Why podcast numbers are slippery
Podcasts are distributed in a way that makes counting unusually awkward. Most shows publish an audio file and a feed, and listeners subscribe or download through many different apps and websites. There is no single gatekeeper that sees every play, as a streaming service or broadcaster might. The result is that “how many people listen to podcasts?” and “how big is this show?” are answered by several different measurement systems that do not agree.
The main ways listening is counted
Server downloads. The host that stores a show’s audio can record each request for a file. Download counts are the oldest and most widespread podcast metric. They are objective in the sense that they are logged by a machine, but they measure file requests, not people listening. An episode can be downloaded automatically to a phone and never played, or played several times from one download, or fetched twice by the same person on two devices.
Filtering and standards. Because raw requests include repeats, bots, preview fetches and partial downloads, hosting companies and industry groups have developed rules for filtering. Typically these involve counting a unique IP address and device combination once in a set period and discarding known automated traffic. Different providers apply different rules, so two services counting the same show may report different totals.
App and platform data. Some listening apps can see play events, the proportion of an episode completed and whether someone follows a show. These are closer to actual listening, but each platform sees only its own users, and the data is usually shared selectively and in its own format. This is a clear example of the limits of platform data.
Audience surveys. Survey firms ask representative samples whether people listen to podcasts, how often, and which shows. These can estimate reach across a population and describe who listens. They rely on recall and are affected by how people interpret the question, a standard concern with self-reported behaviour. People may confuse podcasts with radio programmes, video channels or audiobooks.
Panels with meters or apps. Some research companies recruit panels whose listening is tracked on devices. This improves accuracy for the people in the panel, but panels are limited in size and may not represent all listeners, particularly for small niche shows.
Downloads, listeners and reach are different things
A podcast figure can refer to quite different quantities:
- Downloads per episode: how many file requests an episode receives, usually within a specified window after release.
- Unique listeners: estimated number of distinct people, which has to be inferred and is imperfect because one person may use several devices and several people may share one.
- Subscribers or followers: people who have chosen to receive new episodes, some of whom have stopped listening.
- Reach: the share of a population that listened to a show or to podcasts in general over a period.
- Completion or retention: how much of an episode is heard.
Quoting a total of downloads over the lifetime of a back catalogue sounds impressive but is a different number from the audience for a new episode. Likewise, “monthly listeners” from one service is not comparable with “downloads per episode” from another.
What growth claims can mean
Statements that podcasting is growing rapidly can reflect several different things.
More people listening. Survey reach may rise as more people try podcasts. This is the claim most readers assume, but it is only supported if the same question was asked consistently over time with comparable samples, as in tracking surveys.
More shows. The number of available podcasts can increase without a matching rise in listeners. Many shows are small, and a large catalogue says little about how audiences are distributed.
More measurement. As hosting platforms improve or add tracking, previously uncounted listening can become visible. An apparent jump may then partly reflect a change in how data is collected, a point related to platform rule changes affecting trend data.
Revenue growth. Advertising spending on podcasts may grow because advertisers shift budgets, not necessarily because audiences grow at the same pace. Forecasts of future advertising are projections, not measurements.
Changes in definition. If video versions of podcasts on streaming sites are included in “podcast listening” in one year and not another, trends can change shape without any change in behaviour.
Why comparisons across shows and sources fail
Comparing the audience of two shows is difficult even when both report downloads. They may use different hosting companies with different filtering rules, different measurement windows (the first seven days versus the first thirty), and different release patterns. Shows with daily episodes and shows with weekly ones produce very different numbers per episode and in total.
Self-reported statistics from creators add another layer, since figures in media kits are often prepared to attract sponsors and may select the most flattering period or metric. This does not imply dishonesty, but it is a reason to ask which number was chosen and why.
Audience composition matters too. A show with a small but highly specialised audience may be valuable to certain advertisers despite modest download numbers, so raw size is not the only relevant measure.
Reading a podcast statistic
Some sensible checks:
- Is the figure downloads, listeners, subscribers or survey reach?
- Over what time window was it measured?
- Who produced it, and what method and filtering did they use?
- Is it a single show, a platform or a whole country?
- Is a growth claim based on consistent measurement across years?
Podcast data is still evolving, and no single number captures listening. Treating figures as estimates produced by particular methods, and not as exact counts of attentive listeners, avoids most of the common misunderstandings.