We actually saw a very bimodal split: One group of employees had virtually no measurable change (neither improvement or decline) in any metric, subjective or otherwise.
The other group's productivity plummeted by nearly all metrics (again, from subjective performance reviews to any objective measures we looked at).
The angry internet comment section solution is to fire everyone in group #2 (people who can't work from home), but that group was large enough that no amount of firing or intensive coaching was going to turn the ship around.
The problem isn't as simple as you read about on internet comment sections. In my experience, it's good to have some remote employees but the WFH dynamic collapses when you try to let everyone do it. It's not appropriate for everyone. Normalize having some remote employees while requiring others to be in the office if they can't handle WFH. It's hard.