VizWiz

Launch, grow, and unlock your career in data

June 26, 2018

Makeover Monday: Where are London's happiest bike pickup zones?

, , , , , , , No comments
While I was running to work today along the Thames toe path, I had an idea! I wanted to create a hex map or grid map or some type of grouping of stations to represent the stations in clusters rather than showing every station on a map.

I created a viz last year about American happiness, so decided to use a similar theme. What I did was group stations together based on their location. It takes two calculations:


You then makes the continuous dimensions and place them on the appropriate shelves (Round Lon on Columns and Round Lat on Rows).

I then created a calculation that ranks each "zone" by the number cycle hires and then places them into percentiles. I then take the percentiles and break them up into happiness quartiles.


I set the Location Happiness to discrete, placed it on the Shapes shelf and applied my emoticon shapes. I then duplicated the Round Lat field on the Rows shelf and moved the Location Happiness field to color, changed the shape to circle, moved the marks to the back and assigned colors.

Simple! I like how this turned out.

June 25, 2018

Makeover Monday: When are bicycles hired in London?

, , , , , , No comments
The weather has been wonderful here in London lately, with really warm temperatures expected later this week. I take advantage of the weather by cycling to work whenever I can. It not only helps the environment, it's good for my health, it's good for my mind, and cycling commuting saves me time and money.

I asked Eva to use data from Transport for London's open API about their cycle hire scheme. Data is available back to 2012 and I offered to prep it for her and upload it to Exasol...all 50M+ bike hires worth. I love the weeks when we get to use Exasol because I can ask and answer questions on massive data sets without any performance constraints.

The visualization to makeover this week comes from Sophie Sparks:


What works well?

  • The small multiple layout works great or showing cyclical patterns (see what I did there?).
  • The diverging color scale helps accentuate the peak periods.
  • The shading under the lines makes the viz feel more full and complete.
  • Shading the weekends helps separate them from the rest of the weekdays.
  • Putting the word When in red in the title to match the peak period.

What could be improved?

  • I would remove the section at the top that says "Boris Bikes" and the image.
  • Include some sort of insight as a subtitle.
  • There's no indication of what the y-axis means. I assume it's the number of bikes hired, but it could just as easily be something else.

What I did

  • First, I rebuilt Sophie's viz because I like it.
  • I wanted to focus on the weekday and hourly patterns in the data.
  • Use the TFL blue as a single color for the viz.
  • Provide some interactivity so that people could see when the peaks and troughs in the data are for a specific year or month.

Click on the image for the interactive version.

June 19, 2018

Tableau Tip Tuesday: How to sort first by the most positive values, then by the most negative values in a single chart

, , , , No comments
It was day 1 of Tableau training for DS9 at The Data School yesterday and we were practicing different sorting methods, including sorting by a discrete measure. I was then asked how we could sort positive values from highest to lowest followed by the negative values from lowest to highest. This helps emphasize the best and worst performers. The trick was doing this in a single chart.

Enjoy!

June 18, 2018

Makeover Monday: U.S. Influenza Surveillance Report

, , , , , , No comments
I've seen lots of headlines in the US since March about how bad the flu outbreak was this year, so I did a bit of googling and found this week's Makeover Monday topic. The chart we are to makeover comes from the CDC.



What works well?


  • Clearly marking the x-axis so that it's evident that the weeks don't start at the beginning of the year
  • Including the national baseline for context
  • Chart dimensions scaled properly
  • Using red for the most recent season so that it stands out more

What could be improved?

  • The colors are too bright and are competing for attention.
  • The symbols on the 2017-18 line are unnecessary.
  • The start of the 2009-10 season is wrong, according to the data that can be downloaded.
  • The national baseline should be weekly, not flat across the time period.

What I did

  • I liked the idea behind the original chart, so I kept that but made it look nicer and more focused.
  • I included a summary to set the context for the line chart.
  • I included the national average by week for context.
  • Use a stepped chart to make the weekly change easier to see.
  • Focus the lines on the two outlier periods.

June 15, 2018

Tableau Prep Tip: Returning the First and Second Purchase Dates

, , , , , No comments
If you don't participate in Workout Wednesday, you're missing a great opportunity to learn. For week 24 2018, Ann Jackson challenged us to also use Tableau Prep. The toughest part of the data prep was getting the second purchase date for a customer. First is easy. The trickiest part is that you can't sort data in Prep, so you have to do some workarounds to get what you need.

In this video, I should you how I approached returning the first and second purchase dates for a customer, include some summary measures, then bring them both back together into a single table for visualizing in Tableau.

June 13, 2018

Workout Wednesday: Do Customers Spend More on Their First or Second Purchase?

, , , , , , No comments
Ann is back for Workout Wednesday week 24. The full list of requirements can be found here. Personally, I like how this week's challenge included Tableau Prep. It's great to have an excuse to practice!

The high-level requirements:

  1. Create a data set in Tableau Prep that returns the first and second order for each customer along with the sales, number of categories and number of products sold on that day. 
  2. The data must be wide rather than tall. That is, you must have nine columns in total: Customer ID, two dates, two sales totals, two category counts, two product counts.
  3. Create a dashboard with a scatterplot and two strip plots.
  4. Float everything...YUCK!

Here's what my flow looks like from Prep:
I intentionally did NOT rename my tools so that you wouldn't know exactly what I did. You can see my final output below the flow.

