1. What challenges have you faced in integrating identity solutions into your audio advertising platforms?
The most complex challenge is creating the right configurations that consider the large spectrum of devices and playback mechanisms. By that, I mean if you are listening to a podcast on your iPhone, you are likely within the native player Apple provides with the phone. In this case, there is no data outside of an IP address and the content you are listening to if it’s delivered by one of Triton Digital’s platforms, Omny Studio or Spreaker.
Faced with multiple different solutions for podcast delivery, live streaming, and other methods used to insert and deliver audio advertising, the complexity grows very quickly. There simply isn’t a “one size fits all”, making it almost impossible to apply the same recipe or solution to everything.
Additionally, within the publisher’s applications, the limitations and level of integration are all different making this integration more challenging but just as necessary. We have found ourselves often adapting and re-implementing various SDKs to enable the complete audio ecosystem to maximize ad opportunities for our publisher clients.
Furthermore, when it comes to identity, finding the right solution for different types of delivery methods is essential to achieve maximum coverage.
Existing methods and tools, like IP+UA fingerprinting and external datasets, must be combined with other techniques, such as building a listener profile at an entry point that can then be used along the advertising chain. Ultimately, we take the best information available coming in and extend the listener profile to include as much information as possible; at the end, when we have identifiers to insert into the programmatic bidstream, we know we accomplished the enrichment of that profile to the best level possible.
2. What role does user identity data play in enhancing the effectiveness of programmatic advertising?
At a high level, it helps increase addressable audience ultimately providing more value to advertisers.
To break it down more, it’s important to understand the different types of identity that apply and are available. Most boil down to deterministic identity and probabilistic identity, each with different degrees to them.
Probabilistic Identities are a “best guess” approach. Meanwhile, cookies, MAIDs (known as mobile advertising IDs), and IP addresses are deterministic in nature but are still probabilistic because you cannot uniquely identify an end user, a household, or any true uniqueness to them. A listener, for example, could be listening in the car going from 5G to Wi-Fi as they stop to grab coffee and then on another Wi-Fi network at work. The goal of leveraging both types of identities is to build a profile that is consistent and can help find the common thread in that listener’s journey.
Using solutions like ID5, we will capture the IP along with any other identity details captured by the cookie or MAID, which will then be used to determine the best set of IDs to insert into the bidstream that would represent the consumption pattern of a listener. DSPs do the actual identification part and determine most properties of a listener, while our role is to make sure they have the best “passport” for the listener.
Deterministic identifiers on the other hand are based on generally identifiable information, like an e-mail address or other information. This requires a completely different process and method that ensures the underlying data is protected and private, and that user consent was provided, even though the resulting profile is derived from that information.
Both of the identifiers play an essential role in rendering the listener profile much more addressable and valuable to advertising buyers. Highly deterministic identifiers are applied to publisher applications with logged-in listeners and the probabilistic identifiers are applied in all other situations. Both of them cover different portions buyers are trying to reach and the CPMs they are willing to pay to do so.
3. How has the implementation of identity management solutions impacted your programmatic ad revenue?
I won’t dive into figures, but generally, increasing addressable audiences through the use of identity management solutions increases bid confidence which can yield better results, “The more you seed, the more you can harvest.”
To break it down more, probabilistic IDs boost general addressability, while deterministic identifiers play a much bigger role in bid confidence. This means the frequency of bidding is much higher than the addressability boost as these only apply to a much smaller subset of listeners (those who use the publisher’s applications, have a login, share their e-mail, and consent to share their information).
A DSP might privilege confidence vs addressability depending on the different buying profiles in their systems. This causes different DSPs to bid more constantly on different types of inventories, which widens the spectrum of possibilities.
4. What measures are in place to ensure compliance with data privacy regulations when managing user identities or user data used to enhance addressability?
We always strive to exceed industry standards for data privacy and compliance whenever possible, while still adhering to established compliance guidelines. This is particularly true when it comes to dealing with user identities and user data. Notably, for many solutions, an identifier cannot be created without having some signal representing the consent of the end-user.
Furthermore, any identifier that is based on deterministic data like e-mail, passes through a more vigorous process where we ensure we don’t share any emails or hashed emails (otherwise known as a simplistic encryption), nor access emails.
For this particular process where we’re integrating identity solutions, we use a 3rd party cleanroom made possible through our partner Optable, which allows the publisher to directly encode deterministic data into a UID2 token. This adds an extra layer of protection for the publisher’s data.
5. What best practices do you follow to navigate the complexities of data collaboration in the audio advertising industry?
When it comes to navigating the complexities of data collaboration, I always try to stay open and ahead of the curve. Cleanrooms for, example, are still fairly new to the audio industry but it’s ideal for data collaboration, data matching, and other types of operation. We believe the use of cleanrooms will continue to grow over the next few years, as advertisers and publishers continue to look for ways to improve targeting and retargeting at scale.
One of the growth drivers for the adoption of cleanrooms we see is the promising use in retail, where e-mail-based identifiers are becoming more sought after. For example, a retail business has registered accounts from users who buy online on their website. An audio publisher also has registered accounts that consume audio on their media players. Using cleanroom data matching it’s possible to compare and match the different e-mails and extrapolate identifiers that represent the common audience in a privacy-safe manner where the other’s information isn’t visible to the other party. This is known as a zero-knowledge proof.
In the end, you get a list of UID2, for example, and can use that list as targeting criteria and no party (not the advertiser, publisher, or Triton) transferred any e-mails over an unsecured connection, nor were any e-mails shared.
Sounds like magic? Yes, it’s the magic of mathematics; ZKPs (Zero-Knowledge proofs) are an amazing tool at our disposal that allows for a more privacy-safe advertising industry.
