The iOS 17 changes in Apple Podcasts
The dust is beginning to settle on iOS 17’s changes in the Apple Podcasts app - and the effects of those changes are becoming clearer to the industry.
In this special report: a timeline of the changes; how automatic downloads work and what changed; how the industry spotted it, and the effect it has had on podcasting.
We also cover how Apple Podcasts communicated it to the industry - and what we can learn for the future.
Let’s dig in.
A timeline
On iPhone and iPad, updates to the in-built apps, like Apple Podcasts, are done via an OS upgrade, rather than the usual App Store update experience. One of those upgrades, iOS 17, was announced on June 5.
On that day, Apple announced new features for Apple Podcasts in the upcoming iOS 17. Particularly, the app would support episode artwork; a better now-playing experience; and a improved search mechanism.
iOS 17 became available on Sep 18.
More than three weeks after the public launch of iOS 17, Apple Podcast made a quiet blog post, flagging that there had been “improvements to the way Automatic Downloads are handled”. These improvements were flagged to us as a small change, and as a result, you’ll see them mentioned quite low in Podnews on Oct 12.
iOS 17 was also slower than normal to roll out: with, according to OP3’s open analytics, the majority of users switching to iOS 17 towards the end of November - so the effects weren’t immediately obvious.
But, by Feb 2024, those “small changes” had cost Acast $7.2mn, as the company revalued its advertising contracts after a listening slump of 12%.
Bryan Moffett, COO at National Public Media which manages corporate sponsorship sales for NPR, told us that the effect differs for the type of show. Daily shows with habitual audiences saw a decline of around 10% in monthly downloads, he told us, but “longer shows published on other cadences were, on average, showing 30% declines in monthly downloads”.
Other publishers that Podnews has spoken to tell us they have had to give make-goods to advertisers, after networks suddenly had not delivered as many downloads as first thought.
Data from the Triton Digital US Ranker makes it clear: on average, almost a third of weekly downloads have disappeared year-on-year.
The above data is based on entire publishers, and some may have produced fewer shows this year. For another comparison, Podtrac data given to Podnews suggests an average 15% drop in downloads for most large shows.
How do automatic downloads work?
By default, Apple Podcasts enables automatic downloads of podcasts for all shows that a listener follows. It’s the only major podcast player that does this.
Automatic downloads mean that, when a new episode is posted for a show, Apple Podcasts will automatically download it. As the company’s detailed blog post says, that normally happens in the background, when a phone is charged and connected to wifi.
This causes some problems for measurement, though. Our guide to podcast statistics explains:
Audio can be automatically downloaded to your phone, in case you want to hear it later. Or, audio can be downloaded when a listener wants to hear it - a “user-initiated download”, or what most people mean when they say a “stream”. Your podcast hosting company doesn’t know which it is. It just knows it’s a download. And, no podcast hosting company knows whether a piece of audio has ever been listened-to (a “listen” or a “play”). It can only ever know it’s been downloaded.
Automated downloads are paused when the device runs out of storage, or when the listener stops listening - if “the listener does not play a show they follow for more than 15 days and hasn’t played the latest five episodes”, says Apple.
Pausing of automated downloads didn’t change with iOS 17. What changed is what happens when that listener starts listening again.
The change that Apple Podcasts made
Let’s take the example of The Daily.
The Daily comes out six days a week: five weekday episodes and a longer “Sunday Read”.
Mary listens to The Daily sometimes, when there’s a story that she’s interested in. She occasionally picks up her iPhone, looks at the episode list, finds an episode title she likes the look of, and listens to it.
As long as she listens enough, Apple Podcasts will always automatically download every show to her phone - even if she only listens once every two weeks. But, if she doesn’t listen enough, Apple Podcasts will automatically pause those downloads.
What happens when she starts listening again is where the change happens.
According to Apple Podcasts…
Before iOS 17, when a listener would unpause automatic downloads, the system would automatically download all unplayed episodes.
With iOS 17, Apple Podcasts will not download previous episodes and will resume automatically downloading new episodes.
Mary discovered a new show, so she stopped listening to The Daily. Apple Podcasts stops automatically downloading it after about 15 episodes. Two months and 35 further episodes later, she sees an ad for The Daily, and she remembers what a great show it is - so she listens to one of the recent shows.
Before iOS 17, Apple Podcasts downloaded all 35 unplayed episodes.
With iOS 17, Apple Podcasts just downloads one - the one Mary listened to.
That issue got significantly worse if Mary returned to a show after a longer period of time - say, six months, with, for example, 250 shows that had been unlistened to. Apple Podcasts would attempt to instantly download all 250 episodes: and that wouldn’t go smoothly.
One person told us: “The behavior we’d see in something like this would be that all 250 download requests would be made at nearly a rate of 20 downloads per second and the hosting platform would respond back to all of them, trying to sending 250 IAB certified downloads. Maybe 50 would succeed before iOS shut the app down. On the following day, the app would attempt the remaining 200; but again, iOS would shut it down. With this example, those 250 missed episodes would result in a total of 750 downloads over the course of five days”. Because of the speed of request, frequency capping solutions for advertisers weren’t able to mitigate this issue.
