Chips & Salsa Charts

There are more kinds of charts out there than any one person can keep track of. Robert L. Harris' encyclopedic Information Graphics: A Comprehensive Illustrated Reference discusses about 1,600 types of charts, ranging from the “abstract graph” to the “zigzag chart.” Tufte's tomes showcase roughly 1,800 visuals, and even Microsoft Excel 2005 offers 99 “standard” and “custom” chart types right out of the box. It's no wonder that the best-intentioned business information visualist gets stumped from time to time. God knows, I do.

I'm a fundamentalist, though, and that helps. Blessed with a short attention span and an even shorter memory, I find myself driven to look for the fundamental components of everything. So over my years of creating more charts than I can recall, the need for a “visual fundamental” has been gnawing at me, and I think I've finally got it nailed. Here's the elevator pitch:

Part I: While there may be a nearly infinite number of ways to visually present data, the truly important differences between chart types have less to do with what you wish to show than to whom you wish to show it.

In other words, creating the perfect chart for a given set of data is useless if your audience can't or won't take the time to read it. You can lead a client to water, but you can't make her drink.

But wait, there's more; this elevator ride isn't over yet:

Part II: From an audience perspective, there are three fundamental types of information visuals: 1) visuals that make a simple point; 2) visuals that support a complex point; and 3) visuals that help the viewer discover her own point.

One more floor to go:

Part III: One good way to think about these three types is to compare them to the types of food you might prepare and serve to guests: 1) chips and salsa; 2) BLT with avocado; 3) Boeuf Bourguignon (French beef stew cooked all day in red wine) and a fine glass of Cabernet.

Bing. Doors open. Lunch?

Iron chef

Think about your presentation like a meal that you're planning. Guests will come, hoping to be entertained and to have at least a little something tasty to eat. These guests could be your friends, your co-workers or total strangers; there could be just a handful or an audience of thousands. Whoever attends, they're looking to you to do the cooking. Your ideas are your ingredients, your outline is the menu, and your charts are the dishes themselves. The good news is that for most of the entertaining you'll ever need to do, you've only got three meal approaches to worry about: snacks, a light lunch or a sit-down dinner.

The three are distinguished by their level of complexity and the corresponding demands you put on your audiences to digest them. You will choose which approach to use depending on the level of interaction you hope to have with your audience—and whether you want to initiate discussion, move the conversation along, or close things down.

Data dump

In the spirit of the scientific method, I knew that my hypothesis demanded rigorous testing even though it had been simmering in my mind for quite a while. First thing I'd need: data. Problem: clients don't want me to share their customer data without prior written consent, and I was pretty sure they wouldn't give consent if I told them I was going to compare their customers to stew. Solution: collect my own data.

Luckily, in my basement I have a recently purchased data-collection device, otherwise known as an exercise bicycle, a Diamondback RS400. My data-collection plan was highly sophisticated: each day I would ride the thing for a fixed period of 30 minutes with the resistance setting on “Random” (“Low” felt like a cop-out; “High” seemed a stretch). I'd note the number of miles covered. Not only would I generate a whole bunch of time-series data, I'd get the added benefit of burning off some of my wife's French beef stew. Nice plan.

It had been a while since I've undertaken any kind of serious exercise program, so after a few days all I had were really sore legs and a couple of sad data points. This was going to be a long road; I'd need something to distract me from the daily pain. Serendipitously, a friend moving abroad had forwarded his mail to our address, so we weekly received a selection of new magazines. I now had loads of fresh reading material, and started passing my data-collection sessions enjoying either the dry intellectual rigor of The Economist or the mindless but oh-so-entertaining gossip of People Weekly.

After five months of testing, results emerged.

The results, part I: Chips & Salsa charts

Visuals that get lots of information digested quickly

If you want your audience to “see” data and process it immediately, use simple charts that replace numbers and words with basic shapes. Chips & salsa charts are those that present one or two data dimensions in a rapidly scanned and quickly digested format. Pie charts, bar charts, time series, scatter plots and Venn diagrams are among the most common varieties. The graph to the right (Fig 3) shows how many miles I pedaled each day and is a good example of a chart that simply makes a point. In this case, the chart shows that I started in June at about seven miles per day, accelerated to about ten by late July, suddenly jumped up to just over twelve in September, and then remained at that distance until starting to fall off in December. (The longest distance I achieved—just over thirteen miles—was on the day after Halloween, which begs the question: how much sugar did I have in my system?)

