18 November 2022

Gavin Freeguard
Warning: Graphic Content
12 min readNov 18, 2022


Chart attack

It’s been a while since I moaned about a pie chart, so… why not? (And also, why, BBC Wales News, why?!)

Who hates all the pies? A pie chart of reasons for not working in Wales, published by BBC News

If you are going to use a pie, at least order it from largest to smallest to help give a better sense of which slices are bigger than others (though there is a case to have the two ‘sick’ categories next to one another). Why say in the subtitle that the numbers are in thousands, when you could easily mark that on the labels? And you could put the categories as well as the numbers in the label to make it easier to read everything off the chart area.

Pie in the sky: an improved version of the BBC Wales News pie chart

Better yet, you can fit a bar chart in the same area, which makes the comparison between the categories even easier and is less busy.

Best bar none: an even better version of the BBC Wales News pie chart, which is a bar chart

I’m not entirely sure what ‘Other/discouraged’ includes, either.

It’s also interesting (if that’s not straining the definition too much) that the chart does not appear in the article it refers to. If it’s a useful illustration of the data, why not? Not everyone will have come to it via Twitter (especially at the moment…).

On the subject of assessing charts, there were a few that caught my eye a few weeks ago, but I didn’t quite have time to comment on:

  • Long-suffering readers will know I’m a fan of 100% bar charts in the right circumstances — and those who’ve suffered me delivering data training will know I like to put any ‘don’t know’ category in the middle, so you can read stories off both the left and right axes. But — given the size of the don’t knows — I think Opinium might have been better off with a dot plot here, which would bring out the Labour v Tory story more clearly. I find myself relying on the numbers on the right hand side for the closer results.
  • ‘Line charts are for time series’, as one tweeter reminded me when I tweeted a 2015 election version of this chart. That is broadly right, but there are circumstances — like that one — where they can be useful for other datasets. A few weeks ago both Reform and the Sutton Trust used line charts for data that isn’t a time series in a way I (and many others) would normally criticise — but I find myself thinking that the ordering (so one can clearly see which questions/accents scored lowest) and chart choice allowing different organisations/years to be compared might make it worth breaking convention. Maybe. Perhaps. I would say ask me who the Prime Minister is to test my sanity, but I’m not sure that really works any more…
  • Breaking with convention can work sometimes. Not always. This should have been a dot plot, even if the stacking of data that should not be stacked does distinguish high from low scorers; still, no.
  • And absolutely not.
  • And more pie/doughnut fun to finish. Some decent visuals in this piece, but the House/Senate majority charts at the very top are not among them. True, they suggest how close things are, but a) why the angled start b) why not impose a straight dark line to show the 50% mark and c) why not just use a bar?

Other bits and pieces:


  • Too many tweets I was there, sort of, for one of Twitter’s breakthrough moments in the UK — stood next to then-editor of The Guardian, Alan Rusbridger, at the launch of the Convention of Modern Liberty when he followed the ‘miracle on the Hudson’ via the bird app — Guardian accounts that had been hedging their bets on the next big social media thing subsequently sprang into (over?) use. For all its downsides, I find the possible demise of Twitter (it already feels quieter, he says pretending it’s an exodus not his terrible tweeting leading to less engagement) sad — there’s nothing quite like it, it’s where I get my news and many of these links, etc, and is the place where I share what I’m up to. It may provide an opportunity for many of us to reset our relationship with social media for the better, whether just using it less or finding new and more engaging ways to communicate on other platforms — or for the worse, as different communities atomise and silo themselves on different platforms and servers which may offer even less protection and more complications and risks. Elongate should also remind us how reliant we and our public sphere are on privately-owned infrastructure which may not be permanent (a point I should have made more of here). That’s a long way of saying I’ve signed up to Mastodon just in case — @GavinFreeguard@mastodon.me.uk — and I’m also on Medium as well as on this newsletter. And since chart criticism is a theme of this intro, here’s some more — the zoomed-in axis does, I think, tell a story, but it may not be the one Mr Musk wants it to be (are those valuable new users, or bots?). And the whole episode should remind us that moar technology and engineering is not usually what’s needed — as ever, it’s the human decisions behind it which matter — although the evident glitches in recent days suggest some engineers would be helpful. (Note: I wrote that last night before the surreal experience of scrolling post-iceberg Twitter after this news.)

No newsletter next week. Here are some trains performing Pachelbel’s Canon, just because.

Have a great weekend

Enjoying Warning: Graphic Content?

Tell your friends — forward this email, and they can:



Today’s links:

Graphic content

Statement of intent

Cost of living

Poor health

Capitol punishment


Tales from the crypto

Mappy talk


Out of this world

Energy and environment


Everything else

Meta data

Bills, bills, bills

Musk we?

UK government

UK parliament

Information health

Tales from the crypto

AI got ‘rithm

Open for the best

Everything else


And finally…

Food and drink

Everything else