A small change?
Was it a small change? That’s what we thought. After all, there can’t be many people like Mary: people who “follow” a show but don’t listen to every episode.
Except: we’re all like Mary.
The Infinite Dial 2023 says that 120 million Americans listen to podcasts every month; but only 89 million every week. That’s 31 million occasional listeners, each of whom might be marked as a dormant listener by Apple Podcasts.
In the UK, RAJAR’s MIDAS Study suggests that 63% of podcast listeners listen to more than half of all the episodes they download - so, 37% of all podcast listeners listen to fewer than half of the podcasts they download.
The only organization who knew for certain how much of an effect it would have? Apple Podcasts.
NPR’s Bryan Moffett: “We all knew for years there was a big delta between the data in Apple Podcasts analytics and the data from our logs. But there wasn’t much we could do about it, and we couldn’t correlate it with our log data. Because of that, I don’t think anyone really put together the impact of those catch-up downloads on our shows. Apple could have helped us all understand that better.”
How it was noticed
We’re told that publishers and hosting platforms had individual meetings with Apple trying to understand and communicate the issue, and Podnews has learnt of at least one meeting on this issue more than four years ago. People familiar with the issue tell us that without full data, Apple appeared to dismiss the reports; but there seem to have been many different organisations spotting the problem. “This was an issue we, and Omny Studio, had been working with Apple on for two years since right after iOS 15 came out when we first saw the issue with some shows and excessive downloads of [the] back catalog,” said Rob Walch of Libsyn on an episode of The Feed in October 2023. We’re aware of analytics companies, podcast hosting companies, and publishers who all independently flagged-up an issue.
In 2022, Podscribe, a podcast attribution and analytics service, says that it pinpointed the issue and what was causing it; and the issue was taken to the IAB for potential mitigation in their measurement guidelines. One publisher told us: “The best way we found was to limit dynamic campaigns to episodes no older than about 45 days. That reduced the far right end of the distribution. That’s when we really began to understand what was happening.”
Additionally, Apple has a quality control team that regularly monitors the battery and storage use of all of its in-built apps, and we’ve been told that it was this team that independently spotted the issue.
Whoever spotted it - and it’s likely that it was all of the above, working independently - what is clear is that Apple Podcasts knew about this behavior by late June, having made changes to auto-downloads for the forthcoming iOS 17 release.
It appears, though, that Apple Podcasts was not going to highlight the significant change in behavior that they’d made to the app. It wasn’t flagged as part of the changes in iOS 17’s Podcasts app, nor when iOS 17 was publicly available. It made one, small, change to an undated webpage, as the Web Archive shows:
“If a listener resumes playing that show, or changes the download preferences for that show, Apple Podcasts will resume downloading new episodes.” is what Apple Podcasts said in July 2023.
“If a listener resumes playing that show, or changes the download preferences for that show, Apple Podcasts will resume downloading all unplayed episodes.” is what they’d amended it to, without any notice, by September 2023.
Sounds Profitable, a membership organization of over 150 podcasting companies, may have forced Apple’s hand. Apple seems to have rushed to publicly document this change just a day before Sounds Profitable released a “joint statement” signed by many of its members that detailed it. While some deny the extent of that group’s involvement, Apple Podcasts only appeared to communicate the change once it knew that this third-party statement was due to come out.
Does it matter?
The “download” has never been a particularly good reflection of “listening” even though many want to conflate the two. With this change, Apple seems to have put the two statistics better in line with each other. It’s a good thing.
Some have been keen to paint this as a story of fraud - of publishers selling fake automated downloads to unsuspecting advertisers. That’s not the case. It’s always been known - and priced-in - that automatic episode downloads aren’t always listened-to. This did not come as a surprise to anyone.
It’s also a welcome change for the industry to ensure that more episodes are actually listened-to. As we’ve seen from new Podscribe data, podcast ads are performing better as a result. Particularly, “impression-based” buys on back-catalogs have made a significant improvement. This is good news for the podcast industry, and good news for advertisers.
However, Apple Podcasts is responsible for around 40% of all podcast downloads. Accordingly, Apple Podcasts has an significant influence on the $2bn podcast industry. It presumably has the instrumentation to estimate the effect of any change like this, too.
This change was significant. Without notice, it wiped off $7.3mn from just one podcast publisher’s bottom line; and has significantly reduced revenues and downloads from many others.
Apple Podcasts has made significant unannounced changes in the past, with significant numbers of automatic downloads from the Apple Watch causing the IAB to issue new guidance and a new set of measurement guidelines.
If this tells us anything, it’s that Apple Podcasts has a responsiblity to communicate clearly and effectively if it makes any significant changes to its app. We should be greatly concerned if it doesn’t see the importance of clear communication.
And this also shows that, working together in the open makes us all stronger. Let’s hope everyone in the industry does so.
