I think they're useful if and only if the audience expects box plots -- if it's just the convention of their field, for instance.
To me, the biggest problem most data visualizations run into is that they're trying to convey way too much information all at once.
I learned in my consulting career to communicate one idea at a time. When I need to communicate a few ideas about data, I'll usually assign each idea exactly one visualization.
And for that reason, box plots just aren't that useful. For me, it's not that they're intrinsically confusing -- although to some audiences they are -- but instead that they'll invariably leave too much work to the audience.
If you want to communicate that medians in group A are higher than groups B-G, then you should probably just plot the medians on a vanilla bar chart. Maybe the box plot can sit in an appendix for unusually curious readers.
To me, the biggest problem most data visualizations run into is that they're trying to convey way too much information all at once.
I learned in my consulting career to communicate one idea at a time. When I need to communicate a few ideas about data, I'll usually assign each idea exactly one visualization.
And for that reason, box plots just aren't that useful. For me, it's not that they're intrinsically confusing -- although to some audiences they are -- but instead that they'll invariably leave too much work to the audience.
If you want to communicate that medians in group A are higher than groups B-G, then you should probably just plot the medians on a vanilla bar chart. Maybe the box plot can sit in an appendix for unusually curious readers.