In my experience (and I didn't land ANY of the consulting gigs I really wanted, so take it for what it's worth...) these types of questions are more about the process than the product. Accuracy isn't relevant, but understanding the form and how you're supposed to deal with it is what's important. It's a mini-presentation, so write it out and explain to the interviewer what you're doing while you do it; it should be both visual and mathematical. Where you can really score points here is to talk about the assumptions you make (make a written list after you repeat the problem back to the interviewer for the first time and ask whatever questions you have) and, after you give your solution, where the assumptions break down and how you'd adjust the number to account for those things. In an ideal world, then, you'd give the Cosentino answer from the book scan, THEN talk about things like population density and how they would change the answer slightly. You could give a revised number, but it's the idea generation that's more important at that point--a written list of a couple "extras" at the end can go a long way. Remember that the interviewer is going to take your scratch paper at the end of the interview, so it'll be the only tangible evidence of your response. Honestly, it sounds like your answer was okay, except for the weird stuff about not using a notepad. I don't think you flubbed it as badly as you probably think you did.