I loved Chris’ original treemap (for reasons explained below). When it was made Viz of the Day, I heard lots of people say that it was a terrible choice: “you can’t make any insight out of that treemap”, they said. However, I sat in a big group of customers and partners that day, and showed the viz on a screen. What happened? They engaged in it – the treemap generated curiosity in a way my bar chart doesn’t. The subtle use of highlighting on Chris’ original teases people into exploring the data.
However, one of the first things I’d noticed was that the most common words were also the most common words in English. For my makeover, I wanted to exclude those words. I downloaded that data from Wikipedia.
It turns out that ‘baby’, ‘oh’ and ‘yeah’ are the most common of the uncommon words (if I did this again, I probably exclude the next few hundred common words to start getting to the uncommon ones). I like that “Na” is in this list solely because of Hey Jude.
We needed some Austin-related data for MakeoverMonday live at Tableau Conference. We turned to Restaurant Inspection scores from Austin’s data site.
I went in search of lunch in order to do the makeover, and found myself in Franks, home of hot dogs and cold beer. I sat down and ordered a bacon-infused Bloody Mary. Seriously? Bacon in a Bloody Mary? It was amazing.
Anyway, it got me wondering how well Frank’s had performed in recent inspections. That led my direction. I reduced the entire dataset to just Frank’s inspections. Turns out their last inspection was right on the borderline of failure.
My conclusion? Wonderful Bloody Mary. They passed my Restaurant Inspection!
Andy and I hope you all enjoyed MakeoverMonday live, wherever in Austin you ended up doing it.
Andy and I are very excited to be doing MakeoverMonday live in Austin at the Tableau Customer Conference, on Monday November 7th. This is our chance to say thanks to everyone involved, welcome some new friends, and play with data.
Here’s everything you need to know.
Where and when is it?
It’ll be at Level 2, Westin. We start at 2.30pm and finish at 4pm. Full details are in the data16 app (available on Android, iOS and Windows phone)
Will it be full?
YES. Due to many other factors, we could only secure a room for 100 people.
If you want a spot, get there early!
What if there’s no room?
We’re sorry we couldn’t get a bigger room. But all is not lost! Here’s what we recommend you do:
Say hello to the three people standing nearest to you.
My first thought on seeing this week’s original was to try another way to show distribution. I turned to the boxplot, an under-appreciated chart. Steve Wexler, friend and co-author of The Big Book of Dashboards, really dislikes them, suggesting that laypeople don’t understand them. I disagree, and think that a lack of understanding is only caused by lack of exposure to them.
Hopefully my “How to read a boxplot” instructional image at the top helps those unfamiliar with them!
Boxplots pack a large amount of useful info:
The whiskers spread to show outliers. Glasgow has a high SIMD score, but the data is very spread.
Comparing location is much easier. Consider Glasgow/Dundee in the original and the boxplot: It’s much easier to compare the two cities in the boxplot.
My boxplot still needs more work, which I would do with more time. I think it’s important to know how many data points are in each category. The Shetland Islands has a really narrow box, but that’s partly because there are only 7 items, compared to, say, 133 in Glasgow.
We threw you a curveball today. Only two numbers? This is a great challenge.
I did play in Tableau for a while, but then began to think about what these numbers really mean, and what the goal of the original infographic was. The problem is that comparing national debt to anything else is like comparing apples to oranges. And if you do compare it to other things, you run the risk of suggesting that because it’s so large relative to other things that there’s a problem.
That isn’t necessarily true. National debt isn’t like household debt, or currency, or assets. It’s a funny old beast. Here’s 3 of many amazing articles about this:
All of which isn’t too say that high levels of national debt aren’t a problem: they are.
My makeover’s goal was to make it clear that the different values aren’t the same kind of thing. Since there were only two numbers, it seemed right to pull out the pen and paper!
This week’s original
All of the above is one thing, but at the same time, I concede that the original wasn’t explicitly trying to say that US National Debt and, say, all the currency in the world, are similar. They were just trying to give you an idea of what the value represents. I do think there’s a lot of scaffolding around just a few numbers, but as an infographic to sit down and digest, it was a compelling read.
