Build something interactive
And it doesn't have to be data viz!
Welcome to Data Dash
Compressing an avalanche of thoughts about data into byte-sized chunks. In your inbox every two weeks on Wednesdays.
There’s knowing and there’s feeling
A while back Defense Against Dishonest Charts took Data Bluesky by storm. I recommend reading the piece in full, and one element has stuck with me in the weeks since: interactivity.
I’ve known for far too long that messing with y-axis scaling can make “effects” appear out of nowhere. And there was still something powerful about moving the slider myself.
Back to basics
Making interactive visualizations can be tricky. Luckily, there are also tools that smooth the learning curve from static to interactive plots. Think Altair in Python or ggplotly in R.
One principle to apply from Defense Against Dishonest Charts: don't try to make your interactive plot do everything. One of the reasons those plots worked so well is each interaction focused on one element of how charts can mislead us. Imagine how confusing it would have been if they'd tried to make all their points using a single chart with like thirteen toggles.
Some love for the backend
Building something interactive doesn’t have to involve data viz. One of my more successful data deliverables at a startup involved putting a “guess the result” game before an analytics report. The increased attention on that deliverable reminded me just how powerful a small amount of interactivity can be.
To argue against myself
Most projects don't require an interactive element. Shoehorning something interactive in just because it's fun to build Shiny apps can't always be the move. And there's still a lot of great opportunities to “show don't tell” people important points about your data using thoughtful, interactive elements.
A data thing I liked
Git scraping by Simon Willison
A non-data thing I liked
Delusions of grandeur podcast
(I recommend starting with Heir to the Empire)