Makeover Monday, 2017 #42

Making over a table of Formula E racing results from the FIA Formula E website this week.

I was interested in how drivers progressed from practice to qualification so went with a bump chart. I’ve highlighted the winner but dashboard actions also allow the user to highlight the driver that they are hovering over. This way they can see how each driver’s performance changes during an event because the bump chart is quite hard to follow otherwise.

The viz asks a question and gives the user room to explore the data and come to their own conclusions, however a micro chart in the left hand margin also provides a general answer for those with less time or inclination to interact. The micro chart is based on average rank in qualification and race with a trend line to show the correlation.

You’ll spot that for the “super pole” session only six drivers are involved. To avoid a gap in the bump chart for the other drivers I’ve filled this rank from the previous session but not shown a mark / rank.

If I had more time and data I’d dig into the impact that fanboost has on the eventual winner.

The viz is also available on Tableau Public.



Makeover Monday, 2017 #41

Just a brief write up for now! Andy and Eva chose charts relating to adult obesity in America for the 41st makeover with the source data coming from the CDC. The original charts are part of an interactive dashboard and it is well worth checking out.

Similarly to Klaus Schulte I decided to dig into gender differences. Klaus’s makeover is very nice so definitely go take a look at that.

Here is my attempt:


It is also available on Tableau Public here.

Challenges were laying out the maps to include the more far flung states and territories, and also getting some decent labelling on the area and line charts. I’ve resorted to reference lines for the latest value and standard header labelling for the states (I would rather have had a label in the area shading).

Makeover Monday, 2017 #40

This week was a makeover of a Financial Times article on the UK economy since the Brexit referendum using data from the OECD.

The original


The original visualisation allows us to easily see the UK, but doesn’t allow us to compare to another G7 country as there is no labelling of the other lines or ability to highlight. The chart also lacks context around when the Brexit vote occurred and what the historical patterns have been; does the UK usually have slower growth than the rest of the G7? Missing 2017 Q2 data is a bit of a shame.


I’ve gone with three key elements in my makeover:

  1. Retain the line chart and highlighting of the UK, but show a wider period of time and highlight the referendum date
  2. Show the range of growth across the G7 without showing too many lines. Instead allow the user to choose which G7 country to compare to
  3. Provide a ranking table to highlight the low ranking for the UK in the last two quarters


The viz is also available on Tableau Public.

Final thoughts

My usual approach would be to plot all of the G7 countries as individual lines in a deemphasised colour. This week I wanted to experiment with showing the range of data in a different way. I had hoped that a light area chart would allow me to highlight the UK in red, show another selected country in a mid-grey and still see the G7 range in the background. Whilst the approach does allow the user to get a better idea of stability / variability for each country as they explore the data, the area chart doesn’t really sit in the background when it comes to interacting – e.g. hover over.

The area chart has two parts with the bottom part being shown in white to give the impression that you’re just seeing one area. This required calculations based on min/max values to get the two parts.  An additional challenge was that Tableau doesn’t handle area charts crossing the 0 line particularly well (or didn’t for me at any rate – I ended up with empty shards in the chart!). To work around this I calculated an offset and pushed all data points up by that amount so that everything was above 0. The real, none offset, values are used for labels and tooltips and a reference line was added for a fake 0 line. Upshot: it was a lot of work and fudging and perhaps not worth it. Feel free to download the workbook from the Tableau Public page linked above to see what you think.

I like the way the table at the end tells the ranking part of the story in quite an effective way. I prefer it over a bump chart in this case and in hindsight perhaps the table was actually all that the makeover needed!

Makeover Monday, 2017 #39

In August 2016 Nielson released a report what’s in our food and on our mind. Page 8 of the report included a set of spiral bar charts showing restricted dietary requirements around the world:


The charts on this page were chosen for week 39 of Makeover Monday. Whilst the infographic is visually engaging I’m not a big fan of the spiral bars. That and the way the different diets are laid out makes it quite hard to visually compare various categories (e.g. low sodium to vegetarian).


The finished makeover follows and is also available on Tableau Public.


Initially I wanted to present the data in a Gantt chart format to show the range of responses for each dietary category across the continents, with each individual continent plotted as a mark within the Gantt bars. There were two problems with this as you can see in the screenshot below. Notice how the coloured circles at the extremes of each bar are centred on the ends of the bar and hence spill out of it. Also for some bars you can only see four circles due to two continents having the same percentage:


I could have switched the circles for thin bars that were the same depth as the Gantt bars. Instead I decided to try something else. I was sure I remembered a submission in week 24 with rounded ends to a Gantt chart, so figured I’d have a go at that. I ended up achieving it using lines. Is that the best way? Check out my steps below.

