Our data suggest people keep listening to podcasts even if they’re very long

At the 80,000 Hours Podcast we try to do ‘unusually in-depth interviews’ about ‘the world’s most pressing problems and how to solve them.’

Our average episode length so far is 2h03m. Our longest interview, on the nature and ethics of consciousness, is 4h42m.

While it would be nice to have less competition from similar shows, I’d like to encourage more interviewers to go in-depth, even if that takes their episodes over 2 hours. Why?

Since 2018 everyone has been able to get detailed listening time data from Apple Podcast Analytics, to see how far listeners are actually making it through each episode.

I’ve looked into our own data and, contrary to the expectations of many, people just keep listening to long episodes, at least so long as they’re good.

That longest interview has a completion fraction of 58% — people are getting through fully 2h42m on average.

That’s higher than average for an episode that long — it was a particularly good conversation. To see what is typical, excluding that outlier, take a look at Figure 1. [1]

Figure 1 Fraction of people who start listening to an episode at all who are still listening, by half-hour block. Initial drop-off is huge as people decide whether they’re interested in the topic or guest. But listener retention is fairly good after that.

People do indeed drop off as episodes get longer, but two-thirds as many people are still with me between 3h30m and 4h as were with me between 30m and 1h.

So the benefit of incrementally longer recordings remains high.

One possible objection: maybe I’ve scheduled longer interviews with better guests. While I can’t rule out some effect here, in general that’s not the case — in fact more famous guests (like Peter Singer, Tyler Cowen, or Cass Sunstein) are more likely to be pressed for time and set strict limits on how long interviews can run. [2]

Figure 2: Episode length (hours) against number of people who start listening.

Another possible objection: maybe fewer people are willing to start listening to longer episodes? Not as far as we could see (see figure 2). There’s no relationship between episode length and the number of people who start playing it.

Why longer conversation are also higher quality

So, our data suggest that if you continue a long interview, and it’s still good stuff, people will mostly stick with you (or at least our audience will).

What are the other important benefits of longer interviews?

1. The second hour of an interview costs way less to produce than the first. You’ve already paid the fixed cost of reading their book/paper/article, writing questions, getting the recording equipment set up, meeting them, overcoming your nerves, and so on. I’d say the second hour on the margin costs less than half as much as the first.
2. You’re more likely to get to new questions the guest hasn’t been asked before. There’s not much value to society in just getting someone to answer the same obvious questions again, or summarise the thesis of their new book for the umpteenth time. But that’s hard to avoid early in any given interview because you have to bring listeners along with you.
   
   When I listen to other podcasts, I’m tired of hearing people say things I’m already familiar with for 45 minutes, then having the interview end at that point for no apparent reason.
3. You build personal chemistry over time, and guests get more comfortable. I find the best moments on the show are often past the 2h30m mark, when we’re both more likely to be at ease, let our guard down, be authentic and go off script.

The intuition that there won’t be much more to say after an hour-long interview has always puzzled me.

You’ve probably spoken to your best friends for hundreds or even thousands of hours. Are your conversations less valuable now than the first hour you ever spoke? Maybe you’ve hit declining returns and are running out of things to say. But more likely the trust and understanding you have of each other’s perspectives means there’s more to say and it’s actually more enjoyable to hang out than it used to be. A quick Twitter poll (Figure 3) indicates that’s true more often than not.

When should episodes actually end then? You get too tired. Or you run out of good questions to ask. Or you really don’t have time to go on.

What about guests or hosts whose diaries are jam-packed with commitments? I think they should do ~half as many interviews, but make each one ~twice as long. [3] That way they’ll be able to go into details and address novel considerations or objections, rather than just provide a quick summary of their opinions again and again.

There’s a good reason most of us have ongoing friendships, rather than talk to 3 new people a day, never going back to speak with them again.

If you’re curious about the kinds of questions that come up a few hours into an interview, you can subscribe to the ‘80,000 Hours Podcast’ wherever you listen to podcasts.

—

[1] I tried estimating the length vs listening time relationship various ways and then averaged them, but the results barely varied.

[2] In future when we have a larger data-set I’ll break out those where the length was determined ahead of time by the guest, and see how things look for just those episodes.

[3] In fact they might be able to do 60% as many because they save on all those fixed costs of setting up each one.