Glossary

Rethinking lookalike audiences in the first-party future

Advertising
No items found.
What are lookalike audiences?
Written by
No items found.
Published on
April 28, 2025

Recommended reading

Your guide to reducing wasted ad spend using first-party data

An estimated 23-56% of ad spend is currently wasted (and that’s before third-party cookies are completely deprecated). So how can brands ensure they’re reaching their ideal audiences at a time when consumers expect more personalized — yet privacy-preserving — advertising experiences than ever before?

Key visual for guide to reducing ad waste

Although Google has reversed course on its promise to deprecate third-party cookies, third-party data quality continues to worsen — meaning conventional ways to prospect for new clients are dying.

This article dives into both traditional and emerging lookalike marketing approaches — particularly within data clean rooms — illustrating how to create lookalike audience segments that closely match your best customers.

Lookalike audiences: Definition
Lookalike audiences (also referred to as "lookalikes") describe a targeting strategy where brands serve ads to people who share behaviors and traits with their existing customers. This allows the company to extend their reach to a broader — yet similar — audience.
Magnifying glass with lookalike color samples
Find customers similar to the ones you already have with lookalike audiences

What is a lookalike audience? As their name suggests, lookalike audiences comprise groups of new individuals who share similar characteristics with a predefined audience. This original audience is known as the "seed audience." The seed or source audience typically consists of existing customers or engaged users. It serves as a foundation for identifying common traits and patterns within the organization’s customer base.

The process of creating lookalike audiences begins with carefully selecting appropriate seed audiences. This is normally done at a product or channel level, either because there is a belief that the consumer profile differs greatly or that the quality of data collection varies between channels. 

Lookalike models use data in a very different way than traditional targeting methods to determine which new audiences are the most promising. In traditional targeting, audiences are often chosen based on demographics like age and gender. A lookalike model is an advanced machine learning or AI calculation that analyzes all available pieces of data in the seed audience.

Then, it makes a set of decisions about what indicates a high-potential new customer — rather than just assuming that gender or age, for example, play a role. Factors could be browsing behavior, purchasing patterns, and more. The model then seeks out individuals in the prospective audience who share these traits. This is why having rich, reliable data on your seed audience is crucial to a well-performing lookalike — the more factors the model can calculate, the better quality the result. 

Beyond just data, the model used within the platform that identifies the lookalikes plays a vital role in pinpointing the traits of the best target audience. All this to say: The success of the model relies on both its sophistication and on high-quality data given to it.

Core considerations include audience size (commonly ~1,000 users for stable modeling on the open web), the breadth of signals incorporated (from past purchases to on-site interactions), and the expansion rate (the proportion of new potential customers targeted).

When should you use a lookalike audience?

A powerful marketing tool, lookalike audience targeting becomes indispensable whenever you aim to amplify your reach beyond an existing segment. Lookalike targeting is especially useful when launching new products, re-engaging lapsed customers, or expanding into new markets. By tapping into lookalike modeling, you can efficiently find a similar audience without time-consuming manual segmentation.

Lookalike audience example use case

An example of this in action could be an online apparel store that has just launched a new line of eco-friendly workout gear. They already have a segment of loyal customers who previously bought sustainable products from their site: people who care about ethical fashion and tend to spend more per order.

Instead of starting from scratch with broad targeting, they create a lookalike audience based on that loyal, sustainability-focused customer segment. They upload this first-party data into a platform, and the system finds new users who closely match their behaviors and demographics.

Now, the ads are shown to potential customers who are more likely to engage with the new product line — increasing conversion rates and reducing wasted ad spend.

How does traditional lookalike targeting work?

Before the cookieless era, lookalikes were mainly available on platforms within walled gardens. These platforms put their extensive user data to use to create sophisticated models. For instance, Facebook became synonymous with this approach through its "Facebook lookalike audiences" product. This offer allowed advertisers to enhance their targeting precision by reaching Facebook users who closely resembled their existing audience. 

Since 2018, Facebook has required advertisers to use their own data to create lookalike audiences rather than that from third-party data providers. We’ll dive deeper into the implications of this in the section “Walled garden lookalikes vs. data clean room lookalikes”.

In addition, various demand-side platforms (DSPs) apply their own methodologies, enabling their users to create lookalike audiences. For example, some DSPs use machine learning algorithms to identify patterns within the seed audience. The algorithms then map these patterns to a broader audience. Other DSPs may prioritize contextual targeting, focusing on the content and context associated with the seed audience to identify similar individuals.

While walled gardens excel within their closed ecosystems, DSPs offer a more customizable, expansive approach to lookalike creation. However, the methods they used to create lookalikes have traditionally relied heavily on third-party cookies. This creates significant gaps for the ability of advertisers to create lookalike audiences outside walled gardens.

How to create a lookalike audience without third-party data?

In an era of heightened consumer awareness and strict privacy regulations, striking the right balance between harnessing the power of lookalike audiences and respecting data privacy is more important than ever.

As third-party cookies become less dependable, brands must harness first-party data like CRM records, loyalty programs, and on-site event tracking. As highlighted in the previous section, it’s possible to use proprietary data to create lookalikes in walled gardens. But for those who want to leverage lookalikes on the open web, a data clean room can be the most accessible option.

