Allocation pattern · E-commerce / DTC

Beauty Subscription: ROAS Drifted From 3.5x to 1.6x at Scale

A $8M GMV post-seed beauty subscription brand came in last quarter. AOV $42 with a 62 percent gross margin. Predominantly Meta with some TikTok experimentation. Six months earlier the program was returning 3.5x topline ROAS. By the time the founder reached out, ROAS was at 1.6x and the team was exhausted.

The founder’s instinct was that the audience had been exhausted. The agency was pushing for more creative volume and more spend. Both were wrong.

The diagnosis (3 sentences)

Three problem-aware creative angles had been in heavy rotation for five months: “dry skin,” “uneven tone,” “sensitivity.” Frequency on the top creative was 8.4x per user over 30 days. The audience inside the problem-aware emotional driver had been substantially reached; the angle library was saturated, not the audience.

The allocation move

  • Run the creative-test diagnostic first. New angles against the same audience targeting at 12 percent of channel budget for 21 days. If new angles outperform the existing rotation, creative saturation is confirmed. If they perform similarly, the diagnosis shifts to audience saturation, which is a different fix.
  • In this case the new angles outperformed by 1.6x on first-purchase ROAS. Confirmed creative saturation.
  • Build 8 new angles against a buyer-state matrix: 4 solution-aware angles (“looking for clean beauty that actually works”), 2 brand-aware angles (social proof, founder story), 2 identity-driver angles (the kind of person who uses this brand). Rotate against the existing library at 50 percent of new-creative budget.
  • Reallocate 15 percent of spend from Meta to TikTok over 60 days, where one of the new identity-driver angles tested significantly stronger. The platform change is secondary; the angle change is primary.
  • Switch the operating metric from topline ROAS to contribution-margin ROAS. Topline 3.5x was actually contribution 2.1x. The “drift from 3.5x to 1.6x” was really contribution drifting from 2.1x to 0.9x, which means the brand had been losing money on paid for the last 6 weeks, not just earning less.

When this applies. When it does not.

Applies: DTC brands at $5M to $15M GMV running predominantly on Meta. 3 to 5 creative angles in rotation for 4 to 6 months. Topline ROAS has drifted 30 percent or more over a quarter. Creative frequency above 6x per user over 30 days.

Does not apply: Sub-$1M GMV brands where the addressable audience is still under-reached. $20M+ GMV brands where audience saturation is more likely than creative saturation (run the test before assuming). Step-function ROAS drops, which usually signal tracking break or platform policy change rather than creative fatigue.

A founder line, anonymized

“I wanted to fire the agency. I wanted to push harder on TikTok. The audit told me both were the wrong fix. The angle library was 3 stories repeated for 5 months. We built 8 new angles. ROAS recovered from 1.6 to 3.4 at the same total spend in 12 weeks.”

For the three-cause flowchart that separates creative saturation from audience saturation from channel concentration, see ROAS Dropped From 3.5x to 1.8x: The DTC Diagnostic Flowchart. The diagnostic test is what tells you which fix to apply.

Book the 10-Day Allocation Audit.