MakeoverMonday: The Sugar Tax

Sugar Tax

Our chancellor proposed a sugar tax on soft drinks in his most recent budget. About time, too! We desperately need to improve diets and this is a start.

The BBC’s asked How bold is the tax? They used this chart:

What’s good about their chart?

  • ALL the info is there if you want to find it

What doesn’t work so well?

  • I need to move my eye many many times to decipher and compare things. Just comparing soft drinks between two age groups means I have to do a lot of cognitive work to find an answer.
  • The donut is arbitrary. A pie would have been better as there’s more of a mark to identify. If they REALLY wanted a donut, maybe they could have put the AMOUNT of added sugar as a label inside the donut.
  • The story is about soft drinks, primarily, but the colour for soft drinks doesn’t really stand out
  • The icons are ok, but what’s odd is how far away from the donuts they are. There’s too much white space between the donuts and the top of the icons.
  • Why is alcohol red? That makes it stand out the most, but the article is about soft drinks.

For my makeover, I decided to reduce the data to that most relevant to the story: how much of our added sugar comes from soft drinks. It shocks me to the core how much fizzy stuff our kids drink. 40% of added sugar coming from soft drinks? The BBC article also includes the actual volume of sugar, but we didn’t have that in our makeover dataset. It’s depressing.

Even if you kept all the food sources, the BBC’s approach could be improved by either a stacked bar, or even filling in the holes in the donuts. Notice that I’ve made Soft Drinks red in these examples.

Dashboard 3
As little as filling in the donut holes is enough to improve the readability of the chart: the marks are bigger.
Sugar Tax 1
A stacked bar allows for easier comparison.

Of all the three makeovers I’ve done this week, I think the pie might be my favourite: the part to whole relationship of the sugar really jumps out.

MakeoverMonday: Women’s Right Around the World

I went tall this week. There’s an interactive version here.

This month is women’s history month. Andy and I wanted to find a chart to makeover to mark the month. We found a great candidate at The Guardian.

Click here for the original.

What’s good about the original?

  • It’s a powerful dataset.
  • The interactivity lets you explore countries for yourself
  • The clear tick/cross icons on the right highlight where a country is equal or unequal in each category.

What could be better?

  • The circular layout serves no purpose. In fact, it confuses: it’s very hard to follow the categories around a circle to get to anything specific
  • The widths of the arcs are set by population, but that doesn’t have any bearing on women’s rights.
  • Colour is used arbitrarily.

The viz is not much more than a lookup table where you click to see details. The Guardian should have gone with a tabular layout at the least in order to make it easy for the viewer to find country/category.

I originally intended to build the same interactive viz as the Guardian, changing only to a tabular layout. As is often the way, though, I was semi-randomly exploring the data and the above view appeared. The data is detailed so it’s difficult to know how to show everything. In my makeover, I resorted to the tooltip.

Tooltips

With a rich dataset like this, it seems important to provide an interactive rather than static chart. But how do you serve the people who will interact and those who won’t?

My design is an attempt to please all. If you don’t interact, the layout is a sorted bar chart of the countries with the highest number of negative women’s right topics. Even without interactivity, a viewer sees a compelling story.

If you do interact, the tooltips give you context. It’s still imperfect because looking up a particular topic for all countries isn’t facilitated in my solution. Although Highlighting provides a partial solution.

tooltip

Highlighting

I am happy with my use of highlighting to let you click on a category and scan through all the countries to see how they compare, at the same time as keeping all the data visible in the background.

Highlightinh

Data problems

I do need to note that the data is a challenge. I’d say my makeover is a proof of concept and you need to be aware of the following caveats.

  • Lots of countries have answers to only a few of the categories. 31 countries with missing answers are excluded.
  • For all bar one of the topics “No” represented inequality. But many countries had Null or N/A. Is that negative or not? For the purposes of this makeover I classed it as negative. It might be that they simply didn’t provide an answer.
  • Some data is confusing. For example, one category is about the constitution. The United Kingdom has a “no” against all the questions. Is that because it’s discriminatory? No. It’s because there isn’t a constitution. In my viz, it’s still classed as negative. I appreciate this is incorrect, but in the time constraints of MakeoverMonday I do not have time to clean the data fully.

Update

Kristin Henry suggested a positive spin might be better. She’s right. My first update copied the headline and subhead direct from the Guardian. I took the lead from them, and we all know that news sites would rather tell you bad news than good. But this isn’t a bad news blog so here’s a postivie spin:

Women's Rights Around the World (positive)
Click here for an interactive version.

 

MakeoverMonday: Corruption Perception Index

Click here to see interactive version.

For the second week Andy and I have raided our vaults for old dashboards. This week, Andy’s Corruption Perception Index dashboard from 2010.

Andy K original
Andy’s original. Click here to see the interactive.

What do I like about Andy’s original?

  • A simple grid layout lets me see each dashboard region clearly
  • He used size to differentiate most/least corrupt
  • He’s got links to the datasource
  • There are clear instructions above the map and the table

What did I think could be improved?

