Really good challenge. I particularly enjoy the workouts that have practical use cases for every day work. Thanks Luke for the challenge!

What works well?

What doesn't work?

• There's no real story to the data.
• Tables aren't engaging.
• Comparisons are difficult - How do the top 20% compare to the rest?

Really, there's not a whole lot more to say. I started this week with a Google search for some inspiration. I was looking for graphics that caught my eye, interesting statistics, good colors, etc. In the end, I chose to use the colors from this visualisation for my palette because I thought they fit well with YouTube's primary red.

I wanted to answer two simple questions with my visualisation:

1. How do the top 20% of channels compare to the rest? Do they fit the Pareto principle?
2. How many views do the top 10 channels have and how can I best compare them?

What doesn't work?

• Inconsistent colors
• Hard to identify the "story" in the data; The story is buried in the article.
• Pretty busy overall; too much going on
• Tables are terrible for finding insight in the data

I wanted to create a beginning, middle and end to the story, and I feel like I did that. I used a question in the infographic title to help the reader understand what the viz is about. I used dividers to the viz into "parts" of the story and I used the chart titles as legends. Lastly, I used Roboto Condensed font to match the font used in the article.

It's election season in the UK and The Guardian recently released some data related to Prime Minister David Cameron's travels around the World. They provided a table of all of the data at the bottom of their post, but the only viz they provided was this view aggregated by continent.


First, bubbles make comparisons harder than they need to; bar charts are much better for comparisons. This view also doesn't provide any addition insight. I can't answer a simple question like "Which countries did Cameron visit in Europe and when?"

I only had a few minutes this morning to create an alternative and here's what I came up with in about 15 minutes.

The irony of it all.

• I’ve been riding the Facebook shuttle to work for almost a year.
• I’ve been tracking my commute time via Swarm check-ins.
• I decided to drive to work yesterday for my last day.
• I got my first ever speed ticket on my way to work.

I started building the viz below a while ago and decided to finalize it since I had collected all of the data I could. I would like to thank Anya A’hearn for her feedback on my design (note to self: Anya hates lollipop charts). I also need to thank Jonathan Drummey for his help in getting my time calculation formatted correctly (he pointed me to this post).

Anyway, here is my analysis of my commute to work on the shuttle. There are filters for the time of the commute and the schedule. Use the drop down to select various displays: the default is a calendar view and the other views are various dot plots.

Click on the image for the interactive version.

Ever since Alex Song left Arsenal for Barcelona in the summer following the 11/12 season, Arsenal have been in desperate need of a defensive midfielder. They've been scraping by with Arteta and Flamini, but they're both aging and not getting any better. In the 2014 summer transfer window, fans were screaming out for Arsene Wenger to sign a DM. Stubborn as he is, Wenger didn't sign one.

That decision left Arsenal very thin at the back to start the 14/15 season. Fast forward a few months and injuries, as they always do, hit Arsenal hard. They were forced to recall Francis Coquelin from a loan spell at Charlton.


As the saying goes, he's been like a new signing. In his previous spells with the first team, he failed to establish himself, but this time around, he grabbed hold of the opportunity and surely now Coquelin is the first name on the manager's team sheet.

Late last week, I ran across this article from HITC Sport. In it, Dan Coombs says "Coquelin has shown already that he is playing defensively at an equal or higher level than Song's final season at the club." Dan then provided this simple table:


This table is perfectly fine. It's a table and it provides a great way to look up the data. In the end, the point of the story is to show how effectively Coquelin is playing compared to Song. I created this infographic in Keynote with the hope that it is more interesting, more engaging and makes comparisons easier.

Last week, Saturday Down South posted this table, which shows the number of players over the last decade from each SEC college football team that played in the NFL.  Current stars are noted.


Simple enough, yet so annoying that they aren't sorted in descending order, though it is true that an alphabetical sort is faster for finding a team. Magnitude is also more difficult than necessary to understand in a table.

With a few simple clicks, I created this interactive version in Tableau.


On my version, the bars are ranked and I included the % of total SEC players in the NFL for each team to give the numbers more context.

You can download the data here and the workbook here.

This week on the Football Ramble podcast (listen here), they talked about a very strange stat they have been observing this Premier League season:

So far in the 2014-15 season, Aston Villa has the highest points per goal of any Premier League team in the last 10 seasons.

This was quite intriguing to me, so this morning while I was enjoying a cup of coffee at Tully's I went to ESPNFC.com and downloaded the standings for the last 10 years. I added two columns to this dataset, which you can download here.

1. Points per Goal (PPG)
2. PPG Position (i.e., a rank of each team based on PPG)

From there, I wanted to understand the difference between Villa's actual place in the EPL table and their place in the fictitious points per goal table. For this view, I created a slopegraph, which I first wrote about creating in Tableau here.

If you're unfamiliar with slopegraphs, Andrew Wheeler summarizes them well in his paper "A Critique of Slopegraphs":

Slopegraphs show values for two numeric variables by line segments for each observation by connecting points on two parallel axes. They are frequently recommended for visualizing the changes in ranked data (Bertin, 2011; Tufte, 2001).

