Why Your Data Visualizations are Bad (and how to improve them!)

In this video we walk through some examples of what not to do when creating visualizations.

The themes that come out of the examples are:
Avoid information overload
Don’t be misleading
Make your graphs easy to read

Next we address techniques to fix these common issues. Some strategies to implement include stepping your audience one-by-one through components of a graph, smartly using coloring & opacity to highlight what’s important, and maximizing your data-ink ratio.

Some other tips can be summarized as:
Line charts are good for showing trends over time (but avoid spaghetti graphs!)
Bar charts are best to highlight differences between categorical variables
Generally avoid pie charts, donut charts, and 3d charts

In preparing for this lecture, a few different sources were consulted...

Storytelling with Data (Cole Nussbaumer Knaflic): www.storytellingwithdata.com/books
How to Speak (Professor Patrick Winston):    • How to Speak  

Other resources:

Keith’s YouTube channel: youtube.com/@keithgalli
Reddit (Data is Beautiful): www.reddit.com/r/dataisbeautiful
Reddit (Data is Ugly): www.reddit.com/r/dataisugly

—---
Video timeline:
0:00 - Introduction & video overview
1:24 - 1. Avoid Information Overload
4:42 - 2. Don't be misleading
6:44 - 3. No hard-to-read graphs
8:15 - Fix information overload (step through components one-by-one)
10:30 - Strategic use of color & opacity
12:05 - Maximize the data-ink ratio
14:23 - Being honest with your visualizations
15:42 - Improving hard-to-read visuals (graph selection, z-pattern)
17:55 - Resources used to prepare this lecture
20:07 - Final thoughts!
—---

Free Dataiku Learning Resource:
knowledge.dataiku.com/latest/courses/intro-to-ml/i…

Twitter: twitter.com/dataiku
Instagram: www.instagram.com/dataiku/

From LEARN Media
learnmedia.io/