Workout Wednesday – Benford’s Law

Workout Wednesday is another weekly challenge by the same people who manage Makeover Monday. Instead of re-imagining a viz, this challenge is to re-create a viz. It’s a great way to discover new techniques and develop better instincts for what is possible in Tableau.

Here’s my *slightly tweaked* version of the original viz.

Benford - Copy

I managed to re-create all the tricky parts! What I changed:

  • I left the font as Arial (I was working off of 9.3)
  • I tweaked the colors – the red and blue on the original felt too loud to me.
  • I re-arranged the text, and revised it to something easier to understand and more accurate.

This is the process I followed re-creating the viz:

  • Filter year
  • First Digit calc using CO2
  • Create bar chart (First Digit, Records)
  • %of total table calc
  • Exclude Nulls
  • Remove lines (including column axis ticks)
  • Setup Parameter, Parameter Calc
  • Put Parameter Calc into First Digit Calc
  • Multiply infant mortality rate, birth rate by 100 in Parameter calc
  • Adjust axies
  • Use reference line to add labels
  • Header
  • Tried using Gantt bars – Couldn’t create two color split at centerpoint
  • Expected Value, Top Point, Bottom Point calcs
  • *Trickiest Trick* Put those calcs on detail! 
  • Create two cell level reference bands

It took me about 4 hours working off and on to finish this week’s challenge, and only went down one wrong path. All in all, I think it went pretty well!

Makeover Monday – Success

This week’s Makeover Monday challenge was a very pretty infographic drawing a very bad conclusion. The original is really, really not okay and I’m not posting it here. IllinoisPixels has a fantastic write up of the issues that is 150% better than what I was planning to write! The only point I would add to their critique is that the order of the reasons for success seemed random. In my makeover I ordered them by the gap between rich and poor, which highlights the differing perspectives.

 

Baader-Meinhof alert! This data is supported by the headwinds/tailwinds asymmetry theory that I learned about on the Freakonomics podcast last week! It’s a good episode, and March is #trypod month, so I’m inviting you to go check it out!

Dashboard 2 - Copy

This week’s makeover isn’t very interactive, but you can still play with it at Tableau Public.

Makeover Monday – Credit Card

For this week’s Makeover Monday we’re working with credit card transactions!

Here’s our starting point:

AMEX Category Summary - CopyAMEX Monthly Summary - Copy

My goal was to create a visual that would help me understand my own credit card statement. For my own needs I would like to see my overall spending, categorical spending, and vendor level spending month by month. I also want to see a part-to-whole view of categories annually and by month. Ideally there would be several views that flow really nicely using intuitive actions and filters. That’s a pretty lofty goal, and my final product definitely falls short for now. I use Mint to track my spending, and working on this makeover made me appreciate their nice visuals a lot more! In fact, I think the Trends section of the website (not mobile!) is pretty much perfect.

Here’s the makeover:

Amex Makeover 1 - CopyAmex Makeover 2 - Copy

Go here for the interactive version.

After starting work, I realized that my preferred views could be grouped into categorical and monthly. By creating the two views, and then linking them with a bunch of dashboard actions I was able to show most of the desired views. I added a map with date on page because I wanted to work with the page feature. I kept it because it’s fun to use Andy’s purchases to watch his travels!

Continuous vs Discrete gave me a real headache this week! I started using discrete months, which make more sense for this data. The labels and size/spacing were perfect.

Amex - Discrete 1 - Copy

…But when filtered, months went missing.

Amex - Discrete Filtered - Copy

We might be able to correct this with a cross join (null or zero spend default for each category/month combination) but it feels strange that I couldn’t find a “include null axis values” setting. So, we’re stuck with continuous for now. Maybe I’ll revisit this if I have time in the future.

 

 

Makeover Monday – Potatoes

This week we’re working on a report of European potato production.

potato-original-1-copypotato-original-2-copypotato-original-3-copy

My biggest gripe with the original is that the colors are set by rank not by country. Poland is blue in figure one, orange in figure two, green in figure three, and red in figure four! It makes transitioning between charts more difficult than necessary. I was also frustrated with the alignment between the visuals and the sections they related to. Because of the sidebar format and number of charts and visuals included, the reader has to do a lot of scrolling to check the visuals while reading.

This week my goal was to make something prettier and easier to read, that would fit nicely in the sidebar and align with the article’s sections. I’m not really sure what the audience for this report is – maybe some poor sap is trying to research Iceland’s potato production – so I also tried to avoid removing data in my makeover.

In my makeover I assigned each country a color using the official national colors for inspiration. I’m not very experienced at building palettes, but practicing is fun! I used an area chart to show tonnes of potatoes because it combines change over time with part-to-whole. It’s not quite as effective as a stacked bar or line chart, but can show the information in a single chart, which would really help the visuals line up with the article.

For the Farms and Areas section of the report I was planning to use a scatterplot comparing number of farms with average farm size, but the MM dataset only included a few key measures so I created a tree plot instead. I think it does a better job of portraying area than a pie chart, and it kinda looks like farm fields too! When creating the makeovers I totally missed the first tab! So, I only have makeovers of the first two sections, but that’s enough for this exercise.

After creating my two visuals, I also built the first two tables in Tableau as dashboards. Then I used this guide to create actions that let the user go back and forth between the visuals and the tables.

potato-makeover-1-copypotato-makeover-2-copy

Check out the interactive version here!

Makeover Monday – New Zealand Tourism

Makeover & Original

Just a simple makeover this week! I wanted to include a map-as-filter section, but the file with the geographic coordinates couldn’t be opened by my personal freebie account, or by my work’s paid but version 9.3 account. I turned the bar chart into a line chart, and included each year instead of limiting it to the last three. I think this makes it easier to compare years and see the overall trend.

This viz is really very simple, but the underlying data has a lot going on. The region column includes “Total (all TLAs)” representing the entire country. Because the regional values are weighted for population, simply aggregating them would under represent larger, more populous regions, and over represent the smaller ones. Use the Total (all TLAs) rows to represent the entire country, and avoid aggregating the regions.

Workout Wednesday – Unemployment

I wasn’t feeling the Makeover Monday dataset for week three, so I did Workout Wednesday instead! (Love Emma’s approach of Trump v handlers tho!) The original required a worksheet for each state, which sounded *NOT FUN* so instead of sticking to the original challenge I got as close as possible using a single trellis chart. 

First I used dateparse on the years, and then created calculations for state and national unemployment rates. By fixing the national rate to the year, we can get them to appear on the same chart. (Note – remember that states have different populations, so don’t use the average of the state’s averages!)

Next I replicated the chart format using a single state. The line was simply the difference between state and national rates. Then I created a calculated field for the endpoint (if last() = 0 then [measure]) and added it via synchronized dual axis. Add a T|F calculation to color and you’re good to go. There was a lot of formatting. The pink shading was done using a reference band, and I used Times New Roman to approximate the original font. The trickiest part was getting the numbers to appear in the right format. ([Measure] > Default Properties > Number Format > Custom and then enter this: “+” #,##0.0%;”-” #,##0.0%;)

Finally, I built the small multiples grid. I had done this before with the lyrics makeover so this was a snap.

It was frustrating to not be in control of the headers, spacing, and axes of the small multiples. While it’s /possible/ to tweak these elements by creating separate worksheets, it isn’t practical or scalable. This is one of the few times I’ve felt truly limited in Tableau and would have preferred working in a platform with less structure and more control.