Declutter Dataviz with Small Multiples

Declutter Dataviz with Small Multiples

Are you making graphs that look like this crap?

SmallMultiples1I won’t make you raise your hand. But let’s just agree not to do this, yeah? It’s SUPER hard to compare the tops of a bunch of bars. Two side by side bars per region ain’t so bad but beyond that and we are just being annoying.

This is change over time, so a more appropriate graph type would be a line graph, like this:


How’s this working for you? Personally, I can’t see a freakin thing. The values in the dataset are all so close together that this graph has basically become a bowl of spaghetti. Well now what?

When the graph gets complicated, one good solution is to break it apart into small multiples. Small multiples are several graphs, all on the same scale, arranged together. Listen dudes, these really gotta all be on the same scale, okay? This is what the same graph looks like, small multipled:SmallMultiples3

Holy smokes, now I can see stuff! Looking good, Region C!

The thing is, in the first two graphs, I couldn’t see the data well enough to be able to pull anything meaningful out of it. I couldn’t even come up with a good title because I couldn’t locate a point to highlight. The small multiples version makes things approximately a billion times easier.

It’s just six line graphs, guys. Super easy to make. Well, I shouldn’t oversimplify it. The tricky part is making sure all the graphs are the same size, including the first one on the left, which houses the axis. The plot area on that one still needs to be the SAME SIZE AS THE REST. How in the world do we make that happen? Not even kidding, I literally got my kid’s ruler and measured the gridlines behind the graph for Region A and then resized the graph for Region B until the gridlines were the same length. Then I looked up the size dimensions of Region B’s graph and typed those same numbers into the size boxes for the rest of the graphs for Regions C to F. Sounds more complicated than it is. Plus how often do you get to break out the ruler? Select all graphs by holding down the Shift key and then Align > Align Top. They’ll be all perfect and tidy and you’ll look like a freakin rockstar.



10 thoughts on “Declutter Dataviz with Small Multiples
  1. Brandie says:

    Perfect timing! This morning I just finished creating a PowerPoint showing how the way you present data can impact ease of interpretation, or even misinterpretation. What a great example of good(ish), better, best!

    Also, what an attention grabbing first statement!

  2. Nic Bookman says:

    To avoid the headache of matching the plot area size using a ruler (and yes, the fun of matching the plot area size using a ruler), couldn’t you retain the axis on all the graphs and simply change the font color of all but the left one to match the background color? Then you just have to make sure the ordering of front/back layering runs left to right.

  3. Carrie Dearnley says:

    Thanks for the idea.

    I also tried making this using a single graph and using a dummy set of continuous x data with blank values between each region’s values. This works pretty well however the distinction between each region’s data is negated slightly by the continuous horizontal grid lines.

    To combat this effect, I (manually) added light grey rectangles behind each region’s data, with white breaks between each rectangle.


  4. Rhonda says:

    Stephanie Green, you and your small multiples just saved a half page in my report and sent some not-so-great graphs to the compost pile. Thank you!

  5. Juan-Paulo (JP) says:

    Love it! This is a great example how to present a lot of data in a more understandable way!


  6. Laura Beals says:

    Great tips, Stephanie. I recently had a similar dilemma–I need to show change over time for several items that used the same axis. If plotted together, you couldn’t follow any one of them (i.e., not helpful at all). My solution was to insert sparklines into Excel next to the raw data, and then I changed the default vertical axis values of the sparklines to have a minimum of 0 and a maximum of 100 (as I was working with percentages) so that they could be compared. Then (this was a bit crazy)–I was able to take screenshots of the graphs and arrange them as needed in my report as pictures. This worked well, I think, as the overall slope of the line was the important takeaway, not the actual values of each data point (as the sparklines don’t allow you to easily see, for example if using your data above, that in 2009 Region A had a value of ~88%).

  7. Lynda Stinson says:

    This is amazing! I have been making line graphs that look just like the spaghetti bowl one. They would look worse, but I was able to convince them to minimize the number of lines. These simple tricks will do wonders to the reports!

  8. Nan Wehipeihana says:

    Thanks Stephanie. While I like to eat spaghetti Im definitely not a fan of trying to extract meaning from it. Cheers

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>

National newspaper publishers meeting at my hotel & wanted to #dataviz flash mob somehow

RT @disparityreport: Attended awesome workshop on data visualization yesterday! FF @evergreendata

The danger of real time dashboards & the need for slow data

Into R? Grab the code to make a few of the #dataviz on my latest book here Thanks @dckallos !

RT @nancyduarte: To be heard, a message must be authentic & audience-centric. Honored to have helped @SenWarren create this talk https://t.…

RT @VizWizBI: How Middle School Math Pays Off via @evergreendata

RT @SAGE_Methods: “‘What’s your point?’ [...] The most important question to ask when creating a data visualization” @evergreendata https:/…