Displaying More Complex Data
I'm going to take a brief break in the sacred cow series to address something that came up in my 9-to-5. It was one of those situations where it was clear there was a need for a better solution, but it wasn't obvious exactly what that solution was. This did not involve presentation slides, but it is no less relevant to similar situations that happen to involve slides.
I was asked to review a report on an evaluation of some solutions against a list of criteria. At the end was a chart much like this one to the right. (The data in this and all following examples are not the same as what I encountered (for hopefully obvious reasons), but should be similar enough in concept to be illustrative.) As you can see, even when you click on the chart to see the full-sized version, it's unreadable. The story that the chart is trying to tell isn't at all obvious and it takes far too much effort just to read it, data point by data point. There had to be a better way.
Remembering what I learned from Edward Tufte, I suggested that perhaps this is a situation where a data table would actually be more appropriate. However, I also added that we might want to use some color to help convey the story rather than make the reader make heads or tails of 64 numbers that look similar enough to visually blend in together.
What I got back was something like this. By this time, we didn't have time to continue to tackle the problem and it was a valiant effort on their part, but this too had its problems. On the positive side, there was good color contrast to visually separate the values qualitatively. However, what hurt this design was how the text was unreadable in some colors. Changing the text colors in the red and green cells only probably would not have helped the readability either. There was also the issue, in my mind, of too many qualitative distinctions. The scale wasn't logical, but adjusting the scale with the four qualities intact didn't improve things either.
Since there was no more time to make modifications, this is what we were left with. What I did do, though, was to tackle this problem as an academic exercise.
The first changes I made were to change the number of quality distinctions from four (Great, Good, OK, Bad) to three (Good, Acceptable, Poor) and try to make the text more readable. To make the numbers easier to read, I got rid of the colored cell backgrounds and instead applied the color to the text. To keep the reader from getting lost in the data, I used alternating shading for the rows, a subtle, but effective technique. What I had now was better, but still could use improvement.
The next change I made was simple, but yet yielded a much easier to read data table. I split the categories up into groups of four. This could represent groupings of categories or it could merely be a visual break; a chunking of data, if you will. Breaking data up into manageable chunks makes it easier for the reader to consume and remember the data you're providing.
Now, this seemed to be a workable solution. I then asked myself what other ways I could realize the same or better effect. At this moment, I remembered the well known and higly effective tables that are found throughout Consumer Reports magazines. Sometimes they contain data and sometimes they just contain symbols. These symbols could take the place of the colored text to provide a more meaningful table. This next version contained my variation of this technique. I created symbols similar in concept to those in the famous magazine and placed them along with the data, which are now in a simple black typeface.
My next thought was, "What if the raw data is not necessary for this given audience? What if the high-level meaning or importance of the values needs to be conveyed and not the values themselves? What we have now is a table with just the symbols and no data, save for the total values at the bottom. Now, instead of giving the reader a whole lot to read and consider, we're giving them what's important to them: the simplified significance of the values.
You can, of course, apply the chunking of the data to these symbols as well. These symbols are information to be consumed just as the raw data was.
The lesson from this whole exercise was that you should always consider and reconsider how you display your data. Consider an approach that is different than what you would normally take. Put yourself in your reader's or audience's shoes and don't stop improving it until it has the effectiveness that you would expect it to have.
