Downloads and impressions are linked, but they are not exactly the same. A download, according to the IAB definition, is a downloaded audio file that lasts at least one minute and has been filtered to exclude invalid and "bot" traffic and occurs in a single day. It doesn't matter what the duration of the complete podcast episode is; once one minute or more of the podcast has been downloaded, it counts as a download.
An impression, on the other hand, is defined by the IAB as an audio advertisement that has been downloaded in its entirety within a podcast download.
In other words:
A download is at least one minute of podcast episode.
An impression is a full ad that has been downloaded within a podcast download.
So the ad (the impression) must be downloaded entirely, but the episode (the "download") does not have to be entirely downloaded. This difference becomes noticeable when you factor in things such as:
The duration of podcast episodes.
People's listening habits and patterns.
Different players and how they handle interrupted downloads.
Let's dig a bit deeper into those points.
Podcast Duration
Short duration podcasts are less likely to have download/impression discrepancies because listeners are more likely to listen to the entire episode in one shot. But podcast episodes often run for an hour or more, in some cases even three hours. Listeners of those long podcasts are more likely to listen to each episode in two or three sessions, which can cause differences between download and impression data if the sessions are not on the same day, as we'll see in more detail in the examples below.
Listening Habits
Aside from listeners splitting their episode consumption across two or three listening sessions, there is also a difference in how people download their podcasts. Some listeners download via a podcast app and might even subscribe and download new episodes automatically in the background. This means the episodes are almost always downloaded in full, in a single shot. Downloading and listening are completely separate functions, because the listener plays the download from the stored file on their device days, weeks, or even months after they were downloaded.
Other people listen to podcasts directly, meaning they listen as the episode is being downloaded to the device. This is how smart speaker users typically consume podcasts, but listeners can also use this method with mobile or desktop players. In this case, downloading and listening are linked; if the listener stops listening before the episode is finished, the audio file might not have fully downloaded. When the listener resumes listening it can trigger a second download that starts in the middle of the episode, for what amounts to a single "listen" across two sessions and two downloads. (See the examples below for a more detailed view of this.)
Player Behavior
Not all players handle interrupted downloads the same way. Some continue downloading the entire episode after the listener stops listening and some truncate the download when the listener presses Stop. Different players have different cache methods, so a partially-listened to but fully downloaded episode might play from cache the next day or it might be flushed and re-downloaded.
Examples
The examples below show how the points described above can play out in the real world of podcast downloads. In every case, the example refers to listeners using direct downloads, not episodes downloaded in advance via podcast app or subscription.
Example 1: one-hour episode listened to over two days.
Listener opens podcast player, selects a new episode, and hits Play. She hears a preroll and then the podcast starts.
She listens to 5 minutes of a 60-minute podcast and then closes the app.
The player has downloaded 30 minutes of the 60-minute episode before it was interrupted.
Next day, the listener opens the podcast player and resumes listening to the episode.
The player notes that it has only part of the episode and downloads the final 30 minutes.
Result:
Listener heard one pre-roll (impression).
Listener heard one full episode of the podcast (over two sessions).
IAB marks this as two downloads (as it occurred over two days).
The second download started in the middle, so it did not include the pre-roll.
Final tally: two downloads, one impression.
Example 2: long episode listened to over three days, with pre-, mid-, and post-roll ads.
Listener starts a long podcast on Monday. He hears the pre-roll then listens for 20 minutes and stops.
On Tuesday he listens for another 20 minutes, hears the mid-roll, then stops again.
On Wednesday he finishes listening to the podcast and hears the post-roll.
Result:
This could count as up to three downloads, depending on how much of the episode was downloaded before the listener pressed Stop, and how the player deals with incomplete downloads.
Regardless of the download count, there is one pre-roll impression, one mid-roll impression, and one post-roll impression, as each of those ads were downloaded only once.
The impression data is aligned with listener behavior, but not with what you might expect as download count. (You would expect one download, three impressions, but might see two or three downloads and three impressions.)
Example 3: long episode with a single mid-roll listened to over two days.
Listener listens to first 30 minutes of a two-hour podcast then stops listening.
60 minutes of data is downloaded during the 30 minutes of listening.
The midroll impression at the 45-minute mark is recorded as delivered, as it was completely downloaded, although the listener has not yet heard it.
The next day, the listener listens to the final 90 minutes of the two-hour podcast.
The player discards the previously downloaded data (including the unheard mid-roll) and reloads the podcast starting at the 31-minute mark. The re-download includes another mid-roll.
The mid-roll impression at the 45-minute mark is recorded as delivered, as it was completely downloaded.
Result:
Although there was only one "listen," it is counted as two downloads.
The mid-roll was heard only once, but it was downloaded twice, so it is counted as two impressions.
Example 4: same as 3, but the two listening sessions occur on the same day.
Listener listens to first 30 minutes of a two-hour podcast then stops listening.
60 minutes of data is downloaded during the 30 minutes of listening.
The midroll impression at the 45-minute mark is recorded as delivered, as it was completely downloaded, although the listener has not yet heard it.
Later the same day, the listener listens to the final 90 minutes of the two-hour podcast.
The player discards the previously downloaded data (including the unheard mid-roll) and reloads the podcast starting at the 31-minute mark. The re-download includes another mid-roll.
The mid-roll impression at the 45-minute mark is recorded as delivered, as it was completely downloaded.
Result:
Although the episode was downloaded twice, it is counted as one download because both downloads were on the same day.
The mid-roll was heard only once, but it was downloaded twice and is recorded as two impressions.
How prevalent is this kind of discrepancy?
Triton Digital has analyzed the data for millions of download sessions. While the data varies widely depending on podcast length and other factors, the general guideline is that 15-25% of all downloads start in the middle, as described above. This means that you cannot rely on download data alone to provide accurate guidance on impression data.
Keep in mind that IAB does not allow for split content on impressions, so there is no need for concern that a split episode puts your impressions at risk. All delivered impressions will be correctly recorded, even if they don't correspond directly with listening (see examples 3 and 4, above).
Good news: this isn't really a problem
Triton Digital tracks ad requests and impression delivery directly at the content delivery source, so our forecasting and delivery engines are already optimized for IAB impression delivery. You don't have to worry about download or multi-day session variances when tracking impressions. The platform cleans that up for you.
Similarly, Triton Digital uses sophisticated download session management to track when multiple downloads are part of the same listener request, even if they are spaced well apart. As a result, we can effectively deliver both a single listener experience across multiple listening sessions through all major podcast apps and players and also support the IAB standard that requires a limit on the timeframe of those download sessions. Result: good listening experience for the listener, and good impression tracking for you.
The key thing to remember if you're selling advertising based on impressions and trafficking via dynamic ad insertion is to avoid confusion and other issues among your stakeholders by not using downloads to track ad delivery or to forecast ad opportunities. Because as outlined above, a download is not the same as an impression, and is not the same as a single listening session.