Canonicalized

  • Tableau Sidekick
  • Blog
  • Contact

Interactive Highlight Tables/Heatmaps with Tableau Set Actions & Transparent Sheets

Dorian Banutoiu
Updated on: February 27, 2019 — Reading time: 5 minutes 12 Comments

In this article, I am going to pick apart and try to improve on a prevalent chart type used to analyze cohort retention: the impact plot (also called triangle chart).

It is basically a highlight table, or a heat map with the numbers displayed.

Here are a couple of examples:

impact plot

cohorts google analytics

This graph type that is often difficult to understand. Especially by people who are seeing it for the first time in their lives.

Time is present on both axes. In these examples, each row represents one cohort (by acquisition time frame), and the columns show the percentage of users that are retained X days/months after they have been acquired.

I believe the primary purpose of this type of chart is to show hotspots. To direct our attention towards areas with higher than average retention rates. In some cases, these are easy to notice, while in other cases not so much.

Quantity encoded colors are not so straightforward to compare. Displaying the actual values doesn’t solve the visual problem, it just shows some numbers.

My Tableau approach

I took a stab at building one of these impact plots in Tableau, and this is how it looks like:

impact-plot-tableau

I made a couple of adjustments to the traditional way of display:

  • added bars to be able to compare the number of users in each cohort visually; used a light grey color for them since this part of the story is not the focus of our analysis
  • dimmed down the contrast on the number colors; I believe the exact numbers shouldn’t be our focus either
  • added arrows to highlight big drops from one month to another; these are adjustable by the user
  • used a diverging 3-color palette for the heat map; I think it shows the hotspots better than a standard one-color gradient would:

one-color-gradient

I know, the changes don’t make a huge difference. It’s pretty much the same chart with some sprinkles.

But hold on! The fun is just starting!

Adding Interactivity with Tableau Set Actions

Let’s take a moment and think about how we should read this type of chart.

The first direction is vertical

Our eyes move this way when comparing the retention rates in a particular week across cohorts:

read-cohort-vertically

Some points I tried to improve by adding interactivity:

  • focus on one week at a time
  • look at a more accurate evolution across cohorts for a particular week
  • compare the trend and a specific value to the weekly average

Here’s what I came up with:

Notice how it’s highlighting the current cohort with a dot in the sparkline beneath each week.

Update

Shortly after publishing this post I got a comment which I really dig! Both the comment and the fact that people are sharing their opinions publicly.

Sparking up discussions will only lead to progress, and I encourage you to speak your mind whenever you get the chance!

As far as the comment goes, Daniel pointed out two possible problems he found with my approach.

First, cohorts are discrete variables which cannot be connected with lines as I did with the sparklines that are showing up beneath the table.

You should (almost) never connect discrete variables with lines or areas. But in our case, the cohorts are represented by measures of time (week of acquisition). Which is why I consider my approach acceptable.

If the cohorts would be defined by discrete categorical groups such as user location, acquisition channel, etc. the cohorts values would be better represented with bars or dots.

Here’s how the interactivity would look like with tiny bar charts:

Notice that the bar charts are vertically oriented. This was another thing pointed out in the comment: since the cohorts are vertically aligned in the highlight table, horizontally reading their evolution might arise problems for end-user.

I usually read time-series horizontally, so for me, it works. But it’s also acceptable to visualize them vertically.

We can also read it horizontally when we want to follow the path of a specific cohort.

horizontal-read

Interactivity to the rescue once again!

We can focus on each row at a time using a highlight action. And we have some perspective with the line chart popping up on the right bottom corner.

The big drops are pointed out in the line chart as well to draw attention to potential problems.

How to pointers

You can always download the workbook from Tableau Public to see how it works, but I want to share with you some ideas.

1. Set Actions are the primary driver of this dashboard’s interactivity. Although I have used a basic highlight on the table, the lines that show up are powered by set actions:

  • adding color to the sparklines
  • highlighting the hovered cohort in the right-side line chart
  • displaying a dot for the week in focus

If you want to stay current, I recommend having a look at Lindsey Poulter’s blog. She’s been on a high-kick lately showing us how to make use of Set Actions!

2. Transparent sheets

By disabling the highlight table’s background, I was able to show the sparklines at the bottom of every column.

As you can see, this wouldn’t have been possible without excluding the null values from the table:

impact-plot-null-values

I was able to hide the nulls using a table calculation filter:

impact-plot-exclude-nulls

For the sparklines I had to create a new sheet and align it “perfectly” behind the cohort table on the dashboard:

sparklines

align-transparent-sheets

I colored all of the lines white and displayed only the one in focus with some set actions magic.

If we were to use a regular heatmap, we wouldn’t need a transparent background for it.

3. Filtering entire sheets

As you might have noticed, the line chart on the right only shows up when you hover on a bar item:

show-sheets-filter

I pulled this off with a filter action set up like so:

filter-whole-sheet

I used a dummy field that doesn’t affect the sheets in any way. We just need the selection to clear when we move our mouse away from the bars.

4. Other ways to approach the interactivity

I feel quite comfortable with how this dashboard works.

I would be interested to see how it would look if we were to highlight horizontally and vertically at the same time. Matt Chambers has recently written a guide on how to do it.

This is my challenge to you, Tableau enthusiast!

If you manage to do it or if you have other ideas to improve, ping me in the comments, on the live chat or on social media!

Context

They say that context is king nowadays. I believe that statement is very close to the truth.

It’s quite challenging to execute in real-life situations, but trying to put matters into perspective can help our visualizations a great deal.

I also believe that every audience has their preference. I feel obligated to point out the available options for you to choose from.

