Data clean rooms are not the endgame for identity
The industry needs to avoid falling into the trap of seeing clean rooms as a substitute for identity, argues InMobi‘s Todd Rose.
Clean rooms can't substitute identity, argues InMobi's Todd Rose / Adobe Stock
Over the past two years, data clean rooms have exploded onto the programmatic advertising scene. Interest in clean rooms has been mostly fueled by the rise of privacy legislation and operating system-level privacy changes that necessitated more advanced and secure ways for parties to exchange and match data in order to protect consumers.
Meanwhile, these same market forces have led to a significant loss in identity signals. As a result, many industry observers have conflated solutions designed to solve for data privacy with solutions designed to compensate for signal loss. So, it’s important to level-set on an important but often overlooked truth: clean rooms can serve as a solution for data privacy and security, but they are not an identity solution.
Data clean rooms provide a neutral space for two or more parties to match, merge, or segment their first-party data; to create new audiences, enable attribution and perform analytics – without ever disclosing any party’s personally identifiable information to the others.
While clean rooms are a useful tool to prevent leakage of proprietary data, they cannot magically manufacture identity signals. In other words, if you don’t have sufficient permissioned identifiers against which to match your data in the first place, a clean room won’t do you much good. Clean rooms are certainly a component in a privacy-first world, but they do not resolve identity on their own.
I might go so far as to say that the current momentum in the clean room space is driven more by fear than function as the industry grapples with answers to the inevitable question from marketers: ‘What are you doing to compensate for the loss of third-party cookies, Apple's identifier for advertisers (or IDFA) and, eventually, Android ID?’
Clean rooms provide a convenient and superficially satisfying, yet ultimately incomplete, answer.
Retailers, media companies, and walled garden platforms have jumped into the fold. They’ve either crafted homegrown clean room offerings or partnered with independent third-party clean room players. But despite all the dialogue and hype, clean rooms haven’t seen as widespread adoption in practice as many would have you believe.
Part of the lack of adoption is inertia. If identifiers still exist – as they do at this moment – then marketers will continue to depend on them. The longer-term issue with today’s clean rooms is that they require you to trust an intermediary to steward and protect data. While it’s better than trusting a non-neutral counterparty, it still means handing over the reins to your data, and it’s still vulnerable to data breaches.
There is a high likelihood that as the market drags its collective feet on clean room adoption, solutions like multi-party compute (MPC) – which eliminates the need for trust entirely – will leapfrog the current crop of clean room players for many applications. MPC can offer some unique benefits and remove some of the barriers that have plagued clean room adoption. MPC is structured as a peer-to-peer system that creates a virtualized environment. Unlike traditional clean rooms that rely on a trusted intermediary to steward proprietary data, MPC does not require any proprietary data to ever reside outside a data controller’s systems in an unencrypted form. This makes MPC truly “trustless.”
Corroborating this hypothesis, a number of the walled gardens – including Google and Facebook – have been very public about their investments in MPC. Moreover, the code base for MPC algorithms has been open sourced by multiple parties, which will certainly help streamline and facilitate adoption.
As MPC matures, existing data operations teams, which are typically already fluent in implementing pre-defined encryption algorithms, will be able to implement them without significant investments in data science or engineering resources. MPC solutions will eventually have simplified front-end interfaces that will allow ease-of-use for even the most modestly technical of users.
Suggested newsletters for you
MPC will likely gain traction as one of the primary means to match two or more data sets and likely supplant data clean rooms for a large percentage of use cases, particularly for attribution and insights analyses. Marketers might find traditional clean rooms more effective when matching multiple data sets or when a multitude of variables need to be passed through and matched.
Before today’s marketers decide to address their identity issues with a clean room, take a look at what’s on the horizon. Familiarize yourself with publicly available tools for MPC and the requirements to leverage it. Assess your use cases to determine if your needs are simple or complex, and make sure you have the right resources and capabilities for your data operations teams.
Todd Rose is senior vice-president of identity solutions at InMobi.