I put all of the pieces together in this simple dashboard, which you can download here. Do you notice anything else interesting or unusual?

Yesterday I had the pleasure of meeting Matt Sundquist, Co-Founder and COO of plot.ly. Plotly is a really interesting new data viz product that allows you to quickly build charts directly in your browser. It's built on top of d3, so it has tons of styling capabilities. If you're looking to build charts with d3, but don't want to do much, if any, coding, then you should definitely check it out. They have a free version.

What is Plotly? The folks at import.io have a great summary:

Plotly is a young graphing and analytics startup, which makes it easy to analyze and visualize data together. Using Plot.ly, users can import data from anywhere, then make and share online graphs using Python, MATLAB, R, Excel, Julia, and their web app. Users can also share work and collaborate on projects.

The Plotly API enables users to analyze and visualize data in one place, and forms an important step in building the infrastructure for data science to be further democratized. Their goal is to be an awesome platform for collaboratively analyzing data and making graphs. 

From the Plotly website, in the workspace area (think of this as a worksheet in Tableau):

Easily make graphs with your data in the workspace. Import data from Google Drive or upload Excel or text files. You can apply fits, stats, and functions to your data.  

Every element of a Plotly graph is customizable. Change the colors, annotate, write LaTeX, toggle graph ticks, and much more. It's up to you!

The best way to get familiar with any new product is to start using it. Since I had just publish a parallel coordinates plot in Tableau yesterday, I thought I would use the same dataset and see how long it takes to build something similar in Plotly. I was pleasantly surprised by the simplicity and intuitiveness of the interface. It took me about 30 minutes to build this view, but a lot of that time was spent tweaking the settings to get them just right, which I'm perfectly ok with given the very detailed control I have over the final display.

Plotly doesn't currently support things like filtering and parameters yet, but for a first pass through, I really like it and will definitely be using it more often going forward. So, here's my first plot.ly viz (be sure to hover over dots to see their awesome interactivity):

Reader Brian Bieber (no relation to Justin), a Data Analyst at Vanguard, sent me this question:

I'm currently running some monthly data where a list of records is assigned a KPI indicator like G/Y/R. I've been asked to produce a follow-up piece that would show "movers" from the prior month's data, just in a simple text view. So I guess what I need to try and do is figure out how to do a lookback/compare from prior month data to see who went from Y/R to G on the positive change side and who went from G to Y/R on the negative side.

What Brian didn't realize was that answer to his question was pretty much in the question itself; he needs to use to LOOKUP() function. I sent Brian a solution, but decided to fancy it up a bit more and add some more functionality using parameters:

The steps for building a KPI movers viz like this are pretty straight forward. I'm using the Superstore Sales data set that comes with Tableau in this example.

Step 1: On a PC, right-click drag Order Date to the Columns shelf.  On a Mac, Option+Drag Order Date. Choose the continuous month option.


Step 2: Right-click on the Order Date pill on the Columns shelf and change it to Discrete.


Step 3: Right-click on the Order Date field on the Column shelf and choose Show Missing Values. This is important because it turns on data densification. Months where data does not exist get displayed.

Step 4: Add the dimensions you would like to slice the data by to the Rows shelf. For this example, I've used Region and Order Priority.


Step 5: Create a calculated field called Last and change the Default Table Calculation to Order Date.



Step 6: Right-click on the Last field you just create in the Measures area of the Data window and choose Convert to Discrete.

Step 7: Drag the Last field to the Filters shelf and choose 0. This will change the view to only show the latest date.


Step 8: Create a calculated field to get the value for the most recent month. Notice that I'm using the LOOKUP function. Change the Default Table Calculation to Order Date.


Step 9: Create a calculated field to get the value for the previous month. The only difference here is we need to offset the LOOKUP by 1. Change the Default Table Calculation to Order Date.


Step 10: Create two parameters: one to determine the lower limit of the top KPI (Great Performers) and one to determine the lower limit of the middle KPI (Average Performers).



Step 11: Create the KPI calculations for the most recent month and the previous month. Notice that I'm using the parameters.



Step 12: It's up to you if you want to follow this step. For this specific example, I wanted to see the KPI for this month and last month in the table, so I've added them to the Rows shelf.


Step 13: Create a calculated field that returns a boolean to compare the values for this month to determine if they are a "Mover". I'm using Brian's definition for this example.


Step 14: Drag the Movers field to the Color shelf. I chose to set the colors to a light gray when True and red when False. This way only the Movers stand out.


Step 15: Optional - Change the Marks type to Shape and drag Movers to the Shape shelf as well, then pick shapes that give the affect your looking for. I chose diamonds for the Movers and circles for everything else.


Lastly I did a few cleanup items before placing the worksheet on the dashboard you see at the top of this post:

1. Moved Order Date to the Detail shelf
2. Reduced the level of the row divider
3. Removed the bold text from Region
4. Added a calculated field to the Filters shelf that checks that the value in the Great Performers parameter is larger than the value in the Average Performers parameter.

This method could easily be changed to accommodate many different scenarios like day over day, week over week or year over year.

Download the sample workbook here.