I hope I did a decent job with that!

Dorian Banutoiu
Dorian Banutoiu

Highly passionate about data, analysis, visualization, and everything that helps people make informed decisions.
I love what I do! I am working to improve speed in every aspect of my life and that of our clients.
I find comfort in helping people, so if you have a question, give me a shout!

More about: Data Visualization

Tableau Case Study: Chaoly unleashes Chinese Social & eCommerce Analytics

A look behind the scenes at how we helped Chaoly use Tableau to bring data to life.

Juicy details about goals, challenges, and factors that have contributed to success.

We share the full list of dashboards that reached the clients’ screens.

Comments

  1. Daniel Sydnes says

    November 28, 2018 at 8:10 pm

    I encountered some cognitive dissonance with your inclusion of sparklines.

    Lines suggest values over a continuous range. By extension, sparklines typically convey variation over time (e.g., pulse, temperature, stock price). Edward Tufte introduced them as “intense continuous time-series” in 1983. Your use here bucks those conventions — tracking values over discrete non-time buckets (cohort groups). Data bars are a better fit for non-continuous series.

    Secondly, the values used in the sparklines are derived by reading the impact plot vertically (top to bottom), but the sparklines show those same values horizontally (left to right). Imagine the impact plot as a topographical map. We would expect a slice through the vertical dimension to show the height of those values in the same plane as the cut.

    Reply
    • Dorian Banutoiu says

      November 28, 2018 at 9:28 pm

      Hey Daniel,

      Thanks for stepping in! I need some time to process your thoughts.

      My idea is that even though the cohorts are indeed discrete values, in this particular case they are represented by measures of time, and they are sorted chronologically. Which I believe is the case in a lot of real-life situations.
      Regarding the horizontal direction of the sparklines, it’s what felt natural to me. But they might not be so easy to understand for other people.

      I am working on some iterations to try a vertical direction for the lines and to replace them with bars. I’ll add these as alternatives in the article so that the audience can make their pick

      Reply
  2. Chris Tan Xian Wei says

    July 25, 2019 at 10:57 am

    Hi Dorian Banutoiu

    Your post is impressive! The dashboard is very comprehensive and interactive.
    I am quite new to Tableau, may I know how do you edit the layout of multiple line graphs? https://uploads.disquscdn.com/images/3510af4055661bdd2a3ee9c4de5eef87d5b673207c76c0750d42c311b0938c9c.png

    I can only make some things like this
    https://uploads.disquscdn.com/images/5300a9fdd7cf758c7faf01fb1ce34a7bd10cc72a04b84894d2f160f8444263cc.png

    Much appreciated if you can help me. Not sure if you still noticed this after 8 months

    Reply
    • Dorian Banutoiu says

      July 25, 2019 at 10:57 am

      Just add a discrete dimension on the column shelf to obtain a small multiple layout

      Reply
      • Chris Tan Xian Wei says

        July 26, 2019 at 7:43 am

        I think you have misunderstood my question. I mean how do you align line graph separately to different places ? https://uploads.disquscdn.com/images/341fe2f9b8df87efc78c466fe94406a5995c10db3758334ad6b49d4a406b4f97.png

        Reply
  3. jeisey says

    September 9, 2019 at 4:20 pm

    Downloaded your workbook but it will not open (using Tableau Desktop 2018.2)

    Reply
    • Dorian Banutoiu says

      September 9, 2019 at 4:20 pm

      You need Tableau 2018.3 to be able to use Set Actions

      Reply
      • jeisey says

        September 9, 2019 at 7:38 pm

        Got it, thanks Dorian!

        Reply
  4. Jeff Moltenberry says

    November 6, 2019 at 6:22 am

    Hi, this is a great looking viz. Would you mind sharing your diverging color scale? That doesn’t appear to be part of what’s in Tableau public workbook.

    Reply
    • Jeff Moltenberry says

      November 6, 2019 at 6:22 am

      Nevermind, found viridis 20

      Reply
  5. Rafael Santos says

    August 5, 2020 at 12:21 am

    Hi Dorian, What is the name of the color pallete you used on the last graphs, can you share it?

    Reply
    • Dorian Banutoiu says

      August 5, 2020 at 10:19 am

      Viridis

      #440154FF #481567FF #482677FF #453781FF #404788FF #39568CFF #33638DFF #2D708EFF #287D8EFF #238A8DFF #1F968BFF #20A387FF #29AF7FFF #3CBB75FF #55C667FF #73D055FF #95D840FF #B8DE29FF #DCE319FF #FDE725FF #440154FF #481567FF #482677FF #453781FF #404788FF #39568CFF #33638DFF #2D708EFF #287D8EFF #238A8DFF #1F968BFF #20A387FF #29AF7FFF #3CBB75FF #55C667FF #73D055FF #95D840FF #B8DE29FF #DCE319FF #FDE725FF

      Reply

Leave a Reply Cancel reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.

Get in touch!

Drop us a line ⇣    or    Schedule a chat


    We'll get back to you in less than 24 hours

    Tableau Partner
    canonicalized

    Latest articles

    • MDS in a Box – An Open Source Modern Data Stack in Dorian’s Way
    • Mobilize Your Data: Creating Mobile-Friendly Dashboards in Tableau
    • Migrating from Tableau to PowerBI? Here’s what you might need to consider

    Insiders only

    Copyright © 2025 · canonicalized.com · Contact · Tableau Dashboard Examples

    Powered by Ideas · Governed by Data

    This site uses cookies for a better viewing experience. By continuing navigation, you agree to the use of this info. Learn more about cookies and how to decline or change your settingsOkRead our Privacy Policy