  • It’s a very straightforward presentation of the data with no real agenda. It doesn’t inspire me to interact because there’s no indication of any conclusions I can make about the world from this data
  • Red/green: not a great colour choice. He does use size as a second encoding of the data, so it’s just about excusable.
  • Sliders for filters, not drop downs. Is this a problem? I addressed this in a comment on one of my previous makeovers. What do you think? Are slider filters good?
  • It looks so dated, don’t you think?

For my makeover I tried to achieve several things:

I tried to pose a different question. Instead of looking just at the CPI, I wanted to focus on whether the world is getting more or less corrupt.

more or less

On the face of it, the news looks ok: 83 countries got less corrupt between 2012 and 2015, whereas 63 got more corrupt. The story is of course more complicated than that. Looking at the world map, it looks like things are really bad in South America, but pretty good in Africa (lots of white means lots of countries becoming less corrupt).

SA and Africa

However, let’s look at the dashboard for Africa:

Turns out the story isn’t as positive in Africa as initially thought:What’s going on?

  1. More countries got more rather than less corrupt (25 versus 20). There’s lots of white in Africa because the countries that got less corrupt happened to be the bigger ones: they are more dominant in the filled map.
  2. Look at the CPI slope charts. Although 20 countries became less corrupt, they already had very poor CPI scores. Sure, they became less corrupt, but that hasn’t made them safe havens for citizens. The picture’s improving, but not by much.

In terms of design, why did I make my choices?

  • Instead of red/green I chose red/white against a black background. I want to make something with had more punch. I think the end result is a little too much.
  • To show change, I chose to do slope charts. When you interact with the dashboard, each line is labelled and highlighted. Without the interactivity, I acknowledge it just looks like spaghetti.
    highlighting
  • I chose filled maps, but could have used dots on the map. I’m happy to agree with you if you felt dots would have been better.
    dots or filled

 

Criticise positively: a lesson from #MakeoverMonday

MakeoverMonday has grown beyond my expectations. The community’s contributions each week are astonishing and create an amazing conversation.

One of the principles Andy Kriebel had when he started this was to be positive: nobody sets out to deliberately create a bad viz. Andy always points out the positive as well as the negative. MakeoverMonday is not about sneering about other people’s work.

In my Premier League Salary makeover post, I forgot this principle. Here’s what I said about the Daily Mail’s donut chart:

But, oh dear, this week’s chart is committing many crimes against dataviz. I wonder sometimes what the motivation behind a chart like this is. Do the editors just expect people to glance at it and think, “Oh, a chart. This story must be backed by data. Therefore it must be valid”? That must be the reason because this chart collapses under serious scrutiny.

And what happened on Twitter? Nick Harris, who wrote the article and collected the data, was upset, and started a defensive debate with me on Twitter.

We eventually worked it all out, but given the aggressive language in my critique, how else would I expect Nick to react?

I should have been much more respectful. I could have emphasized the things I liked about the viz and neutrally explained why I think the charts didn’t work. Had I done that, our debate would have been developed in a much better way.

This also raises a point about our own opinions and beliefs. I am not a fan of the Daily Mail. I don’t agree with its politics or approach to journalism. Did that cloud my reaction to the chart before I even looked at it? If this had been in the Guardian, would I have been fairer? I suspect the answer is yes to both those questions.

Here’s what I need to remember: MakeoverMonday is about building positive conversations. We find vizzes with great data which we feel could be improved in some way. The intention is to generate conversation, not upset the creators.

I was at Tapestry this week where Enrico Bertini said something perfectly captures what I think MakeoverMonday is about: “Use data visualization to generate ideas not truth. Data visualization is a creativity tool.” The variety of perspectives we see based on all the contributions prove that.

There have been many great articles about how our community should critique visualisations. I recommend you read them all:

(thanks to Rob Radburn for feedback on drafts of this post and Leigh Fonsecca for encouraging me to write this)

MakeoverMonday: Fairground Injuries

Fairground injuries continue to decline
Interactive version available here

For the next two weeks, we’re making over dashboards from our past. Andy and I have been using Tableau Public for 6 years: we thought we’d go back into the vault and improve on some of our earlier work.

What have we learnt in all this time and how would apply the learning?

Here’s the original, and below that, some notes on my makeover.

Fairground accident rates

  • The most important not to do is scoff at your early work, or anyone’s work that looks like this. Why? When I built this chart, and added it to the comments in the original Guardian article, it felt amazing. I felt empowered and excited at the ability to take some data and make my own sense of it.
  • I decided this week to go for maximum simplicity. For all the detail, the story across all criteria was the same: injuries are going down.
  • Given that that’s the case, a line chart with some filters seemed like an effective approach.
  • I pared back as much formatting as possible. In this week’s makeover I was trying to produce something you’d see in a print newspaper; something with a very simple, easy to interpret message.
  • The declining line reminded me of a roller coaster. I considered putting a roller coaster icon on one of the downslopes. However, this is one of those cases where you need to remember that this is real data about real people, some of whom lost their lives.