Easy for me to dunk, but in all honesty social media algos are notoriously difficult to tweak for 3 reasons; 1) can’t a/b test cleanly due to network effects. What seems like good results in a small audience test can turn sour when you add in large network complexity. 2) what’s good for the masses can suck for niche audiences. I wouldn’t be surprised if this particular algo tweak improves metrics overall due to mass popular content being better surfaced but turns off people who want to engage with deeper / thoughtful discourse. Tracking key influencer segments can help here. 3) what’s positive for the short term can be terrible for the longer term. Ie, certain changes boost short term usage but don’t align with people’s long term values (compounded due to network effect point from #1). Holdouts can help but aren’t accurate. One has to have conviction on the “feel.” 🫡 @nikitabier