Measuring in Privacy Environments (including Apple Private Relay & VPNs)

Description: This article provides information about Apple Privacy Relay changes and how they impact InnovidXP. It also provides an overview of Innovid’s planned improvements relating to IP obfuscation due to the industry-wide changes.


Background on Apple Privacy Relay 

Apple has been implementing a series of Intelligent Tracking Prevention (ITP) changes (see here and here) and, since 2021, has been rolling out privacy protection changes across markets. These changes impact our ability to track Safari traffic, specifically impacting the accuracy of attribution results by reducing the number of site visits and conversions that can be associated with ad impressions.

Almost all clients who use analysis to understand the TV-driven impact on web visits will be impacted. Depending on the type of measurement and outcomes models used, you may observe a change in attribution metrics from the point of the change.

The impact is hard to quantify as it varies depending on:

  • Geographic location
  • Roll-out speed of Apple’s Private Relay 
  • Particulars of individual campaigns
  • Mix of Safari’s users vs. others who visit your website 

InnovidXP has previously relied on IP addresses to identify different users visiting simultaneously. However, Apple Private Relay breaks this assumption. By obfuscating the IP addresses of Apple devices as they pass through the Relay, many hits from different visitors are condensed into fewer (Apple-defined) proxy IP addresses, particularly in cases where (hashed) IP addresses are used as identifiers. InnovidXP has adapted to these changes, and Google has indicated they are planning a similar change on Google Chrome, but the implementation date is unknown. We can extend our technical solution for Apple Privacy Relay to incorporate traffic from Chrome browsers. We expect other browsers like Microsoft Edge to roll out similar measures in the future.


Session Stitching

InnovidXP has changed how we construct visit sessions out of individual page requests. A visitor may visit multiple pages (i.e., page requests or hits) when visiting a brand's website during a session. We convert the individual page requests to sessions and link the sessions to impressions to provide attribution insights. This is called "session stitching".

How session stitching works

Customers place an InnovidXP web tracker on their website to measure TV-driven web responses. The new solution will use a 1st-party cookie placed by this web tracker as the primary identifier, allowing different users with the same proxy IP to be more readily separated.

2024-02-19_12-33-57.jpg


Subdomains

As explained in the above section, InnovidXP has an enhanced methodology, known as ‘session stitching’, that handles the obfuscation of the IP addresses. However, if an advertiser’s website uses this methodology and has a subdomain URL structure, this can cause unexpected consequences.

For example, in some cases where a subdomain is used within the website URL, the web tracker views it as a different instance i.e., two sessions are recorded instead of one where a visitor has traversed the website from one subdomain to another in a single session.

To support sites where a subdomain is used within the website URL, a subdomain-friendly web tracker can be used. This tracker allows the grouping of all visits to multiple subdomains in the same user session to that specific user.

Web tracking options

InnovidXP currently offers two options for web tracking: standard web tracking and the option to use a specific subdomain-friendly web tracker. Our web response tracker must be placed on every page of your website in both standard and subdomain tracking options, and in all cases, URL accuracy is crucial.

Our standard web tracking option is used as a minimum for tracking to take place. Most websites require standard website tracking only, e.g., yourwebsite.com/orders.

However, if your website uses subdomains, e.g., orders.yourwebsite.com, you need the subdomain-friendly web tracker.

Contact Support at support-xp@innovid.com for help with this.

Website changed to subdomains

If you are using our standard web tracker and your website has changed to include subdomains, contact Support at support-xp@innovid.com so that we can help you use our subdomain-friendly web tracker.


Methodology Attribution Upgrade for OTT/CTV Impressions

As indicated, these changes impact the industry’s ability to track Safari traffic, explicitly impacting the accuracy of attribution results by reducing the number of site visits and conversions that can be associated with ad impressions. 

Safari browsers on desktops, laptops, and mobile devices are impacted by the incremental rollout of these new ITP features, which can affect InnovidXP attribution products, including OTT/CTV, impression-based linear, and our probabilistic Spike model.

How the upgrade works

The upgrade enhances our outcomes methodology to capture TV-driven performance for OTT/CTV campaigns more accurately, giving customers a more representative view of the impact of their campaigns to support planning and optimization.

How this solution evolved

Our Data Science team has conducted extensive research and analysis on the proportion of impressions & response traffic we see coming from private relays and VPN servers. Using that data, we tested various models to upscale OTT view-throughs to properly reflect the amount of Apple Privacy Relay traffic and VPN traffic that may have been “missed”. After testing model validity throughout 2023, we have created a dynamic algorithmic model based on real-time data.  This allows us to provide a more comprehensive view of attributed outcomes.

What data does this change impact?

This impacts attribution metrics such as responses and actions for website and app. There are no changes to measurement metrics.

Who does this apply to?

All customers using OTT attribution with session stitching enabled must have OTT upscale enabled.

What does this mean for my outcomes data?

Sessions known to have occurred but are currently discarded can now be attributed. We anticipate that the measured outcomes will increase between 20-40% across your platforms.

What if I use a non-IP-based identity layer, e.g., Blockgraph?

This solution will only upscale OTT traffic using IP-based identifiers.


Related content

FAQs: Privacy and Security

Was this article helpful?
0 out of 0 found this helpful