Blended ROAS Is Hiding the Channel That Is Dying

Blended ROAS hides channel-level decay, because a strong channel can carry the average while a weak one bleeds. Split the number by channel, by new versus returning, and by device, and the dying channel shows up immediately.

The blended number that never moves

Blended ROAS is the number most teams report, and it is the number that tells them the least. It averages every channel, every campaign, and every customer into one ratio, and the average is stable precisely because it hides the movement underneath. A channel that is dying slowly is offset by a channel that is growing slowly, and the blended number sits still while the business underneath it changes shape. The stability feels like control. It is the opposite: it is the absence of information, and the absence is comfortable enough that nobody splits the number until something breaks loudly enough to force it.

The problem is arithmetic, not dishonesty. A channel spending 60 percent of the budget at a 4 to 1 return and a channel spending 40 percent at a 1.5 to 1 return blend to roughly 3 to 1, and the blended number looks healthy. If the weak channel falls to 1 to 1 while the strong channel rises to 5 to 1, the blend barely moves, and the team concludes nothing changed. Something changed: a channel stopped paying for itself, and the money it burns is money the strong channel earned. The blended number cannot see that, because it was designed to average it away.

The channel that pays for the average

The channel that pays for the average is usually the one closest to the purchase: branded search, retargeting, or the email flow that closes people who were already going to buy. Those channels look efficient because they harvest demand other channels created, and the blended number credits them with the sale while charging the prospecting channel with the cost. The prospecting channel looks weak by comparison, because it does the expensive work of finding new customers, and the blended number punishes it for that work while rewarding the channel that closed the sale.

The fix is not to kill the prospecting channel. It is to stop judging it by the same blended number that rewards the closer. Split the reporting by intent: prospecting and closing are different jobs with different returns, and a prospecting channel at 1.5 to 1 can be the best investment in the business if it feeds the closing channel that runs at 5 to 1. The blended number treats them as competitors. They are a sequence, and the sequence only works if both parts are funded.

The split that shows the truth

Split by channel first, then by new versus returning, then by device. New versus returning is the split that most often reveals the problem, because a blended number that mixes them hides the fact that the business is buying the same customers over and over while new customer acquisition quietly stalls. Device is the second split, because mobile and desktop convert at different rates and a shift in traffic mix looks like a performance change when it is really a mix change. Three splits, run monthly, and the dying channel appears in the first month it starts dying, instead of the quarter it becomes undeniable.

The split also needs a time dimension. A channel that decays slowly looks fine in any single month, because the month-over-month change is small. Plot it over six months, and the slope is obvious. The teams that catch decay early are the teams that look at the slope, not the level, and the slope is only visible when the number is split. A blended number has no slope worth reading, because the channels cancel each other out.

The decision the split forces

The split forces a decision the blended number lets you avoid: which channel gets more money, which gets less, and which gets cut. That decision is uncomfortable, because it means admitting that a channel the team liked is no longer earning its budget, and it means moving money to a channel that is less familiar. But the decision is the job. A blended number that never moves is a report that never asks for a decision, and a report that never asks for a decision is a report that exists to be filed, not to be used. Split the number, read the slope, and move the money. The channel that is dying will not recover because the average looks fine.

Frequently asked questions

What is blended ROAS, and why is it misleading?

Blended ROAS averages every channel into one return ratio. It hides channel-level decay, because a strong channel can carry the average while a weak one bleeds, and it mixes prospecting with closing, which are different jobs.

Which split reveals a dying channel first?

Channel first, then new versus returning, then device. New versus returning often reveals the real problem: the business buying the same customers repeatedly while new acquisition stalls.

Should I judge prospecting by the same ROAS as retargeting?

No. Prospecting finds new customers and retargeting closes people who were already going to buy. They are a sequence, not competitors, and judging prospecting by the closer return starves the top of the funnel.

How often should I split the number?

Monthly, with a six-month slope. A slow decay is invisible month to month and obvious over six months, and the slope only shows when the number is split by channel.