A download chart that falls off a cliff is alarming, and it is usually not what it looks like. Most sudden drops are a change in how the number is counted rather than a change in who is listening. Work through these in order, because the cheap explanations are also the common ones.
Before diagnosing a cause, make sure you are comparing like with like. Three things routinely manufacture a drop that never happened.
A moved publishing day. Most download reporting is windowed — a 7-day or 30-day figure per episode. If you usually publish on Tuesday and this one went out on Friday, a 7-day comparison is measuring a different slice of the week.
A previous episode that did unusually well. One episode with a guest who shared it widely raises the baseline you are comparing against. The next episode has not dropped; it has returned to normal.
A partial window. An episode released three days ago has not finished accumulating. Podcast downloads arrive over days and weeks, not hours, because apps fetch episodes when subscribers next open them.
This is the cause that catches most people, because nothing in your show changed and the fall is abrupt and permanent.
The clearest recent example is Apple’s change in iOS 17 (autumn 2023), which stopped automatically downloading new episodes for subscribers who had not played the show recently. Shows with a large but inactive back catalogue of subscribers saw an immediate step down. Nothing was broken and no audience left — the automatic deliveries to people who were not listening simply stopped being made, and therefore stopped being counted.
That is the shape to look for: a step, not a slope. A measurement change produces a cliff on a specific date and then a flat new baseline. A real audience change produces a gradient. If you can point at the exact day it fell and everything since is level, suspect the platform rather than the show.
Check your host’s status page and release notes for that date, and check whether the fall is confined to one app. Most hosts break downloads out by listening platform; if Apple fell and Spotify did not, you have your answer.
A migration between hosting platforms is the other common cause of a genuine, mechanical drop, and it is the one you can actually fix.
When a show moves hosts, the old feed must 301 redirect to the new one, and every enclosure URL must resolve. If the redirect is missing, misconfigured, or removed too early, apps keep polling an address that no longer serves audio. Subscribers do not unsubscribe — they simply stop receiving episodes, which looks exactly like an audience disappearing.
Check that your feed validates, that the old URL still redirects, and that a fresh request for your most recent episode audio returns a 200 with the full file. Also check whether an analytics prefix was added or removed during the move: adding one changes which system is doing the counting, and the two systems will not agree.
Hosts, prefixes and platform dashboards do not all count the same thing, so a drop can be an artefact of which number you happened to look at.
Under IAB v2.1, a request only becomes a download if enough audio bytes were delivered to play at least a minute, the requester is not on a known bot list, and the same listener has not already been counted for that episode in the last 24 hours. A host that tightens any of those — an updated bot list, for instance — will report fewer downloads for identical traffic. That is the standard working, and the lower number is the more honest one.
It is worth knowing which of your numbers is the audited one. Your host’s IAB-certified figure is what a sponsor will check. A third-party dashboard, a prefix, or a platform’s own in-app analytics may each be measuring something slightly different and none of them is obliged to match.
If the drop survives all of the above, the show is genuinely reaching fewer people, and the cause is nearly always at the top of the funnel rather than the bottom. Subscribers are reliable; new listeners are not. A quiet month of discovery shows up as a slow decline in per-episode downloads while the subscriber base holds.
The practical response is to put a specific episode in front of people who do not already know the show — which is measurable in a way that most podcast marketing is not, because the result lands in the same IAB-counted number you were worried about in the first place.
An overnight, permanent step down almost always means a platform changed how it counts or delivers, not that listeners left. Apple’s iOS 17 change is the best-known example: it stopped auto-downloading episodes for subscribers who had not played the show recently. Look at whether the fall is confined to one listening app, and whether the chart shows a cliff on one date followed by a flat line rather than a gradual decline.
It should not, if the migration is done correctly. The old feed has to 301 redirect to the new one and keep doing so for a long time, because apps poll on their own schedule. If downloads fall after a move, check the redirect first, then check that episode audio URLs return the complete file, then check whether an analytics prefix was added or removed — that changes which system counts, and two systems will not produce the same number.
Yes. Episode-to-episode variation of a third or more is ordinary, driven by topic, guest, whether anyone shared it, and the day it went out. What is not ordinary is a step change that holds for every subsequent episode. Variation is noise; a persistent new baseline is a signal.
Most of a typical episode’s downloads arrive in the first week, but they keep arriving for months, because apps fetch episodes when a subscriber next opens them and back-catalogue listening never fully stops. This is why comparing a three-day-old episode against a month-old one always makes the new one look like a failure.
You can add to them. Paid promotion pointed at a specific episode drives new listeners to that episode, and if the promotion is measured to IAB v2.1 the resulting downloads appear in your own host’s analytics alongside the organic ones. Insist on that reconciliation: a number that does not show up in your host did not happen.
Podbooster runs targeted ads that bring real listeners to a specific episode, and counts a download only when it meets the IAB v2.1 standard — so the number matches what your host shows. $0.80 per verified download, from $100, no contract. Most campaigns finish in two to five days.
Promote an episode →Built and operated by Straw Hut Media, an independent podcast production company.