An article so absolutely shallow and superficial as to be entirely braindead. I won't go into why (if you don't already know, then you shouldn't be writing SQL code in the first place) but I will correct one error. An index scan is not "much better" than a table scan. It's exactly the same cost as scanning a table as wide as that index. In this example, if the table had no other rows than last name and first name, the cost would be identical whether table or index scanning, and the QO would only favor the index scan on the assumption that the index is more likely to already be cached.