The Cost of Bad Design

The next time someone asks me how good data visualization actually contributes to a better bottom line, I’m going to retell this story.

Some time ago, I sat in a meeting with 11 other people. We were reviewing evaluation findings, presented via charts, which were created by another researcher (not present). You know this scene. You’ve been there a hundred times. The group spent 20 minutes just trying to dissect a single chart. It wasn’t the data that was confusing. It wasn’t the chart type, either. But it was the little things like color and labeling that confused the 11 of us (an educated group, too, I should point out). So what did that bad design cost?

6 people are paid $600/day, which means we spent $150 on their confusion
1 person at $400/day = $17
1 person at $300/day =  $13
1 person at $250/day = $10
2 people at $1,000/day = $83

So one poor chart design cost our group $273, which is actually more than the daily rate (salary and benefits) for one of the meeting attendees.

Ouch. And this doesn’t include the time it took for the report author to develop the poor chart in the first place, or the time that person will put in to make the chart clearer. Bad design is expensive. Thus, investment in some professional development around good design or even consultation with an actual graphic design may literally pay off in the end.

Proper Placement of Chart Legends

In my dissertation study (probably the most boring four words to start any blog post), I saw a lot of evaluation reports that grouped all tables and graphs into the appendix. Tables, graphs, and other graphics really must be placed right next to the narrative describing them. Why? When we flip back and forth between pages, we impair working memory’s ability to make sense of the associated words and images. Truly, the ideal situation for our brains is extremely close placement. Whenever we have to seek-and-find to match up content, we hurt cognition. That’s why I was super excited to see Jeff Johnson talking about charts and their legends in his book, Designing with the Mind in Mind.

Jeff also pointed out how it is hard for people to distinguish legend colors when produced in default mode (so two lessons here). This is what the typical chart looks like in Excel, with some modifications I made to clean it up:

The blues and purples do get hard to distinguish, don’t they? So Jeff recommends enlarging the legend colors so they are both more easily distinguished and closer to the actual lines they are associated with, reducing the need for an actual eye movement between the line and it’s corresponding legend entry. You can’t really do this in Excel, though. So I faked it, by inserting a square, matching the color, and putting it right over the line in the original legend. See here:

It’s an improvement, yes. However, there’s still a lot of searching to match up each line to it’s legend entry. So I got this idea from Storytelling with Data to get rid of the legend entry altogether. In the example below, I’ve inserted text boxes with the correct word from the legend, so it is totally obvious which line goes with what and the need for the search-and-find in the legend is removed. Cognition supported. Oh yeah.