VizWiz

Data Viz Done Right

January 17, 2018

Workout Wednesday: Rolling Three Month Sales

No comments
Nice challenge from the boys this week. Essentially you have to create a stacked area chart with the stacks adjusting top to bottom depending on which one is the largest. The title also needs to be the same color as stack that's on top. Finally you need a parameter to determine where to split the colors of the view and where to display the values.

There are some other tricks too. I won't give away my solution, but I'm 99% sure it different than theirs. Get the requirements here.

Creating Maps With Linear Geometries in Tableau

No comments
This blog post is a couple months late as Tableau added support for linear geometries in version 10.4. So what are linear geometries? Essentially the are spatial files that are represented as a single line.

For example, let's say you have a series of locations that represent train stops. In Tableau you can draw the route by connecting the dots via the line shelf. This will result in marks for every station. If the route is a linear geometry (or linestring) instead, it is represented in Tableau as a single mark, meaning the viz will load much faster.

This is useful if you each point isn't important and you care more about the path itself. To help me understand how these work I downloaded shapefiles from the US Census, github and Transport for London. Each of these was a linear spatial file already, meaning I could connect with Tableau and go.

If I had a series of points, like the train stops example, I could use Alteryx to convert them to spatial points, create the path and export as a shapefile.

With that, here are a few example I've built using linear geometries and custom Mapbox maps. Note that the maps may be slow to load in Tableau Public. I'm not sure why because they're super fast in Desktop. Enjoy!

January 16, 2018

Makeover Monday: Rank of Income Distribution in the United States | North Dakota vs. The Rest

No comments
Yesterday Sean Miller posted this beautiful viz for Makeover Monday week 3 on the data.world discussion.


So this morning, since I had the Data Schoolers give Power BI a whirl with the Makeover Monday data, I thought I'd try to rebuild his viz in PBI. It took a pretty long time and there were tons of needless hurdles:

  1. I had to manually color each line to grey. I can't use a group or set to color by.
  2. There's no way to use a parameter type of object to highlight a chosen state; thus in my viz I've manually chosen North Dakota to match Sean's viz.
  3. To make the rank start at the top, you have to create a calculation that makes the axis negative; you can't reverse the axis.
  4. I couldn't move the North Dakota line to the front. I found a couple of blog posts that recommended techniques, but none of them worked.
  5. Creating a rank calculation is a monumental pain in the ass.

I persevered and got something built. I don't love it, but I'm learning more each time I use PBI about it's strengths and weaknesses. I still stand by my previous statement that Power BI IS NOT a data analysis tool; it's a dashboarding tool, that's it.

With that, here's my re-make of Sean's viz.

January 15, 2018

Makeover Monday: U.S. Household Income Distribution by State

No comments
The viz this week comes from the Visual Capitalist. I used this data set for the final interviews for DS8 last week; it was quite impressive what people brand new to Tableau can create in just a week.


What works well?

  • The ordering of the states by the highest income level from smallest to largest is easy to see.
  • The colors are all pretty distinct from each other (as I'm not color blind).
  • The colors seem like an intentional choice to go from red (low) to green (high) with steps of color in between.
  • The labels help add context to the stacked bars and they aren't distracting.
  • Including the data source

What could be improved?

  • There's no indication as to which year this represents.
  • As it's not interactive, you can't sort by a different income level, therefore it's not easy to compare different states without reading the labels, which is slow.
  • Are there region differences? I can't tell from this view.

What I did:

  • To be able to see regional trends, I included a tile map.
  • I used bars that represent the change since 2009 so that we could see which income groups grew and shrank the most.
  • I included context in the tooltips that show each income level for 2016 and the change since 2009 for the selected income level.
  • I added some annotations to aid understanding (thanks Eva for the idea).

January 12, 2018

Visualizing 854 Strava Runs

No comments
Before Christmas, Chris Love sent me this tweet to check out:


Uh...HELL YES I want to create this. Chris suggested recreating it in Tableau, which I plan to do next week The Information Lab France. In the meantime, I decided to follow the simple instructions that Marcus posted on github.

The vizzes are built using using rstats and ggplot.

I ran into a few errors at first; upgrading my version of R was all it took to make them go away. The processing is super fast and the outputs are really, really cool. You also have the option to customize the size of the viz. I'm thinking of getting these printed as posters.

Here are the routes of my 854 runs from 2013-2017 as small multiples.


And here are all of the runs I've done around London.


