1. Ad Analytics Help Center
  2. About
  3. About Attribution Methodology

About the Device Graph used for Attribution

Spotify Ad Analytics utilizes a device graph covering Australia, Canada, Mexico and the U.S. in order to determine which IP addresses are from a household, as well as household reach.

Why do we use a device graph? 

Most digital tracking is based on an individual identifier that is associated with a user’s web browser or device. There are no device or browser-level identifiers in podcasting, as identifiers such as hashed emails, cookies, and device IDs are not available. Because of this, Spotify Ad Analytics relies on IP addresses in order to attribute impressions to website visits and app engagement. You can learn more about the types of data collected for attribution here

For attribution to be accurate, we have to consider the connection type of the IP addresses across the exposed audience. This becomes challenging when many people are associated with a single “noisy” IP address (e.g. an office building, university campus, public WIFI network, etc). We consider household connections to be the most precise available, as they are much more likely to have the same people connecting to them over time.

How can a device graph help? 

Spotify Ad Analytics uses a device graph in order to sift through the noise and identify household IP addresses. IP addresses that do not match to the device graph may still be considered a household if the connection type is not a cell-tower or commercial IP address.Spotify Ad Analytics monitors those IP addresses from the Spotify Pixel to ensure they are behaving like a household. 

For those IP addresses that do not match to the device graph and are a cell-tower or commercial IP address (i.e. noisy IPs), Spotify Ad Analytics uses a shortened attribution window to remove false positives . In order to showcase the effectiveness of a full campaign across all IP address connection types, Spotify Ad Analytics models results for impressions that fall outside of the identified household IP addresses. You can learn more about modeling here

Frequently Asked Questions

  • Where is the device graph available?
    This device graph is available in Australia, Canada, Mexico and the U.S.
  • How does attribution work in markets where the device graph is not available?
    In markets where Spotify Ad Analytics does not utilize a cross-device graph, we look at the total pool of IP addresses and the connected device count associated with each to identify IP addresses that behave like household IP addresses, in order to perform attribution. 
  • What is the average match rate for your device graph?
    Though it varies based on the campaign, we generally see a 40-80% match rate. We offer both an unmodeled and a modeled view of campaign performance that accounts for activity from non-matched IPs.
  • Who powers the device graph?
    Our device graph is powered by Tapad.
  • How does the device graph help to calculate reach?
    Reach is calculated by using a weighted average of IP frequency and household frequency. The calculation is:
Frequency = (1.5*{Household Frequency} + {IP Frequency})/2.5.

Household frequency can be calculated by looking at the average frequency of ad exposure for household IP addresses, which are determined by device graph matches and non-noisy IP addresses.