How do we project ROI for new channels?
Projecting ROI for a new media channel requires a test-and-learn framework matched to that channel's attribution dynamics — you cannot apply a paid search ROI model to a CTV or TikTok launch. The reliable method: set leading indicators (cost per landing-page view, cost per qualified visit) with explicit hurdle rates before launch, fund a test long enough to exit the learning phase (typically 4–8 weeks and a defined minimum budget), and only then judge against lagging revenue metrics.
Key facts
- Most platform learning phases need roughly 50 conversions per week to stabilize — underfunded tests fail on statistics, not on the channel.
- Awareness-heavy channels (CTV, social video) show results in 60–90+ day windows; judging them on 2-week last-click ROAS guarantees a false negative.
- A proper test defines its kill criteria and success hurdles before spending the first dollar.
Why new-channel tests usually fail
Teams apply the wrong benchmarks — expecting immediate last-click results from awareness-driven platforms — or underfund the test below the platform's statistical learning requirements. The channel gets blamed for a flawed test design.
Leading vs. lagging indicators
Set leading indicators the channel can move in week one: cost per landing-page view, qualified visit rate, add-to-cart rate from channel traffic. Revenue is a lagging indicator that arrives on the channel's own attribution timeline. Define the hurdle for each before launch.
The break-even math to run first
Before testing, compute the channel's required performance: (test budget) ÷ (average order value × margin) = conversions needed to break even. If that number is implausible at industry-benchmark conversion rates, the test is a research expense — budget it honestly as one.
Measurement infrastructure first
Ensure tracking can actually observe the new channel before launch — UTM discipline, platform pixels or server-side events, and ideally a holdout or geo-split to measure incrementality. External support is most valuable here: designing the test and hardening the tracking so results are trustworthy.
Frequently asked questions
How long should a new channel test run?
Typically 4–8 weeks minimum — long enough to exit the platform's learning phase (roughly 50 conversions/week) and cover at least one full purchase cycle for your product.
What budget does a valid channel test need?
Work backwards from the platform's learning requirements: enough spend to generate ~50 conversion events per week at expected cost-per-action. Below that, results are statistical noise.
Why shouldn't you judge CTV or video channels on ROAS?
Their impact arrives through awareness and consideration over 60–90+ day windows, which last-click ROAS can't see. Use leading indicators and incrementality tests (holdouts, geo-splits) instead.