How Pandan counts downloads

Reference · Last updated: 5 November 2026

A download is one listener fetching one episode on one day. This page lists the rules Pandan applies to get there, in the order it applies them. They follow the IAB Podcast Measurement Technical Guidelines (version 2.3, September 2026); section numbers below refer to them. Your plan is based on this number.

For why it can differ from what other services show you, see Why download numbers differ.

Terms

Request: one request from an app or browser for an episode’s audio file.
Listener: one IP address with one app or browser, as named by its user agent (§6.2).
Download: one listener fetching one episode on one day (§6.1).

Five requests for the same episode from the same app on one day are one download. On five different days, they are five downloads.

Where requests are seen

The audio addresses in your RSS feed and on your podcast website point to media.pandan.fm. That server notes each request and passes the app on to the file in storage. Only requests that reach it can count.

The rules

Published episodes, full requests. Only requests that ask for an audio file (GET) count, not those that only ask about it (HEAD, §5.5.2). The file must belong to a published episode, including an audio file the episode had before it was replaced.
No bots. Requests without a user agent, from bots on the OPAWG user agent list, and the duplicate requests of the Apple Watch don’t count (§5.5.2).
One download per listener, episode, and day. All requests for one episode on one day from the same IP address and the same user agent are at most one download. Of the windows the guidelines allow, Pandan uses the calendar day in UTC, from midnight to midnight (§5.5.1). IPv6 addresses are shortened to their first 64 bits before this (§5.5.4).
Enough of the file. Those requests are a download if at least one of them asks for the whole file, for the file from some point to its end, or for any part larger than two bytes. A request for only the first two bytes counts when it comes from your podcast website on Pandan, where a browser sends it when someone presses play. From anywhere else, it doesn’t count, as in §5.5.2.
No implausibly busy addresses. An IP address with more than 1,000 downloads on one day counts for none of them (§5.5.2). We will set this number from real traffic.

When it’s counted

Pandan counts every 15 minutes, so today’s number is still filling up. A month is a calendar month in UTC. Account settings shows your account’s downloads this month and last month, in all and per podcast.

Pandan keeps the requests for 90 days, with the IP address replaced by a hash that changes every day. That lets us count a day again if a rule changes. After 90 days, only the number of downloads per episode and day remains. The privacy policy describes this in full.

Where we differ from the IAB guidelines

We see what was requested, not what was delivered: the file itself comes from storage (§5.2).
The guidelines count a part of the file once it covers one minute of audio. We count any part larger than two bytes; every one we have seen so far was far longer.
The guidelines leave out requests from data centers. We don’t yet, because it needs a database of networks that we don’t have.