Though they promised real-time data on Election Day of the sort that was previously only accessible by campaigns and network executives, it turns out that Palo Alto-based startup VoteCastr pretty well blew it on Tuesday just like all the traditional pollsters did. As Variety explains, VoteCastr got their projections wrong in five of seven swing states, saying that Clinton would be winning in all of them based on their voter turnout counts.
As you can see in their final tallies, which were posted to Slate at 7:55 p.m. ET, they had Clinton taking Florida, Iowa, Ohio, Pennsylvania, and Wisconsin, none of which came true. The only states they got right? New Hampshire and Nevada.
Also, in Nevada, they were reporting results for Jill Stein even though she wasn't even on the ballot there. "She is not on the ballot but was included in our survey which is used to develop the model," VoteCastr wrote on their site. "We messed up and we are correcting the Nevada results accordingly."
Additionally, as Variety reports, VoteCastr's visualization maps didn't update all morning, and at 2 p.m. ET were still static and showing only early voting data.
So much for "fascinating insights into what's happening on the ground," which the startup had promised. They had said this was the first time this experiment was being tried on such a large scale, and clearly it didn't go well.
Just one more thing in a long, depressing list of things that didn't go well this week.