A dataset showing that restaurants had zero customers during lockdowns can't possibly be accurate, since they're still open for takeout. I'm sure there's some real trend being exhibited here, but how do we know people aren't just not using OpenTable?