Building the viz was pretty simple. Scaling the axes the same is something I do a lot, but I do expect that to trip up people. Creating the 45º line is something I've written a tip on before, but Ann's has a twist as it has to be behind the dots. Sneaky!

The sucky part was floating everything. I started by tiling everything, literally writing down the position and size for each element. Then I floated them one by one and entered what I wrote down to put them back in their proper position. From there, it was a little bit of tweaking to move the axes closer together.

Nice challenge. It wasn't overly complex and required me to reach back into the memory bank.

June 10, 2018

Makeover Monday: Tourism Density Index

, , , , , , , No comments
For week 24, Eva presented us with something called the tourism density index, which basically means how many tourists come into a country compared to that country's population. Here's the original viz:


What works well?

  • Really good explanations for how they define overtourism and undertourism and examples for each
  • Providing the exact figures for each country
  • Colors are easy enough to distinguish
  • Sorting the countries from lowest to highest
  • Splitting the view between the highest 9 and the lowest 9

What could be improved?

  • Circles are inherently difficult for comparisons. Are they measure by area or diameter? Either way, the circle in a circle in overkill.
  • Why does the size of the light green circle change once the dark green circle is a larger value? That makes no sense at all.
  • If the exact numbers were not included, it would be impossible to compare countries.
  • Why show the top 9? That seems like an unusual way to select the countries.

My Goals

  • Focus on either the raw values or the percentages. I'll figure this out once I explore the data.
  • Make it easier to compare countries.


June 7, 2018

Workout Wednesday: How does sales compare in the Current Period to the Previous?

, , , , , , , No comments
It's been eight weeks since I've done Workout Wednesday. Sometimes you have to reprioritize things to get other things done. For me, WW was something I could cut out to free up more time for finishing the Makeover Monday book (pre-order here).

But I'm back and this week Rody gave us this challenge. Read all of the requirements here.

I had an idea straight away how to do this and in all it took about 30 minutes. The date offsetting took some tinkering, but the rest was pretty easy. I'm glad Rody is back from his hiatus too. His challenges aren't as brutal as Ann's.

Click on the image for the interactive version.

June 5, 2018

Tableau Tip Tuesday: Split, Pivoting and Union with Tableau Prep

, , , , , , No comments
Now that the manuscript of the Makeover Monday book (pre-order here) has been sent to the publisher, I'm going to do my best to get back into a rhythm with weekly tips. This week, I'm going to show you how I used Tableau Prep to prepare the data for my Makeover Monday viz this week. It involves splitting eight columns into two sets of four, renaming fields, pivoting the data, and then unioning it all back together.

You can download the flow here.

June 4, 2018

Makeover Monday: The UK Gender Pay Gap Across Salary Bands

, , , , , , , , , , No comments
This week's data comes from the UK government and more specifically the Valuation Office Agency. I was alerted of this data set by Aisling Roberts, who had written a great article on LinkedIn that questions whether people will actually take any actions based on the data.

Let's start with this viz from the official report:


What works well?

  • The symbols make it clear this about females and males.
  • The BAN in the middle tells us what the bonus pay gap is.

What could be improved?

  • Both icons are filled to the same level, making it look like there is no bonus pay gap. These should be filled to the actual values for each gender.
  • The icons don't add much value.
  • The title could tell us a whole lot more.
  • There's no source listed nor no timeframe.
  • The gridlines aren't evenly spaced between 0% and 50%.

I must admit, this is a tough data set. Hopefully the explanations I wrote on data.world provide sufficient context. I found it most useful to look at a specific company and look those values up in the data provided to ensure I understood what it means. Given that I found the data overwhelming, I decided to focus on the pay bands since that's what Aisling focused on in her article. 

From there, I started to build lots of charts, but found the number of companies overwhelming. Therefore, I decide to limit the data to those companies located in the City of London (i.e., those with a postcode that starts with EC). I also knew I need to do some data prep so that I could compare females and males in each pay band more easily. I turned to Tableau Prep for this.

The flow works like this:

  1. Remove columns that aren't needed
  2. Splitting the data up into two streams, one for the female columns and one for the male columns.
  3. Pivot the data so that the pay bands are listed down instead of across
  4. Add a column for the gender
  5. Union the data back together
  6. Export to an extract

Pretty straightforward and this short amount spent prepping the data made the gender comparison significantly easier. I first wanted to understand how the median proportion of females and males in each pay band by the size of the company with a City of London total (NOTE: the total only represents companies that reported).

Click on the image for the interactive version.


This simple view makes it incredibly evident that the proportion of females declines as the pay band increases. Males would be the inverse. It's particularly stark in the largest organizations. In the City of London, there are only three employers in that range (British Telecom, Royal Mail, and Sainsbury's Supermarket).

The heat map helped give me an overview of the data and felt ready to create something more detailed. This time I wanted to look at all companies together by gender by pay band compared to the overall median for each gender. I also wanted to provide the user with the option to choose a specific company. When they do, that company gets highlighted.

Click on the image for the interactive version.

What first struck me in this view is the clear, overwhelming patterns down and to the right for women. This gave me a great impression for how big the gender pay gap problem is.

The gender pay gap is not a myth. These are facts, facts that show women are underrepresented at higher salary levels. Don't let this discussion get lost. Check out your own company. How are they performing? Ask them to share the data within your organization. Transparency is a key to fixing this discrepancy.
Subscribe to: Posts ( Atom )