I think it is a bad thing, both pragmatically and idealistically, to pay a ransom. However, the idea that it could be faster is correct at least some of the time which convinces people to pay. When it works, which is by no means guaranteed, the decryption software can be run to somewhat quickly bring computers up. You still need to do a lot of work from there, but you're doing it while the rest of the machines are working. The cost of interruption can be high, so that can seem appealing to someone who is either ignorant of or trying not to think about the high chance that it doesn't go as well as they picture. That's why people pay. If they knew how often that key doesn't work or the infection comes back for another payment, maybe they wouldn't.
Also, those who choose this option tend to be doing their jobs badly, which can be catastrophic in a situation like this. For example, when the computers are functional because the ransomed decryption was run, it can be very hard to convince management that you need to gradually take them down for a complete reimage anyway, but if you don't do it, then at the very least, the same vulnerability exists for someone to exploit and quite likely the original ransomware is still there. Several businesses have gotten a series of infections, either from the same source or from multiple ones, until they did the proper thing and rebuilt securely.