In Podcast Insights, you'll find a variety of analytics reports to help you understand when, where, and how your podcast audio is being consumed. This includes data from both owned and operated players as well as third-party distribution channels.


Due to technical limitations such as caching, IP recycling, and increasing privacy regulations, additional processing is required. Below, we explain our methodologies for calculating Downloads, Uniques, and additional Filtering. These methods have been certified by the IAB Tech Lab.


As the landscape evolves, we may refine our filtering parameters and experiment with new methodologies to improve our metrics. This document will be updated periodically to reflect any changes.


IAB Tech Lab Podcast Measurement Independent Certification


Download

A download is counted in response to a user-initiated action to request the podcast file directly from SoundStack’s CDN. This can be done via owned and operated players or third parties like Apple Podcast, Spotify, and more. A download is counted based on an HTTP GET request and lives for up to 24 hours. For requests from the same user for the same podcast episode within this window, requests are de-duplicated to show the correct number of “true” downloads.


SoundStack uses a fixed time window of 24 hours, as required to comply with the IAB Tech Lab Podcast Measurement Guidelines. 


Only downloads that fully make it to SoundStack’s CDN are possible for Dynamic Ad Insertion and Podcast Insights. For third-party players that cache the podcast content, the user request does not make it to SoundStack, and is thus not included in Podcast Insights. This issue is not unique to SoundStack, and is an industry issue that is constantly evolving.


Uniques

Unique listeners are computed based on:

  • IP address of the user (IPv4 or IPv6)
  • User agent of the user
  • Date, in UTC


Unique listeners are counted using HyperLogLog, an algorithm used to approximate the number of distinct elements in a large data set with less than 2% error margin. HyperLogLog, or HLL, is used by many analytics platforms and is industry-accepted.


Time Window for a Unique Listener

Because of the nature of podcast consumption with a high percentage on mobile devices and residential ISPs, IP addresses of each listener change frequently, by day or even hour, and those same IPs can be re-used by new listeners. To account for this, by default, we consider a unique user also within a 24-hour window, following the IAB Tech Labs guidelines. 


Recycling of IPs

IP’s are typically recycled for a single listener in three ways:

  1. Through third-party services used to protect user identity, such as a proxy or VPN (ie Apple Relay)
  2. NATing (Network Address Translation), where many users are behind a corporate network or using an ISP with limited IPv4 space
  3. Mobile and residential networks, where the IP Address is “leased” and given back to the provider at a specific interval or hopping between cell network towers.


Please discuss with your customer representative or customer support for those who wish to increase or remove the 24-hour window for unique users. This is configurable on a per-organization basis.


Filtering

We do not count downloads (or make requests for audio ads) from known lists of third-party server providers, such as AWS, Cloudflare, DigitalOcean, Google Cloud, OVH, and more. This list is maintained internally and updated regularly every month. To update the list, we look at large sources of traffic requests correlated to ASNs (Autonomous Systems Number) to determine if the traffic is originating from data center IP ranges, and if those IPs belong to legitimate services like VPNs.


Additionally, we utilize OPAWG User-agents-v2 for correlation of User Agents within Podcast Insights and blocking commonly known bot user agents. This list is also refreshed on monthly intervals.


If you have any further questions or concerns please do not hesitate to contact our Support Team through the ticket submission form or by emailing us at support@soundstack.com.