I got some real insight this week: all states ebb and flow for/against each candidate at pretty much the same pace. If that’s the case, why do the candidates pour resources into swing states, since all changes are reflected at the same level on the national scene?
This is reflected in the GIF below:
The US election is around the corner and this week we turned our attention to the polling data. We’d like to thank Drew Linzer for allowing us to use the data from the Daily Kos site.
My goal was this: how could you tell a story in a different way to all the poll trackers, without a map? I decided to drop the data on the independent candidates (sorry Johnson and Stein) and focus not on the actual polling percentage of Clinton and Trump, but the difference between their polling percentages. The gap is more interesting to me.
I goofed around with just making the chart of the gap, which was interesting. You can see that while Trump’s been in the lead a few times, he’s never pulled out a big gap, or held onto it for long.
Once I’d drawn that chart, it was then a simple case of adding State to the column shelf and realising that there was beauty and insight in the pulse of all 50 states. By adding the animation, I hope I’ve emphasised that all states go up and down at the same pace.
One thing, though – my chart is an elaborate way of showing the same thing as this line chart:
But where’s the fun in just doing a simple line chart? 🙂
The Daily Kos tracker is great. There’s a challenge with poll trackers: how do you make them interesting? The Daily Kos tracker is pretty similar to the ones on FiveThirtyEight, HuffPo, WSJ, etc. The good news is that in this election, the data was volatile, so the trackers were interesting charts to look at.
How do you communicate what the dots, marks, and lines on your chart show? Most often, you’ll use a legend. They work well, but check out the this from the Huffington Post. They created a Paragraph Legend (as I’m going to call it).
Why’s this great? I mocked up what this might look like if we used a regular legend. Try and decipher the chart using the “traditional” approach:
In order to decipher the chart you need to read the paragraph. Then the chart. Then go to the legend. Then back to the paragraph. Then back to the chart. Finally you might understand what’s on show.
Now look at the Paragraph Legend. Read the paragraph, look at the chart, and then maybe back to the paragraph once more. I found it much much easier to decode the chart with the Paragraph Legend. Like all small things, this is harder for the designer, but an improved experience for the audience.
[This is the second time I’ve reused Andy Kirk’s amazing idea to blog short posts on great things they see in dataviz. All credit goes to Andy for the idea. I’m going to call my series “It’s the small things….”]
I had a great time keynoting at the Crunch Conference in Budapest last week. What a great city and what a thriving tech scene!
My keynote was the Beautiful Science of Data Visualization: my favourite subject! The original content was developed by Jeff Petiross. My version has evolved from his, but they’re essentially covering the same content.
I was really impressed by Carlos’ sketchnotes. Too often, sketchnoting doesn’t actually capture info in a way I want to read it. However, Carlos creates sketchnotes which are amazing summaries. Go check out the rest of his stuff!
This week’s Makeover features a simple, effective stacked chart from FT.com. Rather than find multiple new stories in the dataset, I focussed only on the original story: satisfaction in 5 European cities.
Decision 1: if you show % satisfaction, you can get away with not showing % dissatisfaction since they’re almost binary. (that’s not entirely true: there is nuance in the differences between “very” and “rather” satisfied/unsatisfied) but I do think it’s valid.
Decision 2: bring in the delta for extra context. Berlin is not only the city with the highest satisfaction rate, it’s overtaken London in the last 3 years. That story is not visible in the original.
Decision 3: improve on the comet chart. Once I’d made the first two decisions, I figure it would be a good time to draw a comet chart. However, they are hard to read, as I’ve written about before. I think I’ve solved the problem by fading the trail. Do you agree?
Decision 4: set the x-axis to go from 0-100%. Berlin’s the highest satisfaction but it’s still below 50%. Setting the x-axis at 100% is intended to highlight the hidden levels of unsatisfaction.
Decision 5: Formatting. I didn’t see a reason to format it any differently to the FT’s original. I love their background colour! I did move the title to the blank space created by the long x-axis though.
The ordering of the colour legend is confusing. I read it left-to-right, top-to-bottom. But it’s order top-to-bottom, left-to-right. I initially thought one colour was missing from the chart until I realised this.