One of the things that stuck with me most from my BI classes was a lecture on faithful intuitive representation of data, for instance the distinction between percentages and percentage points, choosing the right charts for certain figures, why some people use logarithmic charts and why they are bad.

He also brought up the usage of y-offsets, the thing mentioned in the post. He showed similar (but a little less dramatic) charts of significant changes that looked a lot less impressive once you "zoomed out" and set the lower bound to zero.

He then made it very clear that he did not want to see any of us do that in the exam.

I encountered it later at work, where someone had generated a chart of some quality measure that had a manager perk up because it looked unstable and decidedly mediocre. You can guess the issue. The actual figures were a lot less damning, but at a glance, they looked very wrong.