I had fun with this one, deciding to do a connected scatterplot. There is much debate about the efficacy of connected scatterplots. I agree they require an extra mental leap to comprehend, but once the meaning is grasped, I think they are powerful.
As Hans Rosling proved, they are especially effective when animated, hence the GIF. Following the animation is like following a story, the up/down and left/right movements adding drama and anticipation for the viewer. A static version of the chart is below.
The original this week, on first glance, wasn’t too bad. There’s nothing inherently wrong with a dual axis line chart. However, once you try and work out which axis is which and which colour is which, you realise the creator of this didn’t think through the positioning of the labels!
A simple makeover would have been to fix the labelling and make things much clearer using colour to help the reader. Here’s what that would look like:
I’m happy to concede even that the simpler approach may be better than my connected scatterplot version!
Boy, the Americans need to get saving. That’s what I discovered this week. The source chart was a stacked bar. This is an interesting choice by Andy: the chart wasn’t a total disaster, but there’s certainly room to play around with it.
Also, stacked bars have been much debated recently. Stephen Few has argued that, if a time series is involved, you should always use lines to represent the category instead of bars. Cole Nussbaumer and others have made the case that while the lines allow for clearer comparisons, they don’t show “part-to-whole” relationships as well as stacked bars. You should go and read the fascinating debate on Stephanie’s and Steve’s blogs.
Here’s this week’s original chart:
What do I like about the original?
It asks a clear question
You can look up every single value, even if it takes a while
What could be improved?
Double-labelling the category is overloading. I chose to just use the category name (Baby Boomers) and not show the ages. I acknowledge that means that if you don’t know the ages of those categories you’re stuck. However, the message is still clear, I believe
The colour scheme is arbitrary
Moving your eyes between the legend and back again is difficult.
So how did I go about my makeover?
I first tried a diverging stacked bar, a method developed by Steve Wexler. He places the “negative” categories to the left of the axis and the “positive” categories to the right. This allows you to see more easily the overall positive/negative leanings.
However, my diverging stacks didn’t thrill me, so I thought I’d try a unit chart instead. I like the way they simply represent percentages. I was always struck by the brilliant simplicity of the ONS’ “How well do you know your area project” and wanted to do something similar.
I confess I had to construct a separate dataset to make the unit, because in Tableau every dot needs a separate record (although I suspect Joe Mako could make it work with just two rows)
Once I had the data, I played with shapes and settled on blank/filled circles to show the threshold between adequare and inadequate savings. I chose a colour scheme which still showed the different inadequate saving levels (e.g. “no savings account” and “under $1.000”), but all within a red palette. I also added a parameter to allow you to set the level of “enough savings”. You can play with that by opening the interactive version.
I also created a horizontal and vertical version. The horizontal one works well in tweets, the vertical one is better in a blog. Font choice this week was inspired by Kelly Martin’s amazing Font Choice blog post.
I was just on a call with Zen Masters Steve Wexler, Jeff Shaffer and Robert Rouse. We were talking about formatting labels, and Robert was saying “Well, of course, you can just drag the labels around.”
“Wait. What?” I said.
“Click on the label and drag it,” said Robert.
And thus I discovered a cool new trick. How many one-off charts have I struggled with because Tableau didn’t quite put the label where I expected it? (Answer: hundreds, at least). This trick is going to make MakeoverMonday much easier!
All you do is turn labels on, and to move a label, click on it once, then drag it.
EIGHT years I’ve been doing this Tableau thing, and there’s still new tricks to learn!
This week’s source chart is a shocker. There’s very little right with this chart. The title is good, because it asks a question. After that it pretty much falls apart. The actual percentage value is very hard to see. The donut doesn’t help. The percent change circle is totally confusing. They also used a negative scale to size the circles which requires mental hurdles to overcome.
My makeover is at the top.
Comparing growth and decline between two time periods immediately cried out “Slope Chart”. It could have been a side-by-side bar chart, but I am not a fan of those. A dumbbell chart might have worked too, but showing the direction of time is a challenge. I tried Comet Charts to show this, unsuccessfully, a few years ago.
I really wanted a pure slope to work, with all lines within one pane.
However, this didn’t work for me. Too many of the lines were too close. I had to do callout labels for Television. The four lines along the bottom are all a little too close to each other for my liking.
Instead I resorted to a pane for each item. Here’s the version without shapes:
It’s still easy to compare the slopes of each line but not quite as easy. The cognitive load is a little higher this way. As you can see at the top, I chose to add icons. I did this to make it easier to identify each line without the requirement to move your eyes to the top of each pane. Icons are more appealing than labels. However, using icons can be a challenge: what’s a universally identifiable icon for “smart thermostat”? How do you tell the difference between a Smartwatch and Fitness Monitor icon?
We chose this week to stick to greyscale. Remaking this chart was straightforward. Because there are multiple panes, I don’t need to use colour to identify each product.
I added the annotations to help users read the slopes, if they’re not familiar with them. Being in greyscale this week was a good reason to soften the text to a very light grey. This helps the annotation disappear when you want to focus on the marks. If the labels were as black as the lines, it all gets too much:
Like my friend Rob, I never got wildly into David Bowie. His passing, however, made me realise how he was one of those people who are fundamentally a part of my music history, even if they weren’t on regular rotation, or if I never got beyond his singles.
I thought I’d look at when I listened to Bowie over the last 6 years. It’s interesting – he was a real “binge” artists for me. The line shows cumulative listens over time. Look at how the line goes up in steps.
