For a recent project, the client was using the presentation layout below and asked me to help them improve it using Excel. Here’s the pivot table and pivot chart they shared with me as a starting point before I got involved:
Here, I share some patterns that I used for this project and plan to use again.
Prototype in grayscale
Color with the client at the end. This helps minimize color theory churn, and sometimes avoids it entirely (sometimes gray is ok). In this project we actually just used gray and red, which was simple and effective for communicating the stories we wanted to tell.
When color is necessary, use a color palette that fits the domain of the data or the business
I usually use a monochrome scale of muted gray, blue, black, or red and try to use the Standard palette when exploring choices.
For a clustered bar chart with narrow bars, in general avoid data labels at the end of the bars because they can overlap and cause confusion. Instead, try major gridlines, with the numbers on the y-axis. For example, here are gridlines that are a light gray from the standard colors, which is subtle but still visible:
Don’t ask the client where everything should go
It’s good to ask internally, but show the client a design and don’t ask them the best way to present the data. think about why you’re choosing the format you are, prepare your argument mentally in case you are challenged, and then move forward.
One of my favorite metaphors for this is building a house. Do contractors ask the client if they want the fridge in the kitchen? Or the toilet in the bathroom? As my previous manager used to say, don’t ask the client if they want a toilet over the fireplace--they might say yes and they’ll get burnt in the long run. The point is to present a solution to the client and be confident it is appropriate. Do your homework first so you can defend your design and convince others if necessary.
Consider the size of chunks--how many bars? how many lines? how many multiples? Do you really need 20 bars in your bar chart, or is it more clear if you only show 5 and lump the rest together as “Others”? A research paper by George A. Miller in 1956 (and still cited today) suggests a solution, as suggested by it’s title: "The Magical Number Seven, Plus or Minus Two: Some Limits on Our Capacity for Processing Information.” You can probably remember this title--it is only 8 words!
Watch out for god charts that try to do everything
Instead, consider several smaller pieces to do one thing and do it well.
Do not use line graphs for nominal data (person, region, etc).
Think of nominal as an arbitrary label that should not be connected to other labels with a line, but rather stand on its own.
Instead of gauges, radar charts, spider charts, pie charts, donut charts, circle charts, use bar charts.
Watch out for funnel charts: funnel is a metaphor, not a chart layout. Start with a bar chart and go from there.
Use line charts instead of area graphs, and clustered side-by-side bars instead of stacked bars. Stacked graphs cause misinterpretation because they do not start at zero, but rather inherit their position from the values below them. They have similar weaknesses to pie charts. Stephen Few has a write-up on this here:
Hide or delete unnecessary data ink and objects: numbers on axes, filters, legends, labels, gridlines.
Minimize the ink on a chart, and limit information to one place to reduce confusion for users, make maintenance easy, and simplify the visual design.
Considering the above, here is my draft for the chart presented at the beginning.
As I look at this now, I’m reminded of the “Magical Number Seven” guideline, which makes me want to group some of the bars together into an “Others” category later! This is not perfect, but it is a step towards internal consistency, simplicity, and allows cross analysis through 3 different presentations. The client was happy with the new layout. Right away, they had new insights due to the more accessible visual layout, and they started asking themselves new questions about how they can push their business forward, which is what it’s all about.
Posted by admin