I am a huge fan of coffee. Guaranteed, by the time you see me in the morning, I’ve had at least one cup already. This is how I get my day started and protect the rest of the world from tactless gibberish. The addicting scent of that hot bitter brew jolts my senses awake, and that is key. In addition to the caffeine kick, my senses are no longer cloudy and I’m ready to analyze some data.
That is, I think, one of the best parts of my job. But one of the biggest challenges after you’ve made sense of your data is how to share it with everyone else. What is the best way to tell the story? How much data do you share? What does that stakeholder care about? Bars or lines? 3-D or not? What color do you use in your chart? Cream and sugar? Hot or Iced? Wait.The last two are not about data.
The Question Is The Answer
I attended the Perceptual Edge conference this year to learn about two things: new or improved ways to analyze data and how to better present the data. The two things are not exclusive. It’s a simple and logical idea and certainly makes sense, but I think it’s easy to forget so it is worth reiterating every so often. During the “Now You See It” workshop, we began by discussing how important your eyes are to data analysis AND the presentation of it.
You have to use your eyes to find patterns, trends, discrepancies in the data and make connections. You might use a simple scatter plot to discern patterns or fourteen pivot tables to slice and dice your data. You have to use your eyes to assess your graphs to ensure they visually make sense to your audience and that the graph is saying the right thing. The ultimate goal of a chart is to communicate to your stakeholder the answer to the question they have asked in the most concise way possible. In practical terms, I often prefer to get the questions than a request for a certain set of data. To borrow a term from Avinash Kaushik, I can “puke out data” with the best of them, but it’s much more rewarding to dig into a mystery. With the second, you’re likely to find interesting, actionable insights. A data puke is just that.
Data Visualization Is The Cream to Your Coffee
Your visualizations should enhance your analysis and make it better. Your data means nothing if people cannot see the highlight of the data or do not understand it. Luckily, there are many options to showcase your analysis.
It’s easy to pick one method of data visualization over another because it’s prettier or fancier. Since everyone loves a pie chart, it’s a great opportunity to show off your Excel skills. It is also a great way to misrepresent your data and confuse your audience. As an example, let’s talk about my coffee habits. I want to show you the different types of coffee I drink.
Answer me this – which coffee had the highest consumption? I’ll give you a hint, it’s not Verona. Surprised? It will probably also surprise you that the blue and green slices represent the same number of cups. I’m not lying. If you want to mess with people, give them a pie chart and watch them try to use it for any real decision-making. Charts should not be a puzzle one must figure out.
The purpose of a chart is to display data in a way that highlights the pattern, trend or data point. Sometimes, it makes sense to just have a table. Often, it makes more sense to use a graph. However, the graph, and the table, should always be designed so that the audience can easily see the answer. We do all the hard work so our audience can go right from, “Which type of coffee had the highest consumption?” to “Puerto Rican” in just a glance.
The user should always be able to answer the question. You don’t need to get fancy with special effects and background noise. Unless you’re in marketing, in which case, all bets are off.
By the way, the same dataset below was used for both the pie chart and the column chart. The difference between the two charts above is significant in one major way: the column chart can actually answer questions.
|Coffee Type||# of Cups Consumed|
From Coffee to Cappuccino
It is not enough to decide to avoid 3D and pie charts like the plague. Go back to the question being asked. In the case above, the question was “Which coffee did I consume the most?” The answer is in the table above and you could just serve that. It is a perfectly fine table. The coffee types are listed in alphabetical order and you can scroll down to find the largest number. A better table would rank them from largest to smallest.
|Coffee Type||# of Cups Consumed|
Not only do you instantly see Puerto Rican coffee ranks the highest, you can also see the order in which they all fall. I haven’t altered the data in any way, except sort it by rank. Your chart should do the same. By doing that, you make it even easier for the reader to answer his question.
You’re still using milk and coffee to make a cappuccino, but the method of delivery makes all the difference.
The Perceptual Edge workshop did not just go on and on about pie charts and 3D effects. Stephen Few led the workshop in multiple discussions on different ways to make charts better for stakeholders. The general consensus was the default settings are not good enough.
We ended up with a list of guidelines for making charts. Here are, in my opinion, some of the best ones:
Let’s analyze my coffee drinking habits for this year. Let’s say I treat myself every so often but have not been keeping track of how often I go throughout the year. Obviously, we can’t use a pie chart.
That pie chart doesn’t tell you a thing except there are 12 months in a year. It hurts my eyes. Don’t do it. Just don’t. Ever. A time series should always, always be a column or line graph.
Don’t lie, for some of you, this is double or triple.
The column graph can tell you fairly specifically how many times I went to Starbucks each month. Now you can see, holy frappuccino! May was a big month. You can also sort of see the pattern here. It looks like, on average, I go to Starbucks about twice a week. So now we want to know what happened in May.
Aha, it’s much easier to see the trend. So while it’s obvious that something different happened in May, there seems to be something toward the end of the year too. It’s not quite cyclical but there is something to investigate. Maybe we should look further.
Clearly I need a coffee support group.
There’s definitely a pattern here. The point is to remember that there are multiple ways to tell a story. Any one of them might be true. The story your graph tells should answer the questions being asked as concisely as possible. The answers to our questions, though, is Half Price Frappuccinos promotion in May and Pumpkin Spice Lattes showing up on the menus in September.
Regardless of the visualization you choose, it should be able to answer your stakeholders’ questions. And, remember, pie charts are evil.