There's probably no "perfect" way (firm yes or firm no) to delineate a "landslide". Perhaps consensus is best. Everyone agrees 1936, 1964, 1972, and 1984 were landslides (even though in each case if you polled 10 voters at random, on average 4 would have voted for the loser, which hardly sounds like a romp); 2000, 2004, and 2016 were close; and 2008 was somewhere in between.
True, there's no cut and dry way to determine if an election is close, a landslide, or somewhere in between, although there are some that everyone agrees on. My personal definition of a landslide is if a second place presidential candidate gets less than 100 electoral votes (or the combined total of all losers with EVs is less than 100), and in a non-presidential election, where the winning candidate gets 60% of the vote or more.
I don't consider popular vote margins in categorizing presidential elections since the national popular vote has no bearing on the outcome anyway. (Or put another way, the popular vote only matters at the state level.)