Facebook Podcasts - a coding error means it's not as big as you think
Podnews has discovered errors in the coding of Facebook Podcasts which is delivering wildly incorrect download data to some podcasters.
Gary Arndt, from the Everything Everywhere podcast, logged into his podcast host to see that his total downloads had increased by five times on Dec 25. Five times as large!
His podcast is hosted on Libsyn, which offers a choice of stats. It defaults on “Unique”, the legacy stats that Libsyn give all podcast publishers. Since Gary is a larger publisher, he’s also opted for “IAB” stats - calculated the same way as most large podcast hosts do - and on clicking “IAB”, the increase in numbers simply disappeared.
On looking at the “downloads per user agent” list, Facebook was #1 - by far - in Libsyn’s legacy stats; but Facebook had largely disappeared when viewing the IAB stats.
Something odd is going on with Facebook.
What was going on?
We went to our own podcast logfiles to see; and while we’re not seeing a lot of traffic from Facebook, a few entries did stand out.
Above, are two partial downloads from Facebook - a 556KB download of our Dec 22 show (which was 3.5MB) from an iPhone, and a 489KB download of our Dec 23 show (which was 2.8MB) from an Android phone.
These downloads aren’t large enough to contain a minute’s worth of audio. (Our Dec 22 show has 56KB of ID3 tag, and at 112Kb/s, the file downloaded would have only played audio for 35 seconds.)
The IAB specification says (page 11) that “to count as a valid download, the ID3 tag plus enough of the podcast content to play for 1 minute” should be downloaded. These downloads aren’t big enough for that.
In short, these aren’t “real” downloads.
Libsyn’s IAB statistics have correctly removed these bogus downloads: but Libsyn’s legacy stats have not.
Facebook Podcasts is making many bogus partial downloads
If it’s happening with Libsyn, and happening with our own hosted podcast, we wondered if it was happening with others.
So, we checked with Buzzsprout, one of the largest podcast hosts, to see if Facebook was doing this on its platform, too. They told us:
After digging, more than 47% of our partial downloads in the last 24 hours have come from “FB4A/Facebook” and “iOS/Facebook”. I took a deep dive into a few of the most popular episodes on Facebook and definitely see a trend where those episodes have a disproportionate number of partial plays from Facebook awaiting completion. In one case, Facebook accounted for almost 18% of the episode downloads, but 99% of the partial downloads for that same episode: and the number of partial downloads was more than 100% of the complete episode downloads!
Buzzsprout is IAB certified, so its users wouldn’t see these partial plays in their stats.
However, if you’re not with an IAB-compliant podcast host, or you don’t get IAB-compliant numbers, you will see significantly higher numbers than you’d expect from Facebook - and those numbers aren’t real.
What we think is causing this
As you scroll down your Facebook feed, the app is constantly loading more data, including posts, images, video and audio.
We think that when you see a podcast in your timeline, the Facebook app is pre-downloading a little bit of the audio file to work out how long the audio is, and to ensure that it’s got 30 seconds of audio ready so you can hit the play button and instantly listen to it. This is against IAB guidelines.
Testing this is difficult, since Facebook hasn’t launched the podcasts product in Australia, where we’re based. However, we ran a local proxy on our iPod Touch, and watched network requests as we tried sharing a podcast that we own (which you can see anywhere in the world). As we shared it, the Facebook app downloaded something from Buzzsprout, the podcast’s audio host without playing anything (and without us pressing the play button).
Facebook already grabs a copy of every episode’s audio for its own systems as soon as they are published, so it could work out the podcast’s length at that point, rather than download the start of the file to do so. That wouldn’t help with the buffering of the audio itself, however - and perhaps its engineers have prioritised speed-to-play in the interface, without knowing what problems that causes.
In this case, we think that this topical episode about Frankincense and Myrrh was shared by a number of people within Facebook, so it appeared on thousands of peoples’ timelines - causing thousands of partial downloads of audio.
That doesn’t mean that this podcast was played, however. It can’t have been: the partial download was only large enough to play back around 30 seconds of audio.
This is a bad engineering decision by Facebook - it will cause unnecessary bandwidth use by its app, and will cost podcast hosting companies significant revenue in wasted traffic. It would be ideal if they’d confirm to the IAB standard, and only request audio files when a user as asked for them.
Your podcast’s numbers, however, are unaffected by this if your podcast host uses IAB certified numbers.
But only if.
The perils of using non-IAB numbers
Libsyn is unique in only offering IAB 2.0 stats to customers that pay for them; and always default to its own legacy stats on the Libsyn dashboard.
Libsyn report on its download numbers in its podcast The Feed. However, for The Feed’s numbers, Libsyn uses its 'uniques’ legacy statistics, not IAB certified data. Libsyn told us in 2020 that “the reason we report uniques is because we are telling people to compare these numbers to theirs; and many users do not have IAB numbers.”
Libsyn’s Rob Walch has recently used these legacy statistics to talk about Facebook’s growth on the platform: growth that, we can now see, appears almost entirely bogus: based on partial, incomplete downloads that have nothing to do with plays.
“I do see by the end of Q1 2022 Facebook being in the top five of apps,” Walch is quoted as having said. That might be: but only if you’re looking at the wrong numbers. Perhaps it’s time for Libsyn to give IAB numbers to all its customers: just like everyone else.
|James Cridland is the Editor of Podnews, a keynote speaker and consultant. He wrote his first podcast RSS feed in January 2005; and also launched the first live radio streaming app for mobile phones in the same year. He's worked in the audio industry since 1989.|