Return on ad spend is the most quoted number in performance marketing and one of the least useful. It measures what a platform can see, which is rarely the same thing as what the business earned.
ROAS is not wrong. It is just a single, platform-scoped ratio that hides the variables that decide whether a program is healthy: margin, retention, incrementality, and payback period.
Four numbers that beat ROAS
1. Contribution margin per acquired customer
Revenue is not profit. A channel with a 4x ROAS on a 20% margin product is worse than a channel with a 2.5x ROAS on a 60% margin product. Once you feed margin into your platform reporting, bidding behaviour changes immediately—usually for the better.
2. Blended payback period
How many months of gross profit does it take to repay acquisition cost? A blended payback under twelve months gives you room to invest. A payback over twenty-four months means every new dollar is a bet on retention you may not have evidence for yet.
3. Incremental conversion rate
The share of conversions that would not have happened without the ad. This is the number ROAS cannot see. Branded search and retargeting look spectacular in platform reporting and are frequently the least incremental lines in the account.
4. Retention-adjusted lifetime value
Cohort LTV at 90 and 180 days, segmented by acquisition channel. Channels that bring in customers who stay deserve a different budget than channels that bring in customers who churn. Without this, you optimize for the wrong acquisition.
The uncomfortable test
Turn a channel off for two weeks in a controlled geography. Measure the change in total revenue, not channel revenue. Most teams discover at least one line item that was buying demand they already had.
Build a measurement stack in three layers
Measurement arguments are usually structural. Teams try to answer a strategic question with a tactical tool. Three layers solve it.
| Layer | Question it answers | Typical tools |
|---|---|---|
| Platform | Is execution efficient today? | Ad platforms, bid data, creative CPA |
| Attribution | Which touchpoints assist? | MMM, incrementality tests, GA4 |
| Business | Is growth profitable and durable? | Margin, cohort LTV, payback, pipeline |
Each layer has an owner and a cadence. Platform reporting is daily. Attribution reviews are monthly. Business reviews are quarterly, and they are the only ones that should change strategy.
12 mo
Target blended payback period for programs that want room to scale into new channels.
2 wk
Minimum window for a meaningful holdout or geo incrementality test.
5
Maximum number of headline metrics. More than five and nobody owns any of them.
Guardrails beat targets
A target tells a team what to chase. A guardrail tells them what not to break while chasing it. Mature performance programs run on both:
- Target: grow qualified pipeline 30% year over year.
- Guardrail: blended CAC does not exceed $X, and 180-day retention does not fall below Y%.
Guardrails are what prevent a team from hitting a number in a way that damages the business.
Report the trend, not the day
Daily reporting creates daily panic. Weekly reporting creates weekly decisions. If a metric definition changes more often than the metric itself, you do not have measurement—you have noise with a dashboard.
Pick the definitions, freeze them for a quarter, and let the trend tell you something.