How lookalike audiences work in a data clean room

Brands wanting to dedicate advertising budgets to open web campaigns (as opposed to relying only on walled gardens for the creation of their lookalike audiences) face the challenge of adapting to a more privacy-centric landscape. And that’s in addition to the negative effect on both the accuracy and reach of lookalike targeting on the open web caused by poor-quality third-party data.

This is where a data clean room with lookalikes comes in. With these solutions, you can use your first-party data to expand your reach through advanced AI lookalike models and create high-converting audiences.

However, not all clean room providers go about this in the same way. Often, data clean room (DCR) users are faced with tradeoffs between:

  • DCRs where a trusted party, either a publisher or the data clean room provider itself, has control over all the data — this requires the other party to trust them and their promise to keep the data private and secure
  • DCRs with limited or rigid identity matching capabilities, limiting audience reach, and risking privacy exposure in the segments they produce
  • DCRs based on complex cryptography methods that require a team of encryption experts to work with and offer little analytical flexibility

Decentriq offers a lookalike data clean room solution for brands, publishers, and retailers who don’t want to sacrifice privacy and control for usability. Read the case studies below to find out how well-known brands have unlocked privacy-first addressability with Decentriq.

Lookalike clean room case study: Samsung

Samsung relied on a Decentriq clean room to activate GDPR-compliant lookalike targeting across multiple publishers, using its first-party CRM data combined with netID-based publisher audiences. This approach produced 13 privacy-safe lookalike segments, reaching over 1 million potential new customers and driving scalable cross-publisher advertising — all without relying on third-party cookies. Read the case study.

Lookalike clean room case study: Major Swiss bank

​A leading Swiss bank transitioned from traditional third-party cookie-based targeting to Decentriq's Lookalike Clean Rooms, using its first-party customer data to create AI-driven lookalike audiences across Goldbach’s extensive Swiss publisher network. This strategy resulted in a 129% increase in click-through rate, a 57% rise in page views, and a 44% reduction in cost per page view — all achieved with end-to-end encryption and full data confidentiality, ensuring compliance with stringent financial industry regulations. Read the case study.

Walled garden lookalikes vs. data clean room lookalikes

It’s understandable — these are turbulent times, and marketers are under pressure to do more with less. Adding a new product to their marketing stack is not a top priority right now, meaning the first instinct for many marketers would be to use the lookalike audiences feature available on many walled garden platforms. This way, they can simply upload their first-party data and quickly receive new target audiences based on characteristics identified within their original seed audience.

But this method has two major shortcomings:

  • Users don’t spend the majority of their online time in walled gardens — they spend it on the open internet, like premium online newspapers or CTV

To be specific, they spend 34% of their time in walled gardens and 66% on the open internet.

  • Sharing your proprietary data with an outside entity means you no longer have control over what happens to it.

By braving the world outside walled gardens, advertisers are more likely to be seen and won't have to compete for attention. Additionally, using a solution with privacy guarantees by design makes an important statement about how an organization prioritizes its customers’ data.

Summing up the options available as they apply to most advertisers’ major concerns: privacy, precision, and reach.

Decentriq enables privacy-first lookalike marketing

With Decentriq’s lookalike audiences, first-party data is verifiably never accessible to any party, not even to Decentriq. It enables advanced matching and lookalike audiences, immediately actionable from within a no-code SaaS environment. 

Setup takes just minutes — no support needed from an engineering team. Designed for business users, you can get started uncovering insights right away by inviting publishers you work with directly to the clean room or selecting from our network of publisher partners.

Thanks to Decentriq’s unique combination of confidential computing technology and privacy-by-design approach, brands can tap into their entire body of first-party data for marketing in a compliant way, with greatly reduced legal hurdles compared to other data collaboration solutions. 

With Decentriq, only encrypted data enters the platform, and only approved analytics leave — which means brands can keep their first-party customer data fully private while still running powerful, precise first-party marketing campaigns in collaboration with agencies, publishers, and retailers. Watch a demo below:

“Decentriq has, for a while, taken a principled approach to clean rooms. For example, ensuring their isolation from 'the bottom up', guaranteeing that no parties have the ability to see everything in the clear. Applying their principled approach to lookalikes nicely provides required privacy guarantees to all parties.” — Igor Perisic, former VP Engineering and General Manager of Ads Privacy and Safety at Google

Discover how Decentriq can transform your lookalike strategy

The decay of third-party cookie quality need not spell the end of effective lookalike marketing or use of walled gardens exclusively; rather, it opens the door to privacy-first, transparent, and brand-owned lookalike modeling. Data clean rooms, especially when powered by Decentriq’s secure platform, offer a scalable path to build high-fidelity lookalike audiences, drive superior ROI, and uphold stringent data governance.

Ready to elevate your prospecting efforts? Discover how Decentriq can transform your lookalike strategy with our advertiser data clean room solutionsget in touch to get started on privacy-first customer growth.

References

Recommended reading

Your guide to reducing wasted ad spend using first-party data

An estimated 23-56% of ad spend is currently wasted (and that’s before third-party cookies are completely deprecated). So how can brands ensure they’re reaching their ideal audiences at a time when consumers expect more personalized — yet privacy-preserving — advertising experiences than ever before?

Key visual for guide to reducing ad waste

Related content

Subscribe to Decentriq

Stay connected with Decentriq. Receive email notifications about industry news and product updates.