Good Charts and Good Charts Workbook

These two books are antidotes to bad dataviz. They're for anyone who needs to use data to make a business case, to win people round to a course of action, or just to make simple data as intelligible as it should be. In short, they're for anyone who has to present and explain data, and show why the ideas behind the data matter. They'll help you make better presentations and understand what you need to know in order to make them better.

Scott Berinato is a senior editor at Harvard Business Review. He's a self-confessed dataviz geek, and talks widely on the subject in his work with business organizations.

So, keep listening to hear about the four kinds of dataviz, the three stages to making really good dataviz, and how we don't read visualpresentations the same way we read words.

We'll start by looking at "Good Charts" and then move on to the workbook that goes with it. The bulk of "Good Charts"is divided into four parts, with a strong emphasis on the practice of dataviz. The title of each part tells you exactly what to expect – Understand, Create, Refine, and finally Present and Practice . This is a book with a clear how-to message, setting out to demystify dataviz from the start.

Dataviz is essential in a modern, knowledge-based economy. Cell phone apps deliver information on everything from traffic flow to stock prices. Sports broadcasts overlay game action with statistical analysis in real time. You can watch weather fronts roll in, or monitor your household's energy use, all with a few keystrokes. It's all dataviz.

In the workplace, the skillful use of dataviz shows that you're on top of your area of knowledge, and have a firm grasp of the facts. It also gives you the power to influence others. As you'll hear, using this power ethically is as important as using it effectively.

So, what makes good dataviz? Why is one chart more informative than another showing the same information? Berinato quickly rejects the idea that good dataviz is a question of execution, or any formal rules of presentation. It isn't. It's about setting a context for understanding, driving discussion and engaging your audience. Any visualization which doesn't do that has failed, no matter how technically accomplished it is.

Part One of "Good Charts," Understand, consists of two context-setting chapters on the history and science of dataviz. There are some fascinating insights here, particularly in the chapter on the science of perception.

We don't see data visualization in the same way we see writing. Our eyes don't move across and down the page in a systematic manner. We see visually striking cues, and seek context to link them and give them meaning. We can only process a small amount of information quickly, and we rely on conventions to help us. For example, time on a graph conventionally moves from left to right. Using red indicates that you're giving a warning, while green is calming and positive, and so on. These visual cues are deeply embedded in the way we view visual information, anddataviz presenters should work with them, not against them.

Creating effective dataviz needs careful planning. First, you need to decide whether the message you're trying to convey is conceptual or data-driven.

Conceptual information is qualitative. It concerns processes, organizations and ideas. Data-driven information is quantitative. It involves measurable items like revenue, customer signups, and headcount, for example.

Second, you need to decide whether your dataviz will inform your audience of facts, or invite them to explore new ideas.

Based on these two decisions, you'll be able to classify your presentation into one of four types. If it's conceptual, it may be conceptual and informative, illustrating an idea, or conceptual and exploratory, generating new ideas. Data-driven presentations are also either exploratory, testing or confirming a theory, or informative, expressing quantifiable ideas and information. Most dataviz used in presentations falls into this fourth category, which Berinato calls everyday dataviz.

The next step is to create, armed with a clear idea of the information or ideas you're going to communicate. The key to effective creation is in the preparation. You need clear space, both mental and physical, and you need to think and talk through what you're going to deliver.

Here, the book moves to being seriously hands-on. First, you need to consider the data you have to present, and identify keywords you use when talking about it. These suggest the final form of your dataviz. Keywords involving comparison will likely result in line graphs or bar charts, for example.

Next, you sketch out an outline of the dataviz, working quickly and trying as many different approaches as you think will work. Finally, you'll have a working prototype which you can refine.

The author calls the core chapter on creation Better Charts in a Couple of Hours, and the emphasis is on getting something down quickly, without overthinking the problem. He follows that up with a chapter on refinement. Here, the most important thing is to show information in a way that persuades your audience. You need to make the key information stand out, and remove anything that distracts from it. After all, the most effective dataviz is often the simplest: complex, multilayered graphics require much more effort to understand.

The book contains a valuable and thought-provoking chapter on the ethics of dataviz. The question is simple: when does persuasive dataviz shade into manipulation? You can prove anything with statistics, as the old cliché goes, and some users may be tempted to use dataviz in unethical ways. Clever visualization can make relatively unimportant data stand out, and hide key information. For example, if you change the scale on the vertical axis of a graph, you can make growth rates look dramatically steeper.

To check that you're using information ethically, Berinato suggests you question every piece of dataviz you produce. Are you actively changing the argument your data presents? Have you eliminated or hidden something that could call your argument into question? Perhaps most importantly, how would you feel if someone else presented you with a chart like yours? These are worthwhile questions for anyone who's enthusiastic about their data, but not absolutely sure of their argument.

