VizWiz

Data Viz Done Right

December 15, 2017

Visualizing IT Help Desk Performance

5 comments
Wednesday Eva asked for some help from The Information Lab:
I'm looking for some good examples of Customer Support dashboards (incidents, hours worked, forecasting, etc.).


I did a quick Google search and a search on Tableau Public and didn't find anything that was particularly good or useful (even from the support desk software vendors themselves). So I decided the best way to help was to create an example.

I didn't have any data, but I found this dataset on IBM's Watson Analytics website as a sample. The data set contains 100,000 records yet didn't include a date for when tickets were opened. It included a field for how many days a ticket was open though. What to do?

  1. Create a new column to generate a random date (to the second) between Jan 1 2014 and Dec 31 2016,
  2. Added a column for the Closed Date based on the difference between the Open Date and the days the ticket was open.
  3. Randomly removed the Closed Date for some of the records so that I would have some tickets that were not closed (because that's reality).

BOOM! Sample data set done! You can download it here

From there, I went back to my Google image search to get an idea for the metrics that were important. Many of the designs has a "card" style layout, so I create a design very similar to the P&L statement I created before.

I posted iterations throughout the day and into the evening and got feedback from Eva and Mark Kernke (of Groupon) as he had a similar need. Based on their feedback, here's where I ended up. This is merely a framework, but shows what a IT help desk performance dashboard COULD look like.

I hope you like it! Oh, bonus, there's a mobile version too. Check it out on your phone.

December 13, 2017

Alabama's Special Election: The 13 Counties that Swung the Vote

No comments
Unless you've been living under a rock, you know that Democrat Doug Jones defeated Republican Roy Moore in a special election in Alabama yesterday because the first Democrat to win a U.S. Senate seat in Alabama since 1992.

To help me understand where Senator Jones won, I downloaded the election results from Wikipedia (for the 2016 Presidential election) and from The New York Times (for the 2017 Special Election).

My Goals

  1. Understand the change county-by-county across the two elections
  2. Emphasize the counties that switched from Republican to Democrat

I was struggling with the wording of my dashboard, so I posted it on Convo for feedback from my colleagues. Here's what I posted:


The feedback was fast and furious, just the way I give it to them. Essentially this view wasn't terribly clear, especially the map. Ben Moss basically told me that I created a confusing (a.k.a. crap) viz and suggested using a blue color palette instead to emphasize the change towards the Democrat in each county.


Better, but this could easily mislead the reader into thinking that every county was won by the Democrat. Ravi Mistry suggested grey instead.


Nope! That doesn't work either. So back to the drawing board I went. Ben Jones and Jonni Walker were visiting The Data School today so I asked for their feedback. Ben suggested directional arrows and pointed me to his blog post for creating the arrow shapes I needed.

The next step was to take the slope graph and make it directional arrows, focusing only on those counties that switched from Republican to Democrat. From there, it made sense to split the map into two by election to give side-by-side maps and shade the counties by the party that won and the percentage of the vote.

Lastly, I cleaned up the titles and I was done. Fun exercise and I learned quite a bit about directional arrows and the Democratic stripe that goes straight through the middle of Alabama.

Workout Wednesday: UK Box Office Leaders

No comments
Workout Wednesday week 50 looks at a rocket ship chart of films currently showing in UK cinemas. Read Emma's requirements here.

The data provided doesn't match the data Emma used, so my results don't match exactly. The main requirements that took a bit of thought were:

  • Filtering to only films that are currently showing
  • Showing only the top 10 films (which Emma sneakily didn't include as a requirement)
  • Creating a tooltip on a title (I'd never seen this done before a Data School interview yesterday)

While these were the trickiest parts, I had an idea for how to approach them pretty quickly. My filtering was done differently than Emma.

Fun challenge and learned something new!!

December 11, 2017

Makeover Monday: Building Accessibility in Singapore

No comments
50 weeks into Makeover Monday 2017 edition and I think I met a data set that broke me. To be honest, I simply didn't get it. Well, I eventually did, but it never resonated with me. I'm not sure why.

Anyway, Eva posted this viz for us to makeover:


What works well?

  • The flashing dots tell you where to click.
  • Nice zoom action
  • Showing the rating out of five in the pop up provides useful information.

What could be improved?

  • Just about everything. The viz is all over the place and incredibly busy.
  • The viz needs focus so I know where to look first.
  • There are too many colors.
  • There are too many symbols.
  • The legend is massive! Make it smaller.

