I'm not sure anyone knows for sure who this works for and why but the data so far is pretty interesting. In the first trial, they tested up to 6 months and 43% of people treated got improvements in tinnitus and 29% (4 patients) got a greater than 20 point TFI improvement (scroll up to see charts).
Interestingly, those 4 patients all had severe tinnitus.
It does seem that local cochlear NMDA receptor hyperexcitability (which is what the drug acts on) happens very acutely to everyone with cochlear damage (and the excitability gets propagated up the auditory channels) but that seems to be more variable long term when the central plasticity/"phantom cochlea" hyperexcitability may become a bigger part. This may be why it doesn't work for everyone and less than half of the participants responded.
However, it may be that people with *severe* tinnitus retain the local cochlear NMDA stimulation component of neural hyperexcitabilty longer. And actually studies have shown that the stress of severe tinnitus in and of itself can be somewhat self perpetuating, too, because of dynorphin release (which sensitizes the NMDA receptor further):
Endogenous dynorphins, glutamate and N-methyl-d-aspartate (NMDA) receptors may participate in a stress-mediated Type-I auditory neural exacerbation of tinnitus
In other words, this may work better in chronic cases when the tinnitus is severe. Which would be good news for the worst cases. Phase 2 would answer that better, though, obviously with a bigger sample size.