I know I’ve been a little quiet on the Data Duo the last few months. I have been enjoying Tableau/Alteryx so much at Pluralsight, I haven’t wanted to touch Tableau too much in the evenings. With the feedback/help of our great team, I have built some of the best dashboards of my career in the last few months. Too bad I can’t share any of them. However, I am going to look for opportunities to anonymize/rebuild data sets to share concepts from the real world.
Recently, I was asked to build a series of dashboards (crossing many functions) that would be external facing to a large company. One of the visuals from an existing Powerpoint slide looked similar to the Visitor Funnel shown here. This depicts the flow of users through a set of landing pages, to starting/completing a registration form and ultimately consuming content in an application.
If you’re reading this, I am sure your data viz eyes are bleeding. You probably have your calculator out, furiously typing to see if the numbers are even right. I know you are probably trying to discern if color means anything. Don’t worry, I didn’t forget about the elephant in the room…
I’ll address it now. What the hell is that triangle?
I had a good conversation with the creator, who was very open to feedback. I had this conversation not to poke fun at the resulting visual, but because I was going to rebuild it. It surely will look different and I wanted to educate while exploring his tolerance for something different.
I explained that 11% appears to be 50-55% of the width of the top of the triangle. 33% appears to be about 75% of the width. My brain was too busy trying to not explode to even mention that the triangle made it harder to read the figures, that the icon and triangle were repetitive, etc. etc.
The creator pretty much said the triangle exists to give the perception that it is a marketing funnel. Ok, I can accept that. Moving on…..let’s do something better.
So I did a quick Google search of funnel charts…
It turns out the triangle is not too terrible of an offense…it could have been a rainbow 3D cone. Good Lord.
I started building a bikini chart. I multiplied the measure by -1 so I could use measure values to plot the bar charts.
This was a much better chart. The bars were proportionate to the size of the value at least. I had some good options to label the chart. It was pretty intuitive, but I just thought I could do better.
I thought back to when I was trying to create a tornado chart for the #RevizProject. I thought that might look pretty sweet here. It’s actually not hard to build. Just change the mark type to a line and drop measure values on path. If you keep measure names off the marks card the two lines will connect.
Ok, it’s a funnel, but it still needs some help. I wanted to create tiers within the funnel. Again, it’s not too hard. Make a copy of measure values on columns (it will be on a dual axis later); remove measure values from path and put it on detail; drop the tier discrete dimension on the marks card; and change the tier dimension on rows to continuous. You do have to reverse the axis as well.
Now you can start to see this taking shape. I put the chart on a dual axis and labeled the start of the horizontal lines with just some dashes like a connecting line to the mark.
If you stop here, the rest is pretty much formatting.
I will however talk through some of my decisions in adding complexity. If you want to go a step further, you can download the workbook to look behind the scenes.
At Pluralsight, our whole app is black so black dashboards are totally acceptable. Easy decision made for me.
Each tier represents a count, but a very different event is being captured. I wanted to highlight this using different colors. There were only four so it wasn’t a big deal. I was using a line chart with four different colors so Tableau puts the connecting lines of a continuous palette on a diverging palette. This added some cool purple to my lines, unexpectedly. You will always be surprised when you’re tinkering in Tableau.
I wanted to show additional context at each tier along with percentages of the top tier and drop off rates from the previous tier as you move down the funnel. I decided to add another dual axis to this chart on MIN(-1) and MIN(1). This gave me two more marks cards to use in this chart, which would both be text. So why a dual axis and not a header?
- I wanted to use a highlight action to highlight all the information about a tier without that ghastly yellow color that headers highlight in Tableau.
- I was going to use a parameter to switch calculations. These calculations would have two different formats so I would have two different calculations on the marks card label – can’t do that in a header.
- Using MIN(-1) and MIN(1) on a dual axis gives me two columns of text the same width of the funnel pane. Tableau always ensures the panes in a worksheet are equal width, but this little hack gives me some space back.
I created a reference line using WINDOW_MIN(SUM(-[Number]))*1.6 for two reasons.
- This bought me some room on the left side to label the lines with some larger text.
- Fixing the axis wasn’t an option as the data could change with filters. The entire axis will now change with the data, giving the space needed for the labels on the left and cutting off after the marks on the right.
The juice was worth the squeeze for me.
I did not end up where I thought I would with this chart, but I thought these marketing folks have been talking about funnels since the first day they went to college.
The chart not only looks great to me, but I think it is actually intuitive in displaying the data points that need to be conveyed. That is, as long as they aren’t rainbow 3D cone funnels. So let’s all say it together now…What The Funnel?
Here is a GIF showing the different funnels made with my fake data. Try to not see a funnel, a tornado, an upside-down oil can, a flashlight, an ice cream cone, a bullet, a travel coffee mug or a tumbler….I can’t.
Click on the GIF to see the interactive version.
I hope this well helpful. Thanks for reading.