Thanks for producing this Peter, it’s very helpful. I sent you some metric data on the 80,000 Hours Podcast, but now that I’ve seen the post, I can give you the best numbers for the table. I would suggest putting these figures in instead.
The podcast only started half way through 2017, I’m not sure how you want to handle that.
Those are the maximum number of subscribers recorded at any point in the year. It’s probably a few % too high in both cases, but I’ve found that’s the measure most robust to random measurement variations. The overestimation should also be constant year to year.
Podcast downloads/plays don’t correspond to actual times people listened to a full episode. They include people pressing play but only listening to a few seconds; bots downloading the show; automatic downloads by the podcasting software that are never actually listened to; and so on. So they’re massive overestimates of the number of times an episode was listened to, say, half way to completion. However, the overestimate is likely to be a pretty constant fraction year-to-year, so you can still make relative comparisons.
Thanks for producing this Peter, it’s very helpful. I sent you some metric data on the 80,000 Hours Podcast, but now that I’ve seen the post, I can give you the best numbers for the table. I would suggest putting these figures in instead.
Net new podcast subscribers added
2017 − 4,600
2018 − 10,500
Total podcast downloads/plays
2017 − 87,600 (average episode length 1.61 hours.)
2018 − 517,100 (average episode length 1.97 hours.)
Notes on interpretation
The podcast only started half way through 2017, I’m not sure how you want to handle that.
Those are the maximum number of subscribers recorded at any point in the year. It’s probably a few % too high in both cases, but I’ve found that’s the measure most robust to random measurement variations. The overestimation should also be constant year to year.
Podcast downloads/plays don’t correspond to actual times people listened to a full episode. They include people pressing play but only listening to a few seconds; bots downloading the show; automatic downloads by the podcasting software that are never actually listened to; and so on. So they’re massive overestimates of the number of times an episode was listened to, say, half way to completion. However, the overestimate is likely to be a pretty constant fraction year-to-year, so you can still make relative comparisons.
— RW
How big do you expect that fraction to be? (Or: what percentage of those numbers do you expect to be ‘real listeners’?)
Thanks—I updated the post with that data.