I don’t like “I know people..” kind of so-called evidence.
this means nothing.
I have been doing this informally all summer long. I'll ask people "among your circle of friends and acquaintances, how many of them voted for Trump in 2016 but aren't going to vote for him in 2020 and how many didn't vote for Trump in 2016 but are going to this time?" Every time, the second group was larger than the first. In most cases, the first group consisted of zero people and in several cases, the respondent admitted to being in the second group. Probably as reliable as polls with 1-2% response rates.
Some of my best friends are people that know people...