Another really fun couple days learning. Now that I understand how all of this works, it should make prepping the data in Alteryx and creating the viz in Tableau significantly easier.

January 10, 2018

Workout Wednesday: Fiscal Date Running Sum

No comments
With Rody Zakovich and Luke Stanke taking over Workout Wednesday this year (see the new Workout Wednesday website for the challenges), I think we're in for some quite extensive learning this year. I likely won't be able to participate every week; I will when I have time though.

This week, Rody posted a challenge that required calculating the running total of sales based on a parameter that determines the fiscal month.

My Thoughts

  • Fiscal dates suck; they are way too hard to work with in Tableau (this was my first time ever using them)
  • The calculations are quite straightforward. I figure out what I needed to do by building a table of dates first, then creating the calcs.
  • Getting the x-axis formatting correct took me the most time.
  • It's way more fun issuing the pain of these challenges than it is to receive the pain.

Overall, a fun challenge and I hope to never see fiscal dates again. Click on the image below for the interactive version.


January 8, 2018

Makeover Monday: Using Gauges in Highcharts to Understand Characteristic Preferences In Romantic Partners

2 comments
Have I finally found a use case for gauge charts? That might be up for debate. The purpose of this viz was to practice Highcharts. I thought the gauges, when aligned as I've done them, are easy to compare.

Click here for the interactive version, which doesn't really do much other than animate the filling of the gauges.

Makeover Monday: Which characteristic preferences are most different with British men and women?

No comments
I'm giving Datawrapper another try this week. This time, I imported only the data for Brits and put Women and Men in separate columns. I chose the "Range Plot" option, which we often call a barbell chart or a DNA chart and got this really nice chart out of the box.



What I really like is how Datawrapper automatically colors the difference between the two dots. Really slick!

January 7, 2018

Makeover Monday: What characteristics are most important to British men and women?

11 comments
Makeover Monday week 2 represents my 600th viz on my Tableau Public profile. I'm not saying this to brag; I'm hoping it shows readers of this blog the dedication I put into my craft. I preach to people all the time that the only way to get better is to practice. That's what I do...every single day. I want to learn something new every day. I encourage you to do the same.

For week 2, we looked at this viz from YouGov.


What works well?

  • The title and subtitle that make it clear what the viz is about.
  • Splitting the view up between men and women keeps it from getting too busy.
  • You can easily look up any value.
  • The colors are easy to distinguish from each other.
  • The colors are in the same order for each row.
  • Including the survey dates in the footer.

What could be improved?

  • Making comparisons between men and women takes longer than necessary.
  • Repeating the word "ranked" on each row is unnecessary.
  • While you can easily look values up, your eyes have to go back and forth to the legend.
  • The legends could be reworded to be shorter. For example, change "They have a personality I like" to "Personality".

What I did

I struggled with this data set. I must have spent 2-3 hours trying to find something insightful, trying different chart types, etc. Nothing was working for me. When I get stuck like this, I like to turn to Google images. I searched for "looks vs. personality charts" to hopefully find some inspiration.

I clicked on one image and then some of the related images. This image in particular stuck out to me:


Oh maybe this is something to work with, but wait, I could swear I've seen this somewhere before. Turns out Andy Cotgreave created this for Makeover Monday week 4 back in 2016. How ironic!

So I set out to do something similar. Basically I wanted to take the original stacked bars, keep them separated by gender, and then create BANs and the units chart for each personality characteristic as Andy has done.

With that, here's is my viz.


January 4, 2018

Makeover Monday (Power BI Edition): Poultry and Livestock Consumption

8 comments
This week I've dedicated time to learning new tools. Earlier this week I posted vizzes created in Google Data Studio, Datawrapper and Infogram. Tableau's main competitor is Power BI, so this morning I downloaded and installed PBI Desktop, connected to the data via the data.world connector, and recreated my Tableau viz.

My First Experience With Power BI

Keeping in mind that I knew in advance exactly the chart I wanted to build, here are my initial impressions and experiences with Power BI -

