Love is in the air this week with a makeover of an infographic on valentines day spending in the US.

The original visualisation is pretty good although some of the key data (like average spend per person) doesn’t necessarily jump out. Also there’s nothing to show changes over time, even though the data source does contain that information.

So for my redo I wanted to focus on a very clean presentation of the main trends over time, whilst still highlighting some key stats. I also wanted to offer the viewer the ability to explore the data a little more – something I haven’t done in many of my makeovers this year.

.

The interactive version is available on Tableau Public here.

Getting nice clean spark lines was just a matter of turning off all formatting and adding calculations to provide label text for the first and last points only – a few people have covered that technique this week so check out some of the other blog posts. The heart shapes with the key stats are achieved by using a dual axis chart where the additional series has a single point, in a position specified by a calculation (in the example below at year 2013 and 62.5%, note that with no ELSE only the one point is plotted as the rest are NULL). The point is then given a custom shape, the size beefed up and label text added.

A quick redo of the pie charts in this Business Insider article for #MakeoverMonday week 5.

.

If you’re thinking that something seems dodgy with these charts then you may well be right and should have a read of @ChrisLuv’s comments which are an excellent read.

In addition to these observations the pie chart colouring doesn’t highlight the message – for me Germany sticks out more than the US based on the colour choice. Other #MakeoverMonday participants have commented on the use of pie charts; personally I don’t mind the pie charts here as the message is to compare the US segment to the rest of the segments combined. Although if doing this I would have coloured the other segments similarly and highlighted US as mentioned.

For my redo I went with a bar chart but combined the other G-7 countries into a single bar. Doing so allows the message in the article (whether right or wrong) to stand out more for me, whilst still providing some of the detail for the other G-7 countries.

I’ve experimented with labelling the other G-7 countries by creating duplicate sheets where mark colours are set to fully transparent so that we seem to get a set of labels aligned under each chart – a bit of a hack and I’m not sure it really works given the distance from the labels to the bar segments?

I spent more time looking into the data than on the visualisation for this weeks #MakeoverMonday because the data related to New Zealand. The task this week was to make over the international and domestic tourism spend charts on figure.nz. The international chart is shown below:

The charts are very clean, but showing each year side-by-side makes it hard to read for me. The key seasonality of tourism spend emerges nicely but also makes it harder to spot trends.

The charts use RTI (regional tourism indicator), but that is an old superseded measure. MBIE now use MRTE (monthly regional tourism estimates) which is a $ figure rather than an index.

You can read more about MRTE and RTI here on the MBIE website. One of the key differences I was interested in was that the RTI uses “a relatively coarse definition of domestic tourism … [which] conflicts with the official definition” – in essence some domestic tourist spend is discounted under the RTI.

Running short of time for my makeover I’ve focussed on showing the differences between the two measures:

.

A number of community submissions focussed on events such as the premier of the Hobbit movie, instead I’ve shown the marked temporary increase in tourist spend during the 2011 Rugby World Cup which jumped out from the seasonally smoothed 12-month moving averages.

The visualisation is also available on Tableau Public.

A reviz of global iPhone sales over the last decade for week two of Makeover Monday in 2017.

On first glance the only thing I wanted to change from the original chart was the slight 3D affect on the columns, and maybe the background colour. Other than that the chart has a clear and simple title and highlights the data point addressing the question posed.

Digging into the quarter-by-quarter data there seemed to be a bit more of a story and, for me, the addition of a moving average helped to smooth out seasonality and see this story:

.

Adding the bounding boxes around 2015 and 2016 was a mistake as they impinge upon the hover over functionality in the underlying workbook on Tableau Public. They also add some visual clutter, but I was keen to help differentiate and highlight the two years.

On a wider note, there was some good discussion amongst the community about the dangers of data quality and drawing false conclusions (for an example see my previous blog entry on the pitfalls of using taxable income data to draw conclusions about salaries and wages, or dive into this blog post by Steve Wexler and Jeffrey Shaffer and resulting twitter discussions).

