Sparsity
The percentage of the User-Item Matrix with no rating, typically 95 to 99 percent or more. Sparsity is the nature of the problem, not a data-quality bug.
When to use
Compute it as the share of empty cells: Seret+ is 80 percent empty, and Balagan Books (5,000 users by 20,000 books with 50,000 ratings) is 99.95 percent. High sparsity hurts Collaborative Filtering because two users may overlap on very few items, which favours content-based filtering.
See DS Lec 6.