Using thick lines instead of a Gantt

First up I defined the minimum and maximum percentage (followers) for each diet regardless of continent. This was done using two LOD calculations. Here is an example:


Next I figured I’d need two points with different values to use for the ”path” of each line. Point 1 being the start of the line and point 2 being the end. Luckily we have five points in the data so I just made two of them return a value and the rest return nothing (NULL – implicitly due to the lack of an ELSE):

Now I had a path for points 1 and 2, but I also needed a position to plot those points at on the horizontal axis. I used the same approach as above but this time returned the percentage when min or max:

To put this together I dragged the point onto columns, and the path onto the path card having selected a line mark type. A little bit of formatting to boost up the size and a label at the end of the line resulted in this:

The final step was to add the dual axis chart back in showing each continents actual percentage. I ended up boosting the size of these up and colouring them white except for a continent selected by the user. That way the focus is on the continent that the user wants to look at.

Interesting feedback was that the visual look was quite similar to iOS on/off switches. Oops! That wasn’t something I was aiming for, but I can see why people thought so. The danger of being too similar visually to a concept as familiar as on/off switches is that some users may think that they can interact with the bars as if they are switches.


Great to learn a new technique. Perhaps not so great to have ended up mirroring a UI component in a way that doesn’t follow the conventions for that component! Arguably I haven’t allowed people to compare continents that easily. If I were redoing the visualisation I’d take a look at both of those aspects. There was some great analysis by others in the community looking at different categories of dietary restrictions. I could have used two colours to differentiate diets most likely to be related to health as compared to those more associated with moral/ethical/religious choices. Finally a similar visualisation had a radio button selector for continent instead of a drop down – I definitely should have thought of that given the small number of values.

Makeover Monday, 2017 #38

A day at the races! This week we got a wealth of Strava data from Andy and Eva across two recent events that they have competed in. Find out more here. You’ll see that they each had a set of questions for the community to consider. I was busy trying to reproduce some of the graphs I’ve seen produced from sports watch / trackers and got very bogged down in the data. A quick look at the submissions that were coming through on twitter showed me that those submitting were focussing on a subset of the questions and only tackling one person’s data. Phew – it was good to take a step back! Eva’s question about the second kilometer of the run intrigued me. Friends that participate in triathlons have mentioned the initial pain over the first 500 meters but I hadn’t picked up on a lull after the first 1000 meters.

I’m sure I’ve learnt more from the exercise than Eva will, but here is what I came up with.

The viz is on Tableau Public too.


Makeover Monday, 2017 #37

For week 37 of Makeover Monday we’re looking at UK bicycle thefts based on data from

The reference site linked above has pretty extensive visualisations and I love the way they unfold as you work your way through the site. I’m not going to try to improve upon the site, instead I’m looking for a different angle to show and, given the bleakness of the data (less than 1% of bike thefts resolved!), it’s an opportunity to try a paired back black and white viz.



Trying to remove unnecessary colour is some advice from Andy and Eva in their weekly roundups. Does it work in my viz? Feel free to let me know on the twitter thread - any feedback is welcome. Also check out the Tableau Public interactive version where you can dig into a particular constabulary to find your local story.

A note about the resolution calculation:

The source data came with a range of outcome descriptions which I’ve mapped to three resolution categories: unknown; resolved and not resolved. If you’re interested in digging into this mapping I’ve included it below. The reason for the unknown category is that there seems to be a lead time on getting a resolution (positive or negative) as a case works its way through the system. Having a yet to be determined / unknown category means that we can factor those out of a resolution rate. Otherwise we’d expect to see a tail off in the most recent months as the thefts are yet to be investigated, etc. Arguably a couple of the outcomes that I’ve categorised as resolved are a little unclear and a proportion of those could actually be unresolved or still in progress.



Makeover Monday, 2017 #36

The UN, #MakeoverMonday and #VizForSocialGood came together to challenge the data visualisation community to visualise results from the MYWorld survey on the UN Sustainable Development Goals. You can read more about the challenge on the Makeover Monday blog.

The design is geared towards a tablet as there were indications that the UN would like to use the finished viz on stands with tablets at various UN events like the forthcoming UN General Assembly in New York.

I’ve split the story into three pages to meet different aspects of the brief:

  1. The overall demographics of survey participants (I’ve also used this page as an opportunity to introduce the goals and the MYWorld survey questions);
  2. Responses across the 17 goals for question 2 and 3 of the survey (a sort order selector in the column headings allows the user to consider different rankings); and
  3. The ability for users to compare their country to another country (indications were that stakeholders were interested in, and motivated by, comparisons to their neighbours).

The proposed visualisation is available on Tableau Public with data extracted as at 23-Aug-2017 – screenshots below.

Page 1


It would be possible to include a country filter on this page (in the right hand column header) and to include a call to action prompting the user to consider what participation is like across various demographics within their country – and what they could do to increase or maintain participation.

The graph of participation over time is intended to show how participation is trending in terms of the goal of 1,000,000 respondents per year and a good place to show how many participants are aware of the goals over time.

