America is large, multifaceted, diverse and complicated. The difference in the popular vote was large in political terms, but in absolute terms 48% of the voters voted for the winning candidate.

To claim that the problem was "the data driven approach" seems to be very bold. There are a lot of aspects to a campaign, and flaws in any of them can be fatal. Ironically, we have very little data on the subject of what went wrong for Hillary; although given the election results it was probably a problem in specific regions.

fivethirtyeight.com had Trump as the underdog with a ~30% chance of victory (so, we'd expect this sort of outcome every 4 elections). This suggests that data was painting an accurate picture to me, everyone was surprised but the election result wasn't unthinkable.