of items is either very easy due to huge differences in magnitude, as in this example (power is upwards of 90% of the total), then a stacked chart illustrates that one group of items is much larger or smaller than other groups of items. I don’t think it conveys this any better than a clustered chart with the short items adjacent to the long items.

With such customizations, I agree that a simple or cluster bar chart can convey a part-to-whole relationship very effectively.

If we are limited to standard Excel charts that modestly skilled users can create without your Waterfall Chart utility, then I would ask:

Does a Stacked Bar Chart Beat a Simple Bar Chart
(When We Don’t Want To Resort To a Pie Chart)?

I vote yes – for this dataset. (Feel free to use my stacked bar chart as a good or bad example. )