Storytelling with Data Community Exercise — Visualize a Multi-Dimensional Table

In this post, I’ll walk through my solution to the following exercise, visualizing a multi-dimensional table.

You can find the full original Storytelling with Data prompt here: https://community.storytellingwithdata.com/exercises/visualize-a-multi-dimensional-data-table

Original Visual and Prompts:

Prompt: Imagine you’re part of the People Operations team at a mid-sized tech company.

Your team has just conducted the annual Employee Engagement Survey, gathering responses from employees across four departments. Leadership wants to understand how engagement varies by department and tenure, and what other factors may influence these scores.

STEP 1: Brainstorm how you could visualize this data. Create a list of various graphs that could effectively highlight relationships or patterns.

My first instinct is to go from a table at least to a heatmap. This will hopefully make some things stand out, for example if there is a value higher or lower than the rest, or if everything is sort of flat.

Next I’d probably try some bar charts, or dot strip plots to see:

• does 1 department stand out?
• does one tenure group stand out?
• does a salary band stand out?

STEP 2: From the list you’ve made, create at least three different views. (These visuals should be rough drafts, rather than polished, final versions.)

Since the prompt calls for rough drafts, I’ll stick to (mostly) default settings for this part of the exercise. I’ll use Tableau because that’s the tool I like to use to quickly explore data with dragging and dropping.

First up, a heatmap:

Throwing color on the values, it’s a wash of blue but I do spot that HR seems to be lighter than the other departments, and the 5+ year folks in Engineering and Sales seem to have the highest pay, performance, and satisfaction. Both things seem worth digging into.

Next, I’ll try a scatterplot — focusing on engagement + performance scores

I’ve colored by department, and put shape on tenure. Nothing really sticks out to me for the departments, but it does look like the folks that have been around for less than a year have both lower engagement and performance ratings than their colleagues.

For my third view, I’ll ignore tenure and look at all metrics at the department level with a bar chart:

With this view one thing to me sticks out — HR is paid the least, and has the lowest engagement and performance of the departments.

Last view I’ll try out — a dot plot. I’ll put department on color, and have rows be tenure:

STEP 3: Consider each view and evaluate what pattern or insight it shows. How does each one help (or hinder) understanding? Determine which visual(s) you would share with the leadership team and why.

1. The heatmap is barely a step up from the table. While I do think it is an improvement, it does take some time to parse — and I wouldn’t want to spend time explaining it in a meeting or over slack.
2. I like the scatterplot more than I thought I would. It doesn’t include salary in the view, but to me it’s the easiest to spot a pattern of performance related to engagement, which could then lead to those next level questions about salary.
3. The bar chart I think is just okay. It gets the job done, but there’s not much visual interest and I find myself doing a lot of staring at it to compare across the different categories.
4. The dot plot for me is tied with the scatterplot. I like each department represented as its own unit in the canvas, and the whole page isn’t filled up with color like on the bar chart. One downside is for this dataset, both engineering and sales have the same performance rating so that data point gets lost a bit.

My Final Build

I chose to keep the scatterplot, but also include a bar chart showing salaries by tenure. I’ve arranged it on a small slide sized visual, with some key takeaways highlighted. While the scatterplot is my favorite chart for spotting clusters, I think the added context of the salary is important for further discussions. Keeping these charts separate instead of keeping them all in one table helps highlight some insights for discussion.

Closing thoughts

This is my first published viz in over two months — the longest break I’ve had from publishing work since 2021! I recently moved from Florida to Minnesota, and settling in to my new place has meant more time arranging things in the physical world than on the screen. This challenge appealed to me as a great thought exercise — a simple data set, and the practice is relevant to finding and presenting insights at my day job.

If you’d like to crack it open to see how it works, or just bookmark for future reference, you can find my solution on Tableau Public here: https://public.tableau.com/app/profile/brrosenau/viz/StorytellingWithDataVisualizingaMultiDimensionalTable/StorytellingwithDataVisualizingaMultidimensionalTable