My Goals

  • When I started exploring the data, I noticed the likert scale value recorded for each building. Therefore, I thought of creating a survey visualisation similar to this one from Steve Wexler.
  • Create a map of the average score for each postcode.
  • Provide the survey results at the region level and use that to drill down into that region in the map.
  • Create a mobile-friendly viz.

With those goals in mind, here is my Makeover Monday week 50.

December 8, 2017

Looks vs. Personality - A Tableau Remake

No comments
Yesterday I wrote about a great viz from YouGov and in that post I wrote about some of the things I wanted to see improved in the viz. As a way for me to learn, I decided to try to rebuild the viz in Tableau, mostly because I knew it wouldn't be straightforward, and also to address some of the areas for improvement. There were quite a few tricks I had to mix together to make it work and I learned a ton along the way.

If you ever need a great way to learn, find a viz you like and rebuild it. You won't regret it.

December 6, 2017

Workout Wednesday: Position of Letter Occurrences in Baby Names

1 comment
I think you'll all like this challenge. It came about from teaching at The Data School and learning something new. I love learning every day!

Using this data set of names from the 1990 U.S. census (via data.world), build the same viz that you see below based on this set of requirements:

  1. The letter down the left is the first letter of the name.
  2. Determine the % of total times that a letter appears in the nth position of the names. In other words, the letter A appears in the 2nd position in 27 names, which represents 2.1% of names starting with the letter A.
  3. The viz should NOT display the % of times the letter appears in the first position.
  4. The % of total display should be sized by the number of names in which the letter appears in each position.
  5. The scale should show at 10% increments.
  6. Match the tooltip.
  7. When you click on a letter/position combination, those names should appear in a table below the chart. The table layout must look like mine.
  8. Notice how the names in the table are small. This helps them fit better (but not perfectly).
  9. Match my tooltips on the names table.
  10. When you deselect a letter/position combination, the table should disappear.
  11. Create a legend above the chart that reflects the possible positions of the nth occurrence.
  12. Match the colors on the legend, which should match the colors in the chart.
  13. When you click on a position in the legend it should highlight that position throughout the chart.
  14. Match my tooltips on the legend.
  15. The user must be able to select the nth occurrence they want to view. The options should be 2nd, 3rd or 4th and should look like mine.
  16. The title must update dynamically to reflect the nth occurrence selected.

I think that's everything. If I missed anything, please let me know. When you're done, be sure to post your solution on Twitter as an image (or gif) with a link to the viz and tag @VizWizBI and @EmmaWhyte. To help you, here's a gif for how it should work and farther down is the Tableau Public version.

Good luck!


December 4, 2017

Makeover Monday: Comparing the cost of food and drinks at Wetherspoons

2 comments
Week 49 is here and with it and interesting data set about the cost of food and drink in a chain of restaurants in the UK called Wetherspoons. The fine folks at the Financial Times conducted a fun experiment traveling across the country, ordering the same items at every restaurant with the idea to compare the prices. The FT normally does top notch data viz, but this articles is quite a poor effort.


What works well?

  • Binning the prices helps cluster the restaurants
  • Distinct enough colors

What could be improved?

  • The viz is desperately trying to be a map, but failing miserably.
  • The dots are so big that they are causing massive overlapping.
  • There's no interactivity, so I have no idea which restaurant is which.
  • The title doesn't tell me anything.
  • What do the dots represent?

My Goals

  • Explore the data, particularly focusing on price comparisons.
  • Are there regional differences?
  • Do large metro areas charge more?
  • Many restaurants make a lot of their money from drinks. Does Wetherspoons?
  • Elaborate on the price buckets idea from the original. How are the prices distributed?
  • Use FT-themed colors (which you can find here)
  • Practice highlight actions so that a restaurant can be more easily found on the map
  • Create a custom Mapbox map based on what I learned from our webinar last week.

With these goals in mind, here's my creation for Makeover Monday week 49.

November 29, 2017

Workout Wednesday: Treemap Drilldown

No comments
This week's Workout Wednesday from Emma builds primarily on several previous challenges, primarily:


If you haven't done those weeks yet, I would highly encourage you to tackle those first. They build upon each other and make this week's challenge much easier.

