GL: I’d like you to do some actual literature research into “shape descriptors” — it would be great if you could cite a few papers.
Get more feature metrics!
Great work.

with 20 criteria, or even 8, you might lose a lot by reducing to a 2D dataset. maybe you can make the display somewhat interactive that allows you to “sort” via different metrics. and you still can have a general typeface similarity view, which i think you have going on in this demo. neat job with PCA

You might want to look at different ways of visualizing it – perhaps a user can click a single font and all the others are classified relative to it based on “degree of similarity”. Or, maybe people should choose which metrics are used, or how they are weighted.

I think you should consider making it 1 dimensional, and display all the typefaces in a line.
I also think you should have several different views to look at this in different ways, because I don’t think there is a single ideal way to visualize this.

Add a zoom function.

You make it look easy as pie. 🙂 -SB π

This could be cool as a game. You guess which typefaces are related and get scored on how close they are. See http://fontgame.ilovetypography.com/

When you mentioned referring a font to a friend, I wonder if you could incorporate some sort of font upload: “upload your font and see what else is close.”