Results over time, i.e. trends (ref 52)

This email gives numerous tips... it's analytically full-on. It's also a return to my roots: tables and graphs. So read on for ideas, twelve tips - plus some surprises on a dreadful graph.

Last week's email showed an awful graph from a report on Artificial Intelligence - see below. Don’t worry about words like Passives, Experimenters, etc – they’re described in the report. Of course, we should give graphs a decent lead-in title - and tables too (that's the first of the twelve tips), but don't worry about that either. Rather, study those downward sloping left-to-right curves?! Weird.

Maybe a table would be better (Figure 2). It abbreviates the row and column labels (e.g. Pa = Passive, for instance); that's to save space in this email, not because it’s a good idea…

Many people create tables that look like Figure 2. Don't. Figure 3 is better. It follows tips for decent tables:

Remove as many grids as you can, and those you keep, make them light, not heavy. Grids look ugly, visually dominate and give readers' eyes a black border to leap as they scan the table. You don’t box in your words in sentences, so don’t box in your numbers in tables.

Make compact: in the after, readers’ eyes travel less to take it all in. Everything is in a common eye line.

Remove distracting typography: in the after, gone is all that blue…such frippery might attract us to the table, but it then distracts us from the content. Also, in the after there’s no underlining nor any reversed font (that’s the white-on-blue column headings) – both hinder readability.

We now more easily spot pointless numbers – the ‘Disagree’ ones. They’re "100 minus the ‘Agree’ ones". Even the graph's authors realise the ‘Disagree’ numbers add nothing, for they omitted the 2017 ‘Disagree’ ones from their graph.

Figure 3 also follows more tips for good tables:

Put the most important column next to the labels – the ‘Total’ column. I wish UK newspapers would do this with football league tables – the most important column is ‘points to date’, and it’s on the far right, far away from the team names… to see how teams are doing, readers’ eyes must flick back and forth from far right to far left.

Put other columns in some order other than alphabetical. Here, they're in ascending order of the 'Agree' numbers (which - unsurprisingly - is descending order of the 'Disagree' numbers... I told you those numbers are redundant). Avoid alphabetical, for it leaves big and small numbers mixed up, competing equally for readers’ attention.

Now the surprise. Look back at the original bad graph. Yes, it’s bad, but as a table, it’s got some neat bits. The ‘Overall’ column is next to the labels. Tick. The other columns aren't in alphabetical order, but in descending order of ‘Disagree’ – 61%, 33%, 31%, 15%. Another tick. From the jaws of defeat, the graph finds some success. 

Back to the redo. Study Figure 4. Gone are the redundant ‘Disagree’ numbers. Also, it follows another tip: put comparable numbers in columns - we find it easier to scan down columns of numbers than across rows. ‘Urgent’ numbers are in a column - it's easy to scan down and compare everyone's scores.  

A ‘difference’ column might help, it spares readers from mental arithmetic. Currently, to see who's up or down from one year to the next, we must deduct numbers in one column from numbers in another. If readers want exact numbers, then yes, a ‘difference’ column helps.

Maybe they don’t. Maybe they just seek a ‘feel’ - is stuff up or down or the same? A graph might help... people like graphs, don't they...? Try Figure 5. Or Figure 6.

Typographically, they aren't bad, they follow tips for decent graphs, e.g. avoid black borders around columns (unfortunately, Excel often adds them automatically). Also, in Figure 6, we've put columns horizontally so that labels fit in better and are easier to read (e.g. 'Have one - last year', etc).

Typographically, not bad. But the graphs don't work. Legends and columns conspire to create puzzles for readers to decode. I’ve tested such graphs – no, graphs less busy than these – on thousands of people. And we struggle with them. We struggle to answer even simple questions (“how many are up?”). The graphs don’t work in overview or detail.

The moral of this story? Avoid multiple clustered column charts. And – we haven’t seen any today, so just take my word for it – avoid paired clustered column charts, e.g. ten columns that go blue, red, blue, red, etc for, say, five divisions and their income for 2018 (blue) and 2017 (yellow).

But don't give up on graphs. Instead, try the line chart in Figure 7. The left is 'Last year', the right is 'This year'. Light lines are ‘Have one’, dark lines are ‘Urgent’. How many have gone down this year? Easy – count the downward-sloping lines. Just one – Pa Urgent. How many stayed the same? Five - five lines are flat. The graph surrenders patterns easily.

The graph has various names. Slope graph. Parallel graph. I call it a China Surges graph because I first did it to show that China had surged… What I do know is this: it's been on this planet longer than I’ve been, and I’m no spring chicken. And the FT does it a lot nowadays.

(For help doing them, visit my downloads page and try File 018. If you prefer a file with fewer macros and automation, email me for a more manual version.) 

We're near the end. Time for two final thoughts:

With thanks to Ehrenberg. As a Professor at London Business School in 1982, he wrote eight pages on tables that included some of the above tips. A few years later, I stumbled across his pages and they lit the ‘clarity’ fuse in me. His work kickstarted my obsession.

“Get a life…”: I suspect some people will, by now, think: “Blimey … all that for just a graph – I’ve better things to do with my time” (and they then obsess over the colour blue for their PowerPoint autoshape…). It’s such a shame though – we spend weeks getting numbers on Artificial Intelligence, then take 30 seconds to turn them into an awful graph. Weeks of effort… wasted. Yet it doesn’t take long to read ideas for what to do instead (this email… five minutes?). And not much longer to turn the numbers into something so much better.

Which is a positive note on which to end.

Clarity and Impact Ltd | +44 20 8840 4507 | |

To see more of my previous email updates, click here.

Want to sign up for my monthly emails? Click here.

Want my GDPR policy? Click here. It's a bit irreverent, plus has two jokes.

Clarity and Impact Ltd