  • It took me a few hours to create the viz you'll see below. Since I've never used PBI before, I felt that the amount of time spent wasn't too bad. It took me about 30 minutes to create my first Tableau dashboard back in 2008.
  • Googling for solutions was very helpful; there seems to be plenty of documentation about how to use Power BI.
  • Formatting is very tedious, but you have a ton of control.
  • You can get pixel perfect reports by choosing NOT to snap to the grid. I used snap to grid to align everything quickly, then switched to the non-snap option to move things around a few pixels. I found this quite simple.
  • I didn't have to pivot the data to get the view I wanted. Pivoting the data made creating the chart in Tableau easier.
  • Creating calculations is pretty straightforward if you're used to Excel. The only confusing bit for me was that you have to create a new column, not a new measure. That didn't make sense, but once I knew what to do, it wasn't a problem.
  • Adding dots to the ends of lines required a separate calculated field (or column in PBI terminology).
  • The tooltips look great out of the box and are super responsive, although I didn't see any way to control what appears in the tooltip. For example, if you hover over the dot on the end of the line, you'll see the extra calcs I had to create for the dots. It would be nice to be able to hide those from the tooltip.
  • Formatting fields is done through the Data pane. When you click on the spreadsheet icon you get a data preview, then you click the column header and format it.
  • You can't create custom number formats, which meant I couldn't put a + before the positive percentages.
  • Embedding and sharing to the web takes too many steps. From PBI Desktop you have to Publish to Power BI, then in the browser, you have to Publish to Web. Seems like an extra step to me.
  • I couldn't control the line thickness of the reference line at zero, but it was great to be able to move the reference line to the back (you can't move a reference line to the back with Tableau).
  • Creating the equivalent of a table calc is pretty simple. For example, I used the LOOKUP function to create the change versus 1965. Again, if you're familiar with Excel formulas that should be easy to pick up.
  • I couldn't add a reference line on a continuous date axis.

Overall, I found this a really useful exercise and enjoyed learning Power BI. It's very helpful for me to understand the pros and cons of Tableau and Power BI and helps me appreciate what both tools offer and how they approach reporting.

The biggest advantage for Tableau is how quickly you get into the flow of visual analytics. Power BI, as I see it so far, is a reporting tool, not a data analysis tool. That's a massive differentiator for me and truly sets Tableau apart.

My plan is to use PBI for a few more Makeover Mondays. Next time, I think I'll start with PBI then go to Tableau and see how I do not knowing exactly what chart I want to build ahead of time.

With that, here's my first Power BI report. I'd love to know your thoughts.

January 3, 2018

Workout Wednesday: Looks vs. Personality

8 comments
This is my last Workout Wednesday challenge for you. Next week, Rody Zakovich will be taking the reins. If you're not familiar with his work, check out his blog. Thank you Rody for continuing this project for the Community!!

Ok, so for week 53, you need to recreate this viz, which is a remake of a viz from YouGov and one which I wrote about before.


Requirements

  • Get the data from data.world
  • I used the Raleway font, but if you can't use it, don't worry about it.
  • The dot plot must show a close circle for men and an open circle (filled white) for women.
  • There should be two annotations that point to three places.
  • The three places where the annotations points to must show the dots connected.
  • The dashboard is 600x700.
  • Match the title, legend and footer.
  • Match the tooltips.
  • There must be solid vertical lines for 0 and 100 and a dashed line at 50 with labels for each at the top of the viz.
  • There must be white vertical gridlines every 10%.
  • There must be white space between each nationality.

I think that's everything. If not, let me know. Post a link on Twitter and tag @VizWizBI, @EmmaWhyte and @RodyZakovich. Thanks for all of your participation this year and I look forward to seeing how Rody take it forward.

Good luck!

January 2, 2018

Makeover Monday: U.S. Poultry and Livestock Consumption (Alternative Tools Edition)

No comments
I think we should all be exploring as many tools as possible when practicing data visualization. Makeover Monday provides the perfect platform for practicing, so I'm going to do my best to lead by example.

Yes, I love Tableau and it's my favorite, but learning how other tools work, how easy or difficult they are to use, and whether they allow visual analytics, helps me to understand the marketplace and be a more informed data visualizer.

Given this week's data set is so simple, I decided to give a few other tools a try. Below you'll find several versions of essentially the same chart I created in Tableau using other tools. Are these tools perfect? No, but no tool is; they get the job done though.

Google Data Studio

Datawrapper


Infogram


January 1, 2018

Makeover Monday: How Has U.S. Consumption of Poultry and Livestock Changed since 1965?

3 comments
Welcome to another year of Makeover Monday! This year, we're partnering with data.world to host our data, be a platform for discussion and to welcome other tools to the project.

For week 1, we're looking at this chart from The Atlas about American poultry and livestock consumption. The data comes from the National Chicken Council (yes, there is such a thing) ; note that the source data is slightly different from the chart as the data is more up-to-date.


What works well?

