Or does it?

It turns out that a quality score-based model explains a mere 1.2% of the variability in the data (R² = 0.012). To make matters worse, the data has a ρ-value of 0.46!

Statistically speaking, that’s a completely useless model.

And, it explains why improving your quality score often does little to improve your cost-per-conversion.

If you are wasting 75.8% of your ad spend to begin with, improving your quality score isn’t going to lower your cost-per-conversion by much.

However, if you’ve eliminated most of your wasted ad spend, a higher quality score will reduce your cost-per-click and—by extension—your cost-per-conversion.