Thanks to Knaflic’s Storytelling with Data and Roberto my lecturer, and all of my classmates who provided feedback to make this uncluttered story telling chart using Tableau. A lot was stripped away but I hope the meaning is clear.
Using Tableau to explore performance metrics
Tracy Keys 
Hi Tracy,
It looks like your small problem is beginning to morph itself into a bit of an unwieldy beast. Are you going to use tableau as the tool for data wrangling as well or will you perform this outside of Tableau?
I assume (I know I should not) from your post that the visualisation will be for you explore and understand the data and not for higher management, there seem to be several themes running through the data at the moment, if it is for wider audience it might be worth considering narrowing down the number of themes.
In the spirit of “spitballing” ideas here are some of my contribution for you:
1. I assume that the changes made to the billing system should not impact your customer sign up process ( I assume again) so any quirks like 100% + signup are possible DQ issue. However, it looks like this occurred prior to this period for Mobile and Tablet, is it possible for people to land on the Signup page without going through the funnel?
2. For seasonality, I would suggest a more macro look such as considering the same time period for prior years, as the week ending 26th Feb leads into many public holidays across a number of state and territories rather than looking at just the day of the week.
3. For the final product would you consider a Sankey diagram, I think does a great job of demonstrating the effect of the funnel.
https://community.tableau.com/thread/152115
-r
Tam
LikeLike
Hi Tam. Thanks for this!
I have been looking at Sankey diagrams today and realised my data is not at all in the right format and I have no idea how to convert volumes that land on each page in the conversion funnel into the data like that done in the link you shared above.
It seems as though the data in the link is how many people have travelled an actual path (using eVars or something like that in Adobe) whereas my data is plain old prop data ie visits at a certain point. So I dont think this data will work for Sankey. Is that your understanding of what data is modelled using Sankey too?
LikeLike
Hi Tracy,
Thanks for your post and for managing to deliver your post in a pdf file 🙂
It seems that you have some expertise using Tableau. Good that you are looking for a story to tell for your audience. Just, don’t loose the horizon and don’t forget your target audience when communicating your insights.
As for Tableau, I would like to know more about the challenges of the tool. What would you say to a person who has no experience using that tool and ask you advice about a visualisation tool?
Looking forward to read more of your stories.
Best,
Vanessa
LikeLike
Thank you for your feedback Vanessa.
I definitely did lose sight of an audience, completely. I was just indulging myself really 🙂 So I am going to re-do the charts using the Gestalt principles.
My advice for first time Tableau users is to attend a Tableau Test Drive. They are in the city for about 4 hours and are free! But that said, its a great tool, very flexible and easy to play around with until you find something that works. Like all of these viz tools, understanding what the measures are, and how your data is structured will impact which charts you use. I find that the hardest thing, and I have that problem regardless of the tool Im using!
LikeLike
Hi Tracy,
Clearly a lot of work has gone in creating these graphs.
After reading the post, I wasn’t 100% sure was what question you were trying to solve.
I could see there were 2 questions being answered.
1. Understanding which device converts better?
2. Understanding the different steps between landing page to conversion.
For Question 1 : I think the line graph did the job where it suggested that Desktop was the worst performer consistently.
The heatmap actually shows you too many numbers and colors and there is no way to clearly identify the issue in the conversion rates.
May be just simple bar charts by week by device would also do the trick.
For Question 2 : Once you use line graphs for the funnel, the graphs become so small that you really don’t undertand the trend in them.
My Feedback would be to use Funnel Charts to understand the drop off rates between Step 1 and step 4 of your funnel by device.
Also, if you have the data, it would be interesting to track the conversions by targets/goals.
Then you could use tableau’s forecasting feature to see if the teams are closer to hitting their targets.
Thanks,
Duhita
LikeLike
Thanks for your feedback Duhita. I will try funnel charts, thank you, as it makes more sense than a sankey for the data I have. Do you have any good examples I could read about? And yes, I thought about forecasting too. Its hard to do when there are significant trend breaks though. In truth, my analysis just ended up showing there were errors in the measures in my view ie greater than 100% conversion.
LikeLike
Hi Tracy,
Firstly I think it is so amazing that you were able to learn so much from this tool and at the same time, form a simple and clear data story.
You have mentioned already that the visualisations were created for exploratory purposes so I am assuming that at this stage, the intention or audience is mainly just you.
Having just completed the recommended reading “Storytelling with Data” this morning, I am excited to share a few suggestions from there that could further help with the storytelling/visualisation stage for this data.
As a static visualisation, your main line graph (figure 1) is looking a little like a “spaghetti graph” (i.e. too many coloured and overlapping line graphs). I know that these are interactive in tableau so the other lines can get pushed to the background as you demonstrated (figure 2).
There is a section in the Chapter 9 (p277) where Knaflic introduced some strategies for avoiding the spaghetti graph phenomenon mainly:
1. Plot the lines in the same or muted colour and only emphasise one line at a time (as with your filtered version)
2. Separate the lines either by plotting each line on a separate x or y axis depending on whether you want to focus on magnitude comparison or trend comparison.
You have already started to do that intuitively with figure 3. You only have three categories so it is not as much of an issue as if there were more categories, however I hope that you can gain some helpful feedback. When presenting this information in a set of slides or other static documentation, these tips might help to make it a little easier to follow the main insight.
LikeLike
thank you Corinna! yes you are right. I have read the book now too! I am updating my post with the latest iteration of the charts to illustrate what I have learned. See the PS 🙂
LikeLike