Berinato offers tips on how you can present charts with maximum persuasive effect, while keeping them ethically sound. You can do this through the style of the presentation, for example by emphasizing the ideas shown in your graphics, not the graphics themselves. Or, you can use dataviz to tell stories, or even to subvert your audience's expectations. Showing them a chart that confirms their biases, before revealing a more accurate one which challenges them, for example, can have a big impact.

In closing, Berinato discusses how to criticize and evaluate data visualizations. This is partly to help you hone your critical skills, but also to share tips on what to do in your own presentations. And, of course, what not to do.

The book is full of real-world examples of the importance of good dataviz. Each chapter contains a case study, drawing on the experience of dataviz users and creators. These tie in with the particular concerns of the chapters in which they appear. The book also offers examples of all the major kinds of graphics you can use, and the situations in which you might want to use them.

One thing the book does not discuss in detail is the various software tools that are available to help build dataviz. This is partly because there are so many, and they all perform different functions, and have different strengths and weaknesses. But more importantly, it's to keep the focus on the concepts behind good dataviz. These don't depend on how good your software is, but on how sound your data handling, planning and creative skills are.

So, where does the workbook fit in to the Good Charts project? Well, it's very much a follow-up to the original book, having been published three years later. It's also much more of a how-to guide. As such, it covers a wide range of techniques that work, and warns against ones that don't. While the first book outlines a conceptual framework for producing good dataviz, and walks you through the theory, the workbook allows you to get your hands dirty by working through actual examples of dataviz projects.

The workbook divides into two main parts. The first covers the key skills you'll need to develop in order to make those good charts. The second runs through two dataviz projects in start-to-finish detail.

Each chapter in Part One introduces an important dataviz skill. These cover handling color, presenting information clearly, choosing the right kind of chart, creating persuasive graphics, and making convincing conceptual charts.

The chapters all have the same basic structure. There's an introduction to the skill area, followed in each case by six important principles. These often read like tips, but don't skip over them – they're valuable, and you'll need them.

Next, there's a warm-up exercise consisting of 10 questions. These test what you've read about in the six principles, and introduce you to the choices you'll need to deal with when building dataviz. They're followed by a discussion section. This is not just a list of answers to the questions, but explains in detail why the answers are right.

The chapters each conclude with three tasks involving a series of judgment calls. These are larger-scale challenges than the warm-ups, and test your knowledge of the six principles with some rigor. Each task has its own discussion section, and these really delve into the reasons why you might choose a particular color or chart type.

The repetitive structure of the chapters in Part One works well. In each one, the book tells you exactly what you need to know, and then follows it up with tests of uniform length but escalating difficulty. These offer genuine challenges to your understanding of the six principles. And the skills you master in Part One are stretched still further in Part Two.

Part Two begins by recapping the three key stages of putting together a piece of dataviz. First, you talk it through. Then you sketch it out. Then you prototype it. It's not quite that simple, perhaps, but the sequence of the steps is clear and easy to understand. Berinato doesn't assume readers have read "Good Charts" already, which means the workbook can stand alone. Nevertheless, it will certainly help to have read the theory before you start the workbook.

The practical part of the workbook concludes with two major exercises for the reader to work on, producing their own pieces of dataviz. One is a monthly report, and the other a presentation on plastic waste in the South Pacific. The first is intended to inform as efficiently as possible. The second is a call to action. The report draws cleverly on a series of imagined questions and answers between the author and the marketing manager requesting help with the report. This allows you to practice the essential skill of drawing out important information, by identifying keywords. It also gives you a good idea of the right sorts of questions to ask.

The second exercise tests technical and illustrative skill, and also your imagination. The discussion emphasizes that there's no one good way to carry out this task, and introduces a range of ways to approach the same data. Just working through the possibilities will likely leave you choosing between two or three different approaches. As Berinato states, that's no bad thing. It's more stimulating to have data that gives you real scope for expression than material which leads you to a predictable, boring piece of dataviz.

The "Good Charts" books make for engrossing reading, particularly if you're enthusiastic about data and how to present it. Even if you're a reluctant presenter, there's plenty here to help you improve your visual aids. Both books are fun to read, too. The author has an easy, conversational style, and a nice line in humor. Bad dataviz is called out for what it is, and both books offer some excellent examples of it.

If we have a criticism, it's in the physical makeup of the books. There's a good balance between illustration and white space for ease of reading, but not a great deal of space to sketch out ideas, which you might expect in the workbook, especially.

But that's a minor drawback. Above all, Berinato makes a subject that could have been dry and dull come alive, both visually and as a mental workout. And that's no mean feat.

"Good Charts" and "Good Charts Workbook" by Scott Berinato are published by Harvard Business Review Press.

That's the end of this episode of Book Insights. Thanks for listening.