Which Lookalike Audience Ad Strategies Have Demonstrated ROI?

Lookalike audience ads can report positive ROI, but results may reflect other campaign changes. Learn how to choose seeds and test performance.

Lookalike strategies have reported positive returns, especially when seeded with customers and used in conversion-focused campaigns. But platform case studies usually combine targeting with creative, bidding, or other changes, so they do not prove lookalikes alone caused the result.

Which lookalike audience ad strategies have demonstrated ROI?

The clearest reported results come from lookalikes built from customers or customer activity and run in campaigns optimized for purchases. TikTok’s case studies report these examples, but treat them as directional: they are platform-published results, and the campaigns bundled lookalikes with other tactics.

Strategy Reported result What else was involved
Customer-based lookalikes with product shopping ads Bask and Lather reported 9x average ROAS over two weeks. (ads.tiktok.com) The brand also used organic-content ads, affiliate creators, and TikTok’s campaign recommendations.
Existing-customer lookalikes in a purchase campaign Endy reported 141% higher average ROAS and 58% lower CPA than in the previous six months. (ads.tiktok.com) The campaign combined lookalikes with interest targeting, creative features, and value-based optimization.
Website-visitor and engagement lookalikes Refy Beauty reported nearly 3x ROAS. (ads.tiktok.com) It also used retargeting, native-style creative, and a multi-channel campaign.

These results support testing lookalikes for customer acquisition, particularly when the seed reflects purchase intent or customer value. They are not proof that the audience type alone generated the reported return; ROAS also does not show profit or incremental sales. If you have little first-party data, or broad targeting already performs well, a lookalike may add little and can restrict reach.

Which customer data sources make the strongest lookalike audiences?

For sales or other lower-funnel goals, start with recent converters or high-value customers; for awareness or lead generation, use a seed closer to that goal. A randomized field study found that lookalikes based on later customer-journey stages performed better for downstream outcomes, while moving the seed further down the funnel did not improve upstream outcomes such as clicks. (cowles.yale.edu)

Prioritize clean, deduplicated records that reflect the customers you want more of, such as repeat purchasers, profitable accounts, or qualified leads. Website visitors, app users, and engaged viewers can be useful alternatives when you lack enough converters, but they are weaker evidence of buying intent. Google’s current Demand Gen guidance recommends high-intent seeds, including recent converters; as of November 19, 2026, Google is transitioning these lookalikes to audience suggestions rather than strict reach limits. (support.google.com)

How should you test lookalike audiences against other targeting strategies?

Run a controlled, concurrent test of a lookalike against broad targeting and any relevant interest-based or other prospecting option. Keep the offer, creative, objective, placements, budget, and measurement window consistent, and prevent audiences from overlapping where possible. Use a platform split test or randomized holdout when available; comparing separate campaigns from different periods can confound audience performance with seasonality, creative changes, or auction conditions.

Judge results by incremental purchases or qualified leads, new-customer cost, and profit or customer value, not just clicks or platform-reported ROAS. If the lookalike improves reach but worsens acquisition cost or customer quality, broader targeting may be the better choice. On Google Demand Gen, remember that a lookalike may function as a signal rather than a hard audience boundary, so compare the resulting campaign performance with that limitation in mind. (support.google.com)

Related Reading

Start collecting feedback in 5 minutes.Try the no-code surveys your customers actually answer — free, no credit card.
Get started free

Latest in Attribution & Marketing Analytics

Start collecting feedback in 5 minutes.

Try our no-code surveys that visitors actually answer.

Questions or Feedback?

We are always ready to hear from you.