Downloads vs Listens: What the Numbers Mean
A download is a file request, your host sent the episode to a device. A listen is measured playback, someone actually pressed play. They count two different stages: a download happens when the file leaves the server; a listen happens later, if the audio gets consumed at all. That distinction is why your host and your app never show the same number.
Most hosts that complain "Spotify says 800 but my dashboard says 1,200" are comparing a delivery metric to a consumption metric and treating the gap as an error. It isn't an error. The two numbers are supposed to differ, because one file can produce zero listens, one listen, or several listens depending on what happens after it lands on the phone.
What is the difference between a download and a listen?
A download is the file leaving your host's server to a listener's device, delivery. A listen (also called a play or a stream) is the file being consumed in an app, playback. Downloads come first and are deduplicated; listens come after, are app-side, and are usually not deduplicated, so the same person can count more than once.
The cleanest way to hold the two apart: your podcast host measures audio delivery, and podcast apps measure audio consumption (Transistor). Delivery happens once, when the bytes are served. Consumption can happen never, once, or many times after that. They're often close, but they answer different questions, "did we ship the episode?" versus "did anyone play it?"
How a download is actually counted (the 60-second rule)
The industry didn't leave "download" to interpretation. The IAB Tech Lab's Podcast Measurement Technical Guidelines, the standard most reputable hosts certify against, currently version 2.2 from May 2024, count a download once per episode, per unique IP-plus-user-agent combination, in a 24-hour window, only after at least 60 seconds' worth of the audio file has been transferred (IAB Tech Lab).
Three things in that definition do real work. The 60-second threshold (measured by file size, not playback) filters out a phone that grabs a byte and bails. The 24-hour dedup window means the same device pulling the file three times in a day still counts as one. And invalid-traffic filtering strips out bots before anything is counted. The practical effect is stark: the IAB's own example notes that 10 raw requests from 6 distinct device-and-IP combinations count as 6, not 10 (IAB Tech Lab). A "download" is already a cleaned-up number, but it still only proves a file moved, not that anyone listened.
Which platforms report downloads, and which report listens?
Your host reports downloads; the listening apps report plays. The crucial trap: each app sets its own threshold, so a "play" on Apple and a "play" on Spotify are not the same event. Apple counts a play at any tap above zero seconds; Spotify (since June 2026) only counts a play once someone reaches 30 seconds.
Here's the split, with each platform's own definition:
| Platform | Headline metric | What it actually counts |
|---|---|---|
| Your podcast host (Buzzsprout, Transistor, Libsyn) | Downloads | File delivered, IAB-deduplicated, 60s of bytes served |
| Apple Podcasts | Plays | Plays on a unique device where duration is greater than 0 seconds, including taps to pause and unpause (Podnews) |
| Spotify | Plays | Each time someone watches or listens for at least 30 seconds, a single standard that replaced the old 0-second Starts / 60-second Streams split in June 2026 (Podnews) |
Two consequences fall out of this table. First, app numbers only cover that app's slice of your audience, Apple's dashboard reports a sample of Apple listeners, not your whole show (Podnews). Second, the thresholds don't match: Apple counts any tap above zero seconds while Spotify waits for 30, so the same listening can register as a play on one app and nothing on the other. You can't add the apps up across platforms, and an Apple "play" and a Spotify "play" aren't the same unit even though they share a name.
Why downloads are falling while listens rise
If your downloads have drifted down over the past two years while your audience feels steady or larger, the metric changed under you, not the show. In fall 2023, iOS 17 changed how Apple Podcasts handles automatic downloads: it stops fetching episodes from shows you've stopped playing, which trimmed a layer of speculative downloads that never turned into listens (Podnews).
That phantom layer was always there. Apps like Apple and Overcast auto-download new episodes for subscribers; if someone falls behind, files pile up unplayed, counted as downloads, never as listens. Podnews estimates about 13% of podcast downloads are never played, for a 2,000-download episode, roughly 260 that never reach anyone's ears (Podnews). Add the May 2024 IAB update tightening filtering, plus the shift to YouTube and Spotify video (where a view generates no download at all), and downloads can shrink while real consumption grows. The number went down; the audience didn't.
So which number should you trust?
Use downloads for reach and for advertisers, because it's the standardized, deduplicated currency the ad market still runs on (Bumper). Use plays, streams, and completion rate to understand whether people actually finished the episode. Neither is "the real number", they answer different questions, and a healthy show watches both.
If you want a benchmark for downloads specifically, Buzzsprout's global stats put the median (top 50%) episode at 28 downloads in the first seven days, the top 10% at 413, and the top 1% at 4,611 (Buzzsprout). One honest caveat most benchmark posts bury: this is one host's data, and the dominant platforms (Spotify and its Anchor tooling) publish no comparable public benchmark, so these figures skew toward the independent, Buzzsprout-hosted end of the market, not the whole of it. Treat them as a rough ladder, not a verdict.
Frequently asked questions
Is a download the same as a listen? No. A download means the episode file was delivered to a device; a listen means someone played it. A download can sit unplayed and still count, and a single download can produce several listens if the listener pauses and replays. Downloads measure delivery, listens measure consumption.
Why are my Spotify numbers different from my podcast host? Because they count different events. Your host counts deduplicated downloads (one per device per day, 60 seconds of file served). Spotify counts a play once someone reaches 30 seconds, app-side, and only for Spotify listeners (Podnews). The gap is expected, not a tracking bug.
Does a download mean someone listened to the whole episode? No. Under the IAB standard a download is counted once at least 60 seconds' worth of the file has transferred, measured by file size, not by playback (IAB Tech Lab). It proves the file moved, not that anyone reached the 60-second mark, let alone the end. For finishing, look at completion rate.
Why did my downloads drop in late 2023? Most likely iOS 17, which changed Apple Podcasts auto-downloads so it stops fetching episodes from shows you've stopped playing (Podnews). That removed speculative downloads that were never listened to. The May 2024 IAB filtering update and the move to download-free video distribution pushed the same direction.
Which metric do advertisers care about? Downloads, for now, it's the deduplicated, IAB-standardized number the ad market is priced on (Bumper). The industry is drifting toward consumption metrics like streams and completion rate, but downloads remain the default currency in most ad deals.
To go deeper on where these numbers come from, see how episodes are delivered in what a podcast RSS feed is, how the audio file is attached in the enclosure tag, how an episode is uniquely identified in the GUID, and how titles and artwork travel with the file in ID3 tags.