Generally this approach is pretty effective. See the dashboard by Professor Hichert in Nice Dashboard Examples.

You said “Two or three clustered bars would be fine, but more than that would be too cluttered to be legible. You could rearrange the data to produce an alternate arrangement, but that may not be as useful for this particular data.”

What’s your opinion of using a variant of a panel chart ? I have done it was a small set of X access points on the horizontal comparing 4 different years using in-cell graphs. Seems to work well with my audience.

I frequently use combination charts in a similar way. Often I use an area chart for target values, or typical values, or some other kind of comparative measure, and a line chart for actual values. I don’t mix bars/columns with lines very much, unless it’s a construction like a box and whisker chart, or where I can use a floating bar to indicate a range of values and XY markers for individual measurements.

Thanks for the clarification. I’ve seen the attributes presented in the order I posted. and I imagine like any good methodology, there is some variation in its implementation.

It’s been a while since I’ve read the articles. I’m glad you cleared up the bar chart’s effectiveness in terms of position along an axis. That actually makes more sense.

You can also use this chart on 2 axis for Pareto Charts, where the bars show the Pareto data and the line shows the cumulative %.

1. Position along a common scale
2. Position along non-aligned scales
3. Length, direction, angle
4. Area
5. Volume, curvature
6. Shading, colour saturation

And in bar charts the primary visual attribute is position along a common scale, not length (which is no better than angle at encoding values)