Week 19 – Which cars do the Dutch purchase?

Week 19 was a makeover of a simple table of the top 10 auto exports from the Netherlands. It’s always interesting to see how people approach making over a table because there’s no real visual elements to makeover like a typical week when we provide a poor chart.

Before I review the week, I’d like to give a special thank you to Martijn Verstrepen of The Information Lab Netherlands for preparing the data and translating it this week. That was a MASSIVE help!

MAKING SENSE OF LARGE DATA SETS

The data set this week, and perhaps the lack of a viz of makeover, seems to have freed people up to do more analytical work. I really enjoyed how many people took this relatively large and very wide data set and focused on something specific.

For example, this piece by Kai Cheah demonstrates a nice combination of analytics and storytelling. I love the way he made the area chart look like smoke coming out of the back of the car.

Charlie Hutcheson decided to focus very narrowly on a specific segment of the data, supercar exports. Notice how Charlie didn’t create anything overly complex; it’s a simple bar chart that demonstrates the analysis clearly.

This piece by Adam Crahen focuses on the cars that are most frequently recalled. Again, nothing complex, but Adam has made it look beautiful, simple and engaging. Big, bold numbers and lines that guide the reader’s eyes around the viz are fantastic ways to engage your audience. That is NOT easy to do!

One last example of simple, effective analysis is this slope graph by Steve Wood. What makes this work well, like the others above, is taking a very broad data set and focusing on a specific piece of analysis. In this case, Steve show how the registrations have changed for the five most popular cars over the past two completed year.

TAKING INSPIRATION FROM OTHERS

We’d given feedback recently encouraging people to emulate the work of others that they really enjoy, yet giving the people you took inspiration from credit.

I loved Pooja’s viz from the prescriptions week, particularly the way she incorporated sparklines at the bottom. This week, I decided to emulate her. I’ll tell ya, she’s quite freaking fantastic because this isn’t as straightforward as she makes it look.

Sarah Bartlett was inspired by a viz she saw from The Economist and used that as the template for her viz. Not only did she create an amazing viz, she also wrote a blog post detailing exactly how she did it so everyone else can learn right along with her.

Before I show my favorites from the week, I have four quick lesson learned to pass on that I observed (and was guilty of myself) this week.

LESSON 1: USING INCOMPLETE DATA

This particular data set had a few interesting gotchas that I had a feeling people would fall for particularly with the registration dates.

When exploring data sets like this, look at the individual years. You’ll notice there are gaps. I saw several submission where people made the assumption that there was data for every year when there clearly is not.
Not all years have registrations in every month. The data doesn’t consistently have registration in every month until 1970. So if you’re trying to compare months, you really shouldn’t use anything before 1970.

LESSON 2: USING APPROPRIATE METRICS

When you want to compare the prices of car models, you should probably look at the median or average price. I saw a couple vizzes where people summed the price of the cars. That doesn’t make sense because the price would then be impacted by the number of cars registered.

LESSON 3: MISINTERPRETING THE DATA

I’m the first to call myself guilty as charged here. It’s subtle, but the data includes the date the car was registered. This isn’t necessarily the date the car was sold. In my viz, I made this wrong assumption. I did so with my UK hat on. In the UK, the purchase date and the registration date are the same, whereas in The Netherlands that’s not necessarily the case. I should have confirmed or clarified that.

I don’t know if my viz influenced others to make the same wrong assumption. However, if it did, I apologize.

LESSON 4: STACKED AREA AND BAR CHARTS

Stacked area and bar charts can be very effective…when there aren’t too many colors. There were a couple submissions this week where people used these chart types, but included like 10 different colors, sometimes more! This makes the patterns almost impossible to see. As an example, suppose you have a stacked area charts that shows the % of total cars registered for each year year by environmental label.

For me, there are way too many colors to see any patterns. What I would do instead is group some of the colors together and put the most important color at the bottom. This makes it easier to see the increase then sudden decrease of cars with “A” environmental labels.