I'd go the other way. Ban the owl familiar.
I mean, sure, you can do that. But banning (or nerfing) is generally an inferior solution to buffing, both in video games and in tabletop games. Yes, you need to be cautious about trying to bring things up
just below the desired ceiling of power. And sure, sometimes a nerf (or
very rarely a ban) is really well and truly the
only acceptable response, if something is so ridiculously and unintentionally powerful that it warps the game around it. For something like this though...I don't think the Owl is overpowered beyond belief. It's just got the best package deal. Make the other packages good
for different reasons, and you'll induce a trade-off review, rather than a brute calculation.
There's always going to be a 'best'. That best will just morph from owl to something else.
Only if you presume that every option can be evaluated by a single, fixed metric. The whole point of balance in a non-trivial game system is to find ways to offer
incommensurate but
clearly valuable benefits. E.g., to use 4e terms, having someone with healing and support powers (a Leader, whether by their innate class role or by investing feats, powers, and/or their PP into gaining that role) is "equally vital," in some sense, to having someone with high defenses and powers that force the enemy to choose between bad options (a Defender, same deal as the previous). Leaders and Defenders are incommensurate, because you can't objectively put both of them on a single metric and truly capture what makes each of them worthwhile. Yet both of them are, quite transparently, extremely valuable to have in an adventuring group.
5e, unfortunately, tends to push things in the other direction. Races, with or without Tasha's, get evaluated this way, hence 5e Dragonborn are kind of crappy while 5e Elves (or even moreso Half-elves) are amazing. Within each class, there will almost always be a "best" subclass, because the intent of "role-less" subclasses means every subclass is expected to fill
pretty much the same spectrum of options, with some very slight tweaks. (E.g. Valor Bard and Swords Bard have more combat prowess than Lore....but Lore is pretty much unequivocally better than either of them unless you
really really want to be a mono-class Bard in medium armor.) Classes, same deal, every class is expected to fill (almost) every role based on what you choose to do with it, so we can inherently apply the same metrics to each and get a commensurate evaluation.
With familiars, at least, you can have truly incommensurate stuff. E.g., owls can fly and see at night. Ravens can speak to other people besides their master. Chameleons can do stealth (these are fantasy chameleons!) An arthropod familiar, if you can stomach having one, could climb on walls and ceilings. Etc. Giving every familiar some kind of valuable trait that can't be directly measured against every other familiar's traits seems a much better way of addressing this problem. And it's not like you'd have to fix every familiar. Just ask a player that gets a familiar what they like about the animal they
want to choose, and find something that works in that direction. E.g. if your wizard wants a cat familiar because her player is a cat dad, think about the useful characteristics an intelligent cat could have. I'd argue stealth, destructiveness (that is, vandalism, not combat), and danger-sense (Owls are good hunters, but cats tend to react rather strongly when "something is wrong"). Work with the player to make those things fit their vision of how the character interacts with their familiar, and the problem is solved.
Of course, it would've been better if the familiars were already written with this concept in mind, but we can't always get what we want.