Without saying anything in particular about the US presidential election, it seems like it once again is a good opportunity to question the value of assumptions as substitutes for fact.
In essence an opinion poll is an assumption. Perhaps even more so in the US, where the first challenge of validation your assumption is to just get those you surveyed to actually go and vote.
What the election should (again) teach us is that the only data and insights that matter are actual data and insights, i.e. facts. In this context? The votes actually cast.
Everything else is inherently fake.