  • Using distinct colors
  • Line chart shows the consumption trend well, particularly the increase in chicken consumption
  • Including the word "pounds" on the top axis label
  • Displaying every 5 years on the x-axis
  • Including a solid line at zero
  • Using light gridlines

What could be improved?

  • There's no interactivity, making it impossible to know the exact values
  • Labeling the start and end of the lines would help give the change more context.

My Goals

  • Since the original chart is pretty good as a line chart, stick with lines.
  • Focus on the change since 1965.
  • Include big labels on the ends of the lines
  • Use colors that associate more with the type of meat
  • Remove as many gridlines and axis lines as possible to move the focus even more to the lines

With those goals in mind, here is my first Makeover Monday of 2018. Let's make this an amazing year! And remember to join the conversation on data.world, where you can embed your visualisations live (or simply post an image with a link).

December 27, 2017

Workout Wednesday: UK's Favorite Christmas Chocolates

No comments
Week 52 of Workout Wednesday is upon us. I do have one more challenge ready for next week, then Emma and I plan to turn the reins over to someone else (announcement coming). We're both kind of out of ideas and inspiration; some fresh blood will help keep this going and make it more valuable to the Community.

This week, Emma set out a challenge for us to viz the UK's favorite Christmas chocolates in a single sheet.

  1. Remake the dashboard in one sheet / view.
  2. Copy the colours - the favourite chocolate should be orange, the least favourite dark grey, and the rest light grey. You CANNOT use manual colours based on the chocolate name - it has to be based on the ranking of the chocolates in each brand.
  3. Label the bars with the chocolate name and the percent of people who picked it as their favourite.
  4. Match all the titles, tooltips and labelling exactly

Pretty simple set of requirements. For requirement 2, I came up with a different solution because there are two least favorites for Cadbury Roses. I chose to color both of the least favorites the dark grey whereas Emma only colored the last bar. I think mine is more accurate.

Click on the image below for the interactive version. One week to go!

December 24, 2017

Makeover Monday: Americans Favor Fake Christmas Trees More and More

No comments
And with this post, another incredible year of Makeover Monday is done and dusted. My expectations have been ridiculously exceeded and I can't wait to see how we break down even more barriers and get even more people involved next year.

Here's my challenge for everyone that participates...Get one new person to participate. That's it! Simple! Do it! Everyone will benefit.

For week 52, Eva chose a simple viz from Statista about real vs. fake Christmas tree purchases by Americans. We switched to fake about eight or nine years ago and haven't regretted it one bit. Every year we think about that one tree we have saved.

Statistic: Christmas trees sold in the United States from 2004 to 2016 (in millions) | Statista

What works well?

  • Nice interactivity on the tooltips
  • Colors are easy to distinguish
  • The tops of the stacked bars allow you to see the overall trend of all tree purchases

What could be improved?

  • Including all of the labels on the bar is distracting
  • Remove the shadows from the text
  • Use colors that are associated better with trees (like green)
  • Use a more impactful title and the wrapping is sloppy
  • Use a smaller footer that won't take up way so much space

My Goals

  • Complete something quick; it's Christmas Eve after all
  • Find something interesting in the data
  • Use Christmasy colors
  • Use a title that tells the user what they're seeing

With those goals in mind, here is my last Makeover Monday for 2017. See you next year!

December 20, 2017

Workout Wednesday: State by State Profit Ratio

6 comments
This is the last Workout Wednesday from me this year. Emma still has one more next week. As for next year, we haven't decided what to do yet.

For today's challenge, you'll need two data sets: Superstore Sales (the one we always use for WW) and the hex map template.

Requirements

  1. You are NOT allowed to use blending.
  2. Create a small multiple hex map showing profit ratio by state.
  3. Color the hexagons using the Sunrise-Sunset Diverging color palette
  4. Match the tooltips on all of the charts
  5. Include a summary heatmap by year and month at the bottom of the map
  6. Everything in the dashboard needs to be tiled
  7. Ensure there is no gap between the hex map and the heatmap. The column divider lines must match up.
  8. Include a sparkline of profit ratio by quarter with labels on the ends of the lines.
  9. Ensure there is enough space for the labels on the sparkline, but you CANNOT manually fix the axis. Make it dynamic via calculations.
  10. Clicking on a state should update the sparkline, update the heatmap, and highlight the state in the hex map. Mississippi  and West Virginia are good test cases to ensure it's working.
  11. The dashboard should be 700x700.

That's it! This is one of the easiest challenges this year from me. It's a good way to bring together a lot of what we've learned this year.

Good luck!