I think it's a somewhat useful proxy. You can't measure "friendships formed", so you find something you can measure that hopefully correlates with "friendships

I think it's a somewhat useful proxy. You can't measure "friendships formed", so you find something you can measure that hopefully correlates with "friendships formed". As you point out, it's not without it's flaws, but I think most people would say that some data is better than none.