This chart also illustrates the strength and weakness of a Chips & Salsa chart. Their greatest strength is that they don't require interpretation, their purpose being to accelerate the viewer's grasp of the core data—kind of like getting your guests fed with a minimum of fuss. When presenting a Chips & Salsa visual, you normally show it to your audience for a moment, state the finding and move on–often to another Chips & Salsa visual that supports a complimentary point. In this way, your story builds over time and you can rest assured that your audience is following along, crunch by crunch.

However, this immediacy is also the charts' greatest limitation: because they require little interpretation on the part of the viewer (deep analysis is up to you the presenter), there is little in the chart itself to prompt deep insight. Remember, after you've been to a party where they only served chips and salsa, you might feel full, but you haven't really eaten.

The results, part II: BLT with avocado

Visuals where the whole exceeds the sum of the parts

If your goal is to get your audience to develop some of their own insights, you will need to layer several complementary data dimensions onto the same chart. This is where the BLT with avocado chart excels—more complex than an appetizer, yet still composed of simple individual ingredients. A BLT with avocado chart doesn't just make a point visually—it provides enough information to support one or more complex points.

In this case, several new pieces of information enter the picture (Fig 4): by showing the days I did not workout, we see that a couple days' rest can improve distance, but too much rest gradually degrades performance. By overlaying the number of calories I burned (starting August 31, the day I finally realized the machine recorded this data point as well), we see that perhaps I have not been on the same consistent path of improvement all the time. Since I had the bike on the “random” difficulty setting, the distance I covered is not as good an indicator of total exercise as it might be. If the machine happened to give me an easier workout on Monday, for example, I might well cover more miles than on Tuesday's harder setting without actually getting any more exercise. Showing calories comes closer to the truth of how hard I really worked. Lastly, by adding in the average speed I maintained each day, we can well see the rate of improvement (and decay, too, unfortunately).

Such a BLT with avocado chart may require a headline to provide a summary or guidance on what to look for, but all the data should be there, ensuring that someone willing to really look at the chart will see the story for himself. While this example adds layers onto a simple time-series, several other types of frameworks (also typically used as BLTs) are also appropriate for supporting this kind of layered visual, including bubble charts, Gantt charts, flow charts and quad charts.

Because of this layering, some of these charts' immediate “scannability” is lost in exchange for deeper insight, requiring more time and effort on the part of your viewer. The beauty of a BLT with avocado chart is that when well prepared, it visually replaces many pages of detailed documentation, and does so without diminishing or oversimplifying the validity of the raw data. Like a good light lunch, these kinds of visuals provide the insight and energy necessary to move on and take action.

The results, part III: Boeuf Bourguignon and a fine glass of Cabernet

Visuals that require time and appetite, but are well worth the commitment

Preparing Beef Burgundy is a full-day affair. It begins with simmering chopped onion and garlic and does not end until the beef has stewed in gently boiling red wine for at least six hours. By the time the masterpiece is ready, each ingredient has altered in form and combined with all others to create a taste that is the culinary definition of synergy. Eating French Beef Burgundy also takes time: this is a meal that demands silver and china on an heirloom tablecloth, several sizes of cut-glass crystal, and a lively group of guests ready to eat.

Presenting an equally complex chart requires much the same effort and commitment. Enter The Economist and People Weekly. By layering a record of my reading material onto my distance and calorie information, a new and potentially even more interesting story emerges (Fig 5). The chart seems to indicate that I did better on the bike when I was reading the mentally taxing stuff (that would be The Economist, by the way) than when I was devouring gossip. Imagine that. I had no idea there might be a connection between cognitive focus and ability to concentrate one's energy. Who knew?

There is an old saying in statistics to never confuse correlation with causation, but in this case I'm going to suggest (since the only difference in my exercise regime was reading material) that reading stupid stuff appears to make me physically weaker. If that sounds like a value judgment, it is, but one I will support based on my data. The chart provides enough information and insight to discover a point. And that's the real beauty of a Boeuf Bourguignon chart: by layering many data points together (many of which may not usually be compared at all) a completely new idea becomes visible. Like the meal itself, I wouldn't often recommend such a chart for the typical business presentation. But when you know you've got a great point that your audience can discover for themselves, the effort is worth the time. Amen, and please pass the salt.

The French connection

It's old hat to anyone who has had the pleasure of attending the Edward Tufte show that the greatest information graphic of all time is Charles Joseph Minard's depiction of Napoleon's 1815 “March on Moscow.” To me, this multi-layered masterpiece is the ideal example of a Boeuf Bourguignon chart. It looks good to begin with, and the longer you have to spend with it the better it gets, and the more you can digest. Not surprisingly, Minard was French; from Dijon, in fact, just down the river from Burgundy. I wonder what he would consider an ideal appetizer visual? Now that I'm nearly stuffed with all these visuals, there's just one more thing... I can't help but imagine what dessert chart might look like.