I would move the legend to the top. Some data viz best practice research shows that helps people process since they will read top-to-bottom but with the legend at the bottom they have to jump down to read it and then go back up to the top of the graph to start to interpret it.

A couple edits I'd make to this graph (partly to make it more applicable to common scenarios at OMNI and partly to adjust formatting): * instead of having the n's in the labels I'd have the category. So it would be like: Completed Follow-ups 28% and Incomplete Follow-ups 72%. * Adjust colors of the labels so they match the segments they correspond to. * Then I'd adjust the yellow bar to be a target since that's more applicable to our work except the very specific instance of GPRA tracking and might help people conceptualize when to use it. Change label to Follow-up Target 70%

The only exception to this I'd make is that I'd typically move unknown/not answered groups to the end. I don't know that there is a way for you to build that into the code, since those groups can vary a bit in name/context. Just thought I'd mention in case you can think of a way to do that, or maybe we just add some contextual info on that to the guide so people are prompted to do it themselves (and know how) if it comes up.

In the sample image, there is an extra 0 at the bottom - maybe a lingering axis label?

I don't think we have a preference for graphs where the bars are horizontal vs. vertical, we just want text to be horizontal for ease of reading. So if you have a graph with vertical bar charts, you just want to avoid what happens in MS Office where they will angle the text if the x-axis is too tight. In those cases in Office, I'd give the guidance to make the graph wider or shorten the labels so everything fits horizontally. Of course, as you note here, the graphs with horizontal bars help avoid the spacing issue but no need to convert to horizontal bars if things are working well vertically.