When Glow Theory, a D2C skincare brand based in Austin, came to us in mid-2025, they had a familiar problem: rising acquisition costs, plateau-ing revenue, and a paid media strategy that felt like it was running on fumes. Eight months later, they'd tripled revenue. Here's how.
The Problem
Glow Theory was spending $120K/month on Meta and Google ads with a blended ROAS of 1.8x. Not terrible, but not enough to fund the growth they needed. Their targeting was based on traditional interest-based audiences and lookalikes built from their full customer list — the same playbook every D2C brand runs.
The issue wasn't the creative or the offer. It was the audience quality. They were casting a wide net and hoping for the best.
The AI Approach
We built a predictive audience model using their first-party data — purchase history, browse behavior, email engagement, and customer support interactions. The model identified patterns that correlated with high lifetime value, not just first purchase.
The shift from "who's likely to buy" to "who's likely to become a loyal customer" changed everything. Acquisition cost went up 20%, but lifetime value went up 340%.
The Execution
We fed the predictive model's output into Meta's ad platform as seed audiences. Instead of lookalikes based on "all purchasers," we created lookalikes based on "predicted high-LTV customers." The targeting was sharper. The algorithm had better signal to work with.
Simultaneously, we used the model to build suppression lists — people predicted to be one-time buyers or heavy returners. This alone reduced wasted ad spend by 22%.
The Results
Over eight months: revenue grew from $380K/month to $1.14M/month. Blended ROAS improved from 1.8x to 3.4x. Customer retention at 90 days improved by 45%. The model paid for itself in the first three weeks.
