30 seconds/recording: presumed average recording length
1,000,000 seconds: a bit higher than the total number of seconds of audio recorded
8 months: Amount of time service was around
1,000,000 seconds / 30 seconds/recording = 33,333 recordings
33,333 recordings / 8 months = 4,166 recordings/month = 138 recordings/day.
Assuming that the average user tried it out once and then didn't use it again, VoiceGem was getting about 140 new users a day. If the average user used it more than once, they were getting even fewer new users. Startups are hard.
Edit: for some reason, Josh Constine at TechCrunch decided to reference my arithmetic. I just want to make sure it's clear that I'm happy for Arda and Alexander for finding an exit, and that startups are hard. I'm not trying to be snarky.
all you are going to deduce from this is whatever you put into your model. if they were getting exponential growth then most of the calls would have been recent. and they imply that they had viral growth because use was spreading...
on the other hand, presumably they didn't decide to fuck over a bunch of grannies for fun (even if they did go to palantir). so it was an acquihire. so maybe some part of the description was more positive than the unvarnished truth?