A fun before and after for me was when trying to explore correlations over samples with a lot of features. In particular, on survey results on why teams adopt different programming languages:

V1: Correlations as an interactive heatmap --

http://lmeyerov.github.io/projects/socioplt/viz/rank.html

V2: Correlations as an interactive graph --

https://labs.graphistry.com/graph/graph.html?dataset=PyGraph...

Nowadays when we're dealing with wide event data (security, fraud, operations, etc. investigations), even more fun!