I read the article and I also can't figure out how they measured what they claim to have measured. The results seem made up in much the same way other advertising metrics and results seem made up.
Whenever I used to run Youtube ads ($100k campaigns) the account manager at Google Ads informed us about an automated option of neasuring assisted and unassisted recall. People whom seen the ads were presented with one question surveys on Youtube. It was pretty interesting as you could see what would be a good reach and frequency for your ads.