by Scott Berinato

Good charts matter. Business runs on data, and data is overflowing. 

The insights, opportunities, risks and trends that make up the trajectory of your fortunes are all buried in those millions of spreadsheet cells you manage. You need to be able to spot the right data points and the ones that connect to each other in meaningful ways. You need to talk about them, argue over what they mean, and make decisions with them. You can’t act on any of this if you don’t see it.

Easy, you say. Just click that button in your spreadsheet or analytics software. Out pops a chart. It looks pretty good. Maybe you’ll tweak the colours. You can even put it on a dashboard. 

Job done? No. Job started. 

Software can quickly create charts, but it can’t quickly create good charts. That requires you. Think of what’s spat out of the program as your first step, not your last. From there, you need to refine that chart until it clearly, urgently conveys the most salient information to the intended audience. The software visualises data. You visualise ideas.

It’s at this point that I typically start to feel resistance from some audiences when I speak on this topic. Those click-and-viz software programs are easy and convenient and some people don’t see the value of spending the extra time refining something they believe is good enough.

In particular, two groups often look askance at my contextual, design-thinking approach to better dataviz: analytics and finance. 

I understand this. Typically these two groups are asked to report their data – all of it – dispassionately. Any refinement may feel like less-objective editorialising, or manipulation. But I will argue that all charts are manipulations – just one view into a dataset in which we’re choosing boundaries, parameters, variables, highlights, and background information. 

These are all judgments (often good ones) about what’s important and valuable. You’ve made hundreds of decisions, often without realising it, to arrive at your chart. 

It’s important to manipulate your charts so they convey the most important information clearly, with honesty and transparency. This article will help you do that. 

Using some top-level ideas, derived from my book Good Charts, we’ll focus on a few key ways to improve your charts from the automated output many of you rely on. None of these tips entails major time commitments or skills development – though both of those will certainly be worth the investment for anyone who wants to get better at making data visualisations. Instead, here we focus on a few, simple, easy-to-implement ideas to get better right away. Off we go.

Talk about your data

Visualisation software is good and getting better all the time. But software programs still can’t know your audience. 

Is the chart for one person looking at it on paper? For a room of 1,000 people on a giant screen? Is the audience full of experts? Is it the end of the day when they’ll be tired?

Software also doesn’t understand your goals. Do you want to inform? Persuade? Are you trying to secure funding? Help someone make a decision? In short, software can’t intuit your context. And in visualisation, setting your context is the most important factor in getting to a good chart.

I’ve seen thousands of charts that look great and don’t communicate the right message, or any message at all. They’re just pretty cathedrals built on sand. Conversely, I’ve seen reasonably ugly charts that show me what I need to know with immediacy. They’re shacks on bedrock. I’d always take the latter over the former.

So how do you set your context? Paradoxically, start by putting away your data and your computer. This helps free your mind from feeling chained to the dataset. Then find a friend to chat with. 

Start by addressing the fundamental questions of what am I trying to say, to whom, and where?

Try to avoid talking about data. Just talk about what you’re trying to show. Instead of focusing on how “the data shows a severe downward trend in revenue since we launched the new products,” talk about how “I see a connection between the product launch and lower revenues and if we don’t adjust the product line, the trend may get worse”. 

Note the subtle semantic difference between the first statement, which focuses on what the data says and the second which focuses on what the idea is.

Also, make sure your friends asks you “why?” constantly. This will force you to explain even basic assumptions clearly and may expose gaps in what you know and what your audience knows. 

As you talk, listen for visual words that may suggest the type of chart you want to use to communicate your idea. If you hear yourself saying “the distribution of results was really random,” that tells you something about how to visualise, possibly with a histogram. Or if you say “revenue sources were clustered in a few regions” that suggests a map, or possibly a scatterplot.

Once you get used to doing this, you’ll be surprised at how many visual words are embedded in a ten-minute conversation. Ultimately, you hope to reach a point where you can summarise what you’re trying to say in a sentence or two. Keep that sentence handy to help guide your chart making. 

Note Figure 1 below as an example.

Before

After