Popularity is but one input for "best" -- depending on one's definition of best. There's enough metadata (price, location, cuisine, etc) to correlate with actu

Popularity is but one input for "best" -- depending on one's definition of best. There's enough metadata (price, location, cuisine, etc) to correlate with actual payment data to filter out fast food like McDonalds. It doesn't have to be a totally naive statistical approach that gives "dumb" answers. The idea is that what people actually pay for is a different set of vector inputs compared to what people submit reviews for (Yelp/TA) or what people link to (blog with links to favorite city restaurants.) Google's search index is over 100 petabytes but even that gigantic database is missing lots of data that other entities can collect and convert into uncontested search results.