My thought is, it’s about behavioral science and macrotrend.
People tend to like the posts that already have a lot of likes. It’s the Matthew effect and Herd Effect in behaviors – likes tend to go extreme. So when the feature is hidden from viewers, the distribution of likes among posts tend to get flatter. And it corresponds to the macrotrend that advertisers are more and more leaning towards working with “microinfluencers” – those who have 1500 followers or less rather than “macroinfluencers” – those who have millions of followers based on the observations and researches that people trust those who are closer to them more, and the conversion of the ads are actually higher.
So knowing this, some numbers they could look at are conversion rate of ads, avg revenue of ads, distribution of likes, and the trend and distribution of turning off of ads.