To an extent these sorts of issue are inevitable within the constraints of Makeover Monday. Not everyone can commit the time to really dig into the data, and the recommended time of one hour arguably doesn’t encourage it.  For most people the focus seems to be on honing Tableau and data presentation skills. But this is where I think Makeover Monday can turn a weakness into a strength (isn’t that what it’s all about after all?). Whilst most people might focus on presentation in any given week some will focus on, and discuss, the data. That’s great – as a group we’ve then covered multiple aspects of the original viz and data set.

If we were to somehow curate those discussions back to the datasets page on the Makeover Monday website, and every viz had a standard footer re data source (they should have this anyway) with a caveat that linked back to that datasets page, would that address some concerns? Might it also ensure that any one viz viewed in isolation would guide people into the wider discussion and in depth analysis pulled together by the community? Might that be a positive thing in terms of improving understanding of the original story being told?

I’ve had a go at adding such a standard caveat to my viz this week. It could certainly be refined, as could the viz itself, but seems like a good start.

The first Tableau Makeover Monday for 2017 looked at an article about gender inequality in Australian pay. The article is based on 2013-14 tax year data from data.gov.au. The original article presented the data in two tabular lists which made the comparisons being drawn hard to visualise. Unsurprisingly many of the makeovers represented the gap between male and female taxable income in a selection of occupations. One of the problems with the article, and a number of makeovers, is the assumption that taxable income is the same as pay; that is not necessarily the case as can be seen by digging into the original source data (which seems to cover taxable income from sources other than main occupation). I’ve steered away from mentioning pay in my version and simply tried to represent that in the bulk of cases men will generally have a higher taxable income than their female counterparts. Click on the image to see the interactive version, where hovering over a bubble shows you the detailed figures.

This weeks Tableau Makeover Monday was a challenge to visualise a small amount of data; two data points – total size of US National Debt versus the rest of the world. The original visualisation can be seen on the visualcapitalist website and also include comparisons of the US$ 19.5 trillion debt to thinks like company sizes, oil exports, cash held, etc. The pie chart works well here and the comparisons give some idea of scale.

I’ve stuck to the challenge and just used the two data points. Presented as a tree map but formatted similarly to a credit card – perhaps a good way to make the concept more real?

.

[Edit: Was it a good idea to use credit card imagery in this visualisation? Maybe not! As touched on by Andy Cotgreave in his Makeover Monday blog post, national debt is different to house hold debt. A better analogy might be a business loan – borrowing money to build a business (aka grow the economy). If I were revisiting the visualisation I’d also try changing the title and caption to refer to “Sovereign Debt” (with explanatory note) as the term ”global” could be misconstrued as all debt not just national or sovereign debt.]

This weeks #MakeoverMonday was a look at US presidential election forecasting data by Drew Linzer on Daily Kos Elections.

The original charts plot the average percentage being polled by Clinton and Trump over time, along with percentage undecided and other (independents). Personally I wasn’t sure I could improve on the existing charts or some of the community versions (loving the tile maps!) so instead I’ve focussed on a different angle - it wasn’t always easy to see at a glance who was predicted to win the election and why. Particularly with the complexity of the electoral college voting system.

Showing the states each candidate was predicted to win, ordered by the scale of the candidate’s lead with a running sum of electoral votes seems to work well:

Colour coding those states where the lead is smaller than the percentage undecided adds another potential indicator of confidence allowing us to make our own judgements of the likelihood of a particular result – how many close states would have to swing Trump’s way for him to win at this stage?

I would have liked to have added the overall country wide line chart – like the original – and to have dug deeper into the two states where electoral votes are allocated differently. Animating the chart over time, using the pages shelf, was an interesting exercise that I would have liked to explore further.

The full viz can be found on Tableau Public.

Or check out other community efforts via the #MakeoverMonday webstite.