What’s happening? With Bowie, once or twice a year, I’d think “Bowie was amazing. I’ll put on his greatest hits.” And then I’d be listening on rotation for the best part of a day.
Favourite track? Before even looking at the data, my answer was Heroes. The data proves it –
Stephen Few has been discussing 100% stacked bar charts. He’s opened a great debate with Cole Nussbaumer. I recommend you go and read the blog and the comments. In a nutshell, Steve prefers lines over stacked bars because they allow comparison of all data points at all places in the time series.
My problem is Steve’s assumption that Cole intended people to be able to compare all data points at all times in chart. In one of the comments, he says this:
“The information and eventual understanding that we get from a graph is more important than any impression that it provides unless that impression is all that matters.”
What if ‘that impression’ IS all that matters?
Sure, Steve’s lines allow you to do more accurate comparison of more data points. However, why should that be the only purpose of a chart? That purpose is one he imposed on Cole’s chart. It might not be the objective Cole had in mind when designing it.
Maybe Cole’s objective was to highlight the % missed only (ie the red marks on each chart). The other bar segments are secondary information, not vital to her key objective in making a 100% stacked bar. The “impression” I get from her stacked bar is that Missed Targets have increased to 42%. If her intention was that I get that impression, then it’s a success.
I might argue that if that was her intention, she could have drawn just the % missed data, and ignored the rest. However, this is at the expense of secondary information. Here’s what that could look like:
Steve also wanted other example of multi-segment stacked bars. I found one from my UK Election project:
I acknowledge the problems with this chart: it’s very very hard to see which organisation tweeted most about UKIP, or the Lib Dems, or any of the other central segments.
But my intention was NOT to allow comparison of every segment. My intention (as shown by the title of the view) was to highlight which organisations were tweeting most about the Conservative party. That was my prime goal. The Conservatives are the left-most segment and sorted in descending order: it’s easy to see which orgs tweeted most about the Conservatives.
Other information which can be learnt from this is secondary to my prime purpose and therefore was intentionally compromised by using stacked bars. An interactive version would add tooltips and more contextual data, allowing the curious to discover more in the chart.
I acknowledge the stacked bar isn’t perfect, but I don’t know how else I could have designed the chart so that it answered my prime intention (% of tweets about Conservatives) and allowed the viewer to see secondary information. How would you have redesigned it? If you wish, the data is here.
Remember, every visualisation is a compromise. And every visualisation has a prime intention which must be considered before critiquing it.
I wholly recommend the following posts for further reading:
In fact, I constructed the whole of this post about the chart above. As I wrote the final comment “but I didn’t actually think the original was too bad” I thought, maybe I’d just go back and build the original one in Tableau. Soon I came up with this:
I prefer it to my original makeover. So this week’s submission. is really nothing more than a format change. Why do I like the reformatted one over the redesigned version?
There’s lots of detail in the histogram behind the obvious annotation and colour messaging
The shading highlights the Dreaded-Twos distances. The point of the article the chart is from was about the dreaded-twos, but the original chart didn’t call out those distances well enough.
Regarding the first makeover (the one at the top of this post), here’s the notes I made:
Let me confess: I’m not too happy with the outcome of this one. However, here’s what I was trying to acheive in this week’s makeover:
The original story focussed on the fact that Stephen Curry shoots from close or far, not from “the dreaded twos” distance of 7-21ft.
The original chart didn’t group his shots into the three categories, which was what I felt was wrong. If the article’s about those three distances, then why not emphasise that aspect and group things together.
I also wanted to emphasise the points value. It’s not just about shooting from 18 feet or 22 feet, for example. The decision is also about points. The original chart doesn’t show the points value.
I tried to incorporate a basketball arc into the diagram to visually show where each group sits.
I probably messed around with colour more than anything else. I settled on the purple for the dreaded-twos and found that incorporating that colour into the title and subtitle helps decode the chart.
I did consider pie charts, as they do show part-to-whole very well but with this data, they didn’t come out well enough.
The original chart on Business Insider wasn’t fundamentally flawed this week, it only needed a few small tweaks to make it better. [hence the final version at the top – Andy]
With more time, I would have more fully incorporated the metaphor of the baseball pitch graphic.
What does this prove about makeovers?
Don’t reinvent the wheel if the wheel wasn’t too bad in the first place.
Iteration is vital. Iteration is always vital. Sometimes you need to go off on a tangent to realise you were in a good place to begin with.
In 2016, I’m joining Andy Kriebel on Makeover Mondays. Each Monday, we’ll take a viz, or some data, and create a new perspective. You can join in by checking Andy’s blog. Click here to find the latest data.
I have to admit I struggled with this. What’s the most important thing to show? Is it just salary? Or value? Or surplus value? In the end, I felt that all 3 need to be shown, but finding a good way to show it was difficult. I’m not even sure I succeeded. I wanted the salary bar to stand out, and tried all manner or Gantts/dual axes/stacked bar combinations. Only the one above came close to emphasising total value and surplus value.
What I like about my makeover:
Salary is labelled
Total value is labelled
Surplus value is implied but not labelled
Annotated mark to help comprehension
Sorted bar charts allow you to put the title in the white space, freeing up space for the marks themselves
I enjoyed combining Gantts and Bars on dual axes to make the marks this way
Custom sorting with parameters on the interactive version
What I don’t like:
I’m just not sure how understandable it is!
Formatting is pretty plain. I ran out of time!
I wonder if a simple bar of surplus value would have been enough?