# Is it OK to overlay a line chart overtop a bar chart to show averages in different categories?

I'm a software developer and not a data visualization expert by any means. I was sent a screenshot of how my client wants some data visualized.

Each bar is the average score for your team. The points in the line are supposed to represent the overall average of all users in the system so that a team leader can figure out if their team's score is above or below average.

The examples I find on the internet typically compare a single score with a straight horizontal line showing the overall average.

Is using a connected line that way fine?

Is there a better way this data could be visualized?

• There are multiple things that are....let's say suboptimal about this graphic. But #1 is probably that a line implies that points are meaningfully connected in some way (such as the same item measured over time or across some gradient). But average of all users in each category is not any more meaningfully connected than the bars. – mkt - Reinstate Monica Mar 12 '18 at 16:06
• You could plot bars side by side. You could plot points side by side. If the most important piece of information is the comparison, you could consider plotting the difference or ratio (as points or bars). – mkt - Reinstate Monica Mar 12 '18 at 16:08
• Some of the posted numbers on this plot must be incorrect, assuming $3.5 \ne 4.5$! – whuber Mar 12 '18 at 16:09
• Numbers are for sure wrong. It was just manually drawn in google slides. It seems like my gut was right and the line is not the right way for this to be shown. – Voziv Mar 12 '18 at 19:08

If your goal is to show

in Results your team is above average, but in Accountability you need improvement

then one way would be normalization, and showing percentage. That is, the team doing 130% of the mean on Results but just 50% of the mean on Accountability. Something like that:

The relation there will be team/mean - 1

Also, to see how data can be efficiently represented, check out the book The Visual Display of Quantitative Information

Beware, however, that this method compresses information. As @whuber noted in comments, you lose ability to compare variability of the data. Also, you lose comparative information: how much your company as a whole performs on different metrics ("you have better conflict than accountability")

• I am uncomfortable with this conversion to proportions because (1) it obscures the amount of variability to expect in each result and (2) doesn't fully reproduce all the information the original graphic is intended to present. – whuber Mar 12 '18 at 16:44
• @whuber i agree with you. My feeling, unsupported by OP though, is that such graph is more about giving general picture, not exact quantitative comparison. What do you mean by "variability"? OP's data doesn't show variability of overall data (sigma of "trust", for example) which would be useful. Maybe percentile is better way to present that? – aaaaa says reinstate Monica Mar 12 '18 at 16:57
• Variability (in the form of standard errors) often can be inferred from bar graphs, provided sufficient information about the response counts and scales is available. For instance, if the scale is integral from 1 through 7, then I know that a mean of 6.5 has a relatively small standard error and that a mean of 4 likely has a much larger SE. For bar charts that give counts or proportions of counts, I can actually compute the SEs from the values. Taking the ratios eliminates the possibility of such analyses of the graphics. – whuber Mar 12 '18 at 17:28

As suggested by @mkt the bars could be plotted side by side. I've taken that one step further and nested the bars.

@aaaaaa's answer mentioned "If your goal is to show". It seems to me having a clear goal is helpful in figuring out what plot to use.

I'll have to spend some time to clarify the goal but for the purposes of this answer I think this would sum it up:

Goal: To show company and team performance as well as compare the two side by side.

For those curious, here's a codepen showing how I made the above plot