Page 2


The continent and country filters at the bottom allow the user to focus in on their area of interest. They can move to page 3 for more detailed comparisons of specific goals. An improvement would be to allow the user to click a goal on page 2 to jump to page 3 with that goal selected.

Page 3


As well as the country comparisons this page also allows the user to compare responses by some of the other categories in the survey data – e.g. education level and HDI (human development index). Finally the page ends with a call to action to take the survey and a thank you for exploring the results. An alternative use of this space would be a call to action geared towards meeting the sustainable development goals.

Makeover Monday, 2017 #35

A quick post for Makeover Monday this week. We had data on NFL player arrests from 2000 to August 2017 with the aim to makeover the interactive visualisation here. Whilst exploring the data I was interested to see the free text outcome data. After grouping this I was surprised to see that the proportion of guilty outcomes seemed to be reducing over time whilst the proportion with an undetermined outcome was increasing . In hindsight you’d expect more recent cases to not yet be determined – perhaps they haven’t yet worked through the system? Still I suspect that there is a story here. Of course that story may not be specific to NFL player arrests and I haven’t checked for similar stats across the wider population.



You can access the interactive version on Tableau Public here.

Makeover Monday, 2017 #34

So apparently there was a pretty exciting solar eclipse this week. I guess that’s why Eva picked a NASA data set on solar eclipses for Makeover Monday! Go check out the original article as it’s got a great map of eclipse paths. The data we got for our makeover didn’t include paths, instead it included one coordinate for each eclipse over 5 millennia along with data on the type and duration. Still, the community produced some amazing and informative visualisations.

Like others, I was seeing interesting patterns across longitudes and latitudes during initial exploration. I started off with bar charts based on longitude and latitude bins but decided there was an opportunity to try a radial bar chart in Tableau. My trigonometry is a bit rusty so I turned to Rajeev Pandey’s “how to” guide to get a start on the maths. Thanks Rajeev!

Here’s what I ended up with:


For longitudes it felt appropriate to use the full 360 degrees, but for latitude which stretches from pole to pole I felt a semi circle was a more accurate representation. Labelling the max values on each chart gives some context and I added subtle Greenwich / South pole labels to the first chart on each row to orientate the viewer.

I loved what people were doing to explain the different types of eclipse so added that as hover over text on a central circle. Colour choice was intended to help the charts pop although in hindsight I should have tried a black background. Have a look at Michael Mixon’s excellent submission to see what I mean. I also loved Colin Wojtowycz’s viz which highlighted the interesting data around latitude of partial eclipses.

Finally if you want to try the interactive version of my viz, head over to Tableau Public.

Makeover Monday, 2017 #33

Earlier this year the pudding published an excellent analysis into the myth that various events were a trigger for mini baby booms. The analysis was based on CDC data on natality in America. Fast forward to August and the data set was selected by Andy for week 33 of Makeover Monday.

Instead of the usual makeover I tried to look for a different angle in the data set; after all visually exploring data and finding the stories is one of the strengths of Tableau right? Like a few other participants the variations in the mothers average age and average baby weight at birth intrigued me. My first thought was that I’d made a mess of averaging as we were provided with pre-aggregated data at county level not state level. Luckily we have the number of births so can calculate a weighted average to use at state level:


If you’re not sure what’s going on here then check out Charlie Hutcheson’s blog and his link to Andy’s article on averages of averages.

I’ve focussed on weight at birth in my viz, highlighting the states with the highest and lowest average over the whole period (check out the LOD calculation and accompanying ranking table calculation if interested … oh and if you can tell me how to calculate a sortable column based on the rank so that I can programmatically ensure that the highest and lowest lines always sit on top then I’d love to hear how as I ended up fudging that bit!).

To encourage engagement I’ve also added the ability to choose a state to highlight, so that US users can see where their state sits.

Here is the finished viz:


The viz is also available on Tableau Public here.

One final comment – why did I make the viz so small?

Well, I’ve been trying smaller charts and dashboards a lot with Makeover Monday. There are a couple of reasons for this. The initial reason was to make them more mobile friendly as we all consume so much of our content on smart phones these days. I find it frustrating when I see a cool viz in my twitter feed, go to interact with it on my phone and can’t really engage.

A secondary reason, which I’m mulling over as a good piece of advice, is that small =  less space for clutter = you have to really work to simplify and distil your story to the key points. We often see advice from Andy and Eva in their weekly write ups to keep things simple – e.g. don’t just chuck several graphs at a dashboard and call it a day. If you suffer from this then perhaps aiming for smaller and smaller dashboards each week would be a good learning exercise?

Of course there’s a point where things just don’t suit a small screen, or get oversimplified, or the author just makes the font smaller to make things fit! Have I done that this week? It’d be great to get your thoughts back on Twitter.