The PAC, Priorities USA, spent a good chunk of the cycle testing the effectiveness of ads, some 500 in all. And, along the way, they decided to conduct an experiment that could have potentially saved them tons of money. They took five ads produced by a fellow occupant in the Super PAC domain—the Lincoln Project—and attempted to measure their persuasiveness among persuadable swing state voters; i.e. the ability of an ad to move Trump voters towards Joe Biden. A control group saw no ad at all. Five different treatment groups, each made up of 683 respondents, saw one of the five ads. Afterwards they were asked the same post-treatment questions measuring the likelihood that they would vote and who they would vote for.

The idea wasn’t to be petty or adversarial towards the Lincoln Project, which drew both fans and detractors for the scorched-earth spots it ran imploring fellow Republicans to abandon Trump. It was, instead, to see if Twitter virality could be used as a substitute for actual ad testing, which took funds and time. If it turned out that what the Lincoln Project was doing was proving persuasive, the thinking went, then Priorities USA could use Twitter as a quasi-barometer for seeing how strong their own ads were.

But that didn’t turn out to be the case. According to Nick Ahamed, Priorities’ analytics director, the correlation of Twitter metrics—likes and retweets—and persuasion was -0.3, “meaning that the better the ad did on Twitter, the less it persuaded battleground state voters.”