As for week 48, Emma asked us to build this treemap:


Go to Emma's blog for the requirements. For me, this was pretty straightforward as I'd done the other challenges. The only part that tripped me up was the sorting of the States. Once I got that figured out, it all fell into place.

Good luck! Here's my viz:

November 27, 2017

Makeover Monday: Two Simple Ways to Improve the World

2 comments
Today marks week 100 of #MakeoverMonday, the Community project. Sometimes I sit back and pinch myself at what it's become. Other times, I stand up and think about what more we could do. Without all of you, this project wouldn't be what it is today. Thank you!

To celebrate week 100, Eva and I will be hosting a webinar next Monday and we'll be joined by Andy Cotgreave (yes I know it's week 101). Register here. In this webinar we'll reflect on the first 100 weeks and answer any questions you have. It's going to be fun!

In addition, we have two more webinars coming up this week. You can find all of our webinars here.


This week, Eva picked something relating to 100


What works well?

  • It's quite attention grabbing due to the overall design and color.
  • The big text in the middle catches your eye.
  • Every section is clearly labeled.
  • Underlining each slice with the color it represents.

What could be improved?

  • There's no overall structure or order.
  • At first glance it looks like a donut chart, but each section adds up to 100 by itself.
  • The grey background behind each section label is unnecessary.
  • Within each section, the segments should be ordered by size.

My design this week is heavily influenced by Eva and Charlie Hutcheson. Both of them made very simple visuals that are quite stunning and stop you in your tracks and make you think. So how could I do something similar? Well, here it goes...

November 22, 2017

Workout Wednesday: Fun with Formatting

2 comments
The week 47 challenge builds upon this viz from Zen Master Rody Zakovich which has you taking 100 rows of data and displaying it in a single viz as five columns. Sounds simple right? It's not.

The data this week is the top 100 global brands, which you will need to download here. The data is only two columns: (1) the name of the company and (2) it's 2017 value.

Here are the requirements:

  1. Everything MUST be done in a single worksheet. Yes, the title, subtitle, footer, etc. all must be done in one sheet.
  2. Put the worksheet in a dashboard before publishing that is 1200x650, but DO NOT add anything else to the dashboard.
  3. Display the data in 5 columns, with each column being a grouping of 20 brands. 
  4. The brands must be displayed in order based on their value.
  5. Include the ranking next to each name. These need to be left aligned.
  6. Match the tooltips (easy)
  7. Make sure every line you see in mine is in yours.
  8. The title should be 24pt Tableau Semibold. The subtitle is Tableau Book 9pt.
  9. The footer is 12pt.
  10. The column headers are 14pt Tableau Medium.
  11. Everything else is Tableau Book 9pt.

Click on the image below for the interactive version. Post your solution to Twitter and tag @EmmaWhyte and @VizWizBI. Most importantly...have fun!

That's it! Good luck! 


November 21, 2017

Tableau Tip Tuesday: How to Create a No Whisker Box Plot

4 comments
In a recent Makeover Monday viz review, I mentioned how I find the whiskers on box plots distracting at times. This is only helpful advice if there's a way to create a no whisker box plot. In the box and whisker settings in Tableau, there's no way to NOT show the whiskers. So this week's tip shows you how to work around that.

Enjoy!

November 20, 2017

Makeover Monday: Snapchat is tops with American teens

4 comments
I have three teens, so this week's dataset had me particularly curious. While my daughter prefers Snapchat, my boys prefer Twitter. One big missing piece with this dataset is the lack of demographic information. I'm curious as to how my teens compare to those surveyed.

The original viz comes to us from Business Insider:


What works well?

  • Catchy title that quickly tells the story of the viz
  • Bar charts are simple to understand and they show the pattern well
  • Sorting the apps by the most recent value
  • Including the axis, otherwise we wouldn't know what the labels mean on the top of the bars

What could be improved?

  • The fading colors across the time periods are unnecessary. Include a time axis instead.
  • Change the numbers above each bar to percentages
  • The color legend doesn't match any of the bar charts and should be removed.

My Ideas

I started by looking at slope graph comparing the starting and ending periods.


This tells the story simply, but also doesn't show enough of the change over time. Next, I took the original and turned it into a line chart, labeling only the start and the end and also changing the colors to match the official colors of each app.


I think the line charts help make the change and trends much more obvious than the bar charts in the original. From there, I decided to look at the change since the starting period (spring 2015) to make the growth or decline of each app easier to understand. And with that, here's my Makeover Monday week 47.