Case Study · TrimRX · Incrementality

How We Proved CTV Drove Real, Incremental Revenue for TrimRX

TrimRX, a fast-growing GLP-1 telehealth brand, didn’t trust platform-reported ROAS to judge CTV. We built a 51-market synthetic difference-in-differences test to measure what CTV actually added to their bottom line.

Results at a glance

4.94%

Average weekly revenue lift from CTV

~9.4%

Peak weekly revenue lift (week of May 4)

3.42x

Incremental ROAS

p < 0.03

Statistical significance, with 6 of 7 weeks at p < 0.01

The Problem

Attribution takes credit. It doesn’t prove it.

TrimRX didn’t trust platform-reported ROAS or last-click attribution to judge CTV, and for good reason. Platforms overcount by crediting anyone who was merely exposed to an ad. In GLP-1 telehealth, where a patient might see a CTV spot, click a Meta ad days later, and then finish an intake form, that leaves the true impact of CTV unknown. TrimRX needed a number that could hold up to scrutiny and guide real budget decisions.

How it Works

How SDiD finds the lift attribution can’t

SDiD answers one question: what would have happened without CTV?

01

Turn it on in test markets

We turned CTV on in 16 test states and kept it off in the other 34 states and DC.

02

Build a synthetic twin

The model blends the untouched markets into a “synthetic” twin that tracks the test markets’ sales almost exactly before launch.

03

Measure the gap

After launch, the gap between the real test markets and their synthetic twin is the lift the channel caused.

Approach

Weakness

How SDiD handles it

Platform attribution

Credits touches, not causes

Measures first-party revenue and orders, not attributed conversions

Before/after comparison

Mixes in seasonality, promos, and site changes

Compares against markets living through the same calendar

Simple holdout (one test vs. one control market)

No two markets behave alike

Weights many control markets into a close-fitting twin

Our Method

How we run an SDiD test

Every test follows the same six steps, and the client signs off on the design before a dollar is spent.

01

Pull first-party data

We analyzed 19 weeks of weekly revenue and conversions by state from TrimRX’s own backend. Platform-reported conversions were never the outcome we measured.

02

Account for every other channel

TrimRX’s Meta and Google spend kept running at near-identical levels in test and control states, and went into the model as covariates. That left CTV as the only variable.

03

Pick the markets

An algorithm split all 51 markets into 16 test states and 35 control markets to maximize pre-test balance. The two groups were within 2% on weekly revenue and 1% on weekly conversions before launch.

04

Pre-test the model

All 8 statistical quality checks passed. Revenue and conversion trends correlated at 0.997 and 0.999, parallel-trend tests passed (p = 0.96 and p = 0.97), and week-to-week error against the benchmark was just 2.79% on revenue and 2.74% on conversions. A placebo check against 500 random market splits put our design in the bottom 8% for imbalance on revenue and the bottom 2% on conversions.

05

Run the full flight

CTV launched in the 16 test states on April 13, 2026, and ran for 7 weeks through June 1. Control markets ran no CTV.

06

Read it and act on it

We measured the lift every week against the synthetic twin, with confidence intervals and p-values, then rolled it up into incremental revenue, ROAS, conversions, and CPA.

Keep outcomes simple

Keep outcomes simple

Revenue and orders are modeled directly. Calculated metrics like AOV add noise, so we infer them from the core models rather than test them on their own.

Client Spotlight

TrimRX

TrimRX connects patients with licensed physicians for personalized GLP-1 weight-loss treatment, delivered to their door. With Meta and Google already running at scale, crediting CTV off platform attribution would have been easy to overstate, or to miss entirely.

The result

+4.94%

Average weekly revenue lift, positive in every one of the 7 weeks and peaking at about 9.4% the week of May 4. CTV delivered a 3.42x incremental ROAS, measured without platform attribution.

7 of 7

Weeks with positive revenue lift

3.42x

Incremental ROAS

Test design

We launched CTV through MNTN in 16 test states at a steady weekly budget and held it back in 35 control markets. Meta and Google spend continued at matched levels across both groups and went into the model as covariates. The outcome was first-party revenue, not view-through conversions.

What the Results Changed

CTV went from a question mark to a performance channel

Lift that kept growing

Cumulative incremental revenue rose every single week through week 7 with no sign of diminishing returns. That’s a strong signal CTV can scale.

Results you can trust

The 95% confidence interval stayed above zero the entire test. The lift was real, not noise.

A benchmark to defend

With true lift isolated, TrimRX can defend CTV spend, set a real efficiency target, and weigh it against every other channel on equal footing.

If your measurement lives inside the ad platform, the platform is grading its own homework. Attribution credits views. SDiD measures outcomes. The only way to know if CTV is working is to run markets where it isn’t.

That is the real value of SDiD. It tells a brand both whether a channel works and how far to push it.

That is the real value of SDiD. It tells a brand both whether a channel works and how far to push it.

Where SDiD Fits

Built for telehealth brands with real volume across many states

Telehealth brands are a natural fit for SDiD. They sell in many states at once, they own clean first-party order data, and they usually run several channels that all claim the same patient. If you’re licensed in enough states to hold some back, you can measure what a channel actually adds.

Question a brand is asking

How SDiD answers it

Is CTV driving sales, or just taking credit?

Measures CTV’s lift against first-party revenue

Which CTV partner performs better?

Runs two platforms in paired markets, head to head

Does CTV move in-store retail sales?

Tests regional markets against store-level sales data

Should we scale, and by how much?

Pairs the lift read with a diminishing-returns curve

Want to know what your CTV is really driving?

We’ll tell you up front if your brand has the volume and market spread for a clean test.