Blended ROAS hid what actually broke in 2025

The account review starts the same way every quarter. Someone shares a screen. Blended ROAS is down a few points against last year. Spend is up. Revenue is up. Nothing looks obviously broken. So the meeting drifts to creative. Within ten minutes someone proposes cutting the three ads with the worst reported ROAS. Everyone nods, because the number on the screen agrees.

Those three ads are usually the ones bringing in people who have never bought from you. That is the trap, and it got worse in 2025. The argument here is narrow. Ecommerce did not degrade last year. Acquiring a first time customer degraded. Blended reporting hides that almost perfectly, because your repeat buyers sit inside the same number as your new ones.

What the vertical medians actually show

Northbeam published its Industry Power Rankings on 11 February 2026. The report gives vertical medians for two things side by side. Blended MER is total revenue divided by marketing spend. First time MER counts only revenue from new customers against that same spend. In Northbeam's numbers, first time MER fell in every vertical listed. Blended MER barely moved. That gap is the whole story of last year.

Take Sporting Goods and Fitness. Northbeam reports spend up 25 percent and revenue up 17 percent. Blended MER fell 7 percent, which is a bad quarter but a survivable one. First time MER fell 11 percent. First time CAC, the cost to buy one new customer, rose 17 percent. Health and Wellness is starker. Blended MER slipped 1.5 percent while first time MER fell 7 percent. The blended line understated the damage by more than four times.

Blended ROAS is an average of a problem and a cushion. The cushion is repeat buyers.

The mechanism is dull, which is why it goes unnoticed. Returning customers cost almost nothing to reach again. They come back through email, through SMS, through a bookmark. A growing repeat base holds the blended number steady while new customer cost climbs underneath it. A 1.5 percent move reads as rounding. Nobody calls a meeting over rounding. Meanwhile the part of the business that feeds every future cohort is getting more expensive every month.

The objection to my own case

The Northbeam data is thin, and I would rather say so than pretend. Its Technology vertical shows a 38 percent average spend change against a 1.3 percent median. That is a distribution pulled hard by a few large accounts. The report discloses no sample size, no attribution window and no confidence intervals. It is a customer panel, not a census of the market. Do not treat those medians as your benchmark.

The argument does not need them to be precise. It needs two lines from one dataset, measured the same way, on the same accounts, over the same period. Whatever bias the panel carries, it carries into both lines equally. One line held and the other dropped. The divergence survives the sampling problem even when the levels do not.

Independent work points the same direction. Measured analysed over 10,000 campaigns across more than 200 advertisers in 2025 using geo tests. Geo tests hold out whole regions and measure the lift against them. Median incremental Meta ROAS came out at 2.16 dollars. And 64 percent of incremental conversions came from new or reactivated customers, Measured found. So the lift paid media genuinely creates sits mostly in the group whose economics got worse. That is an awkward pairing if you report blended.

The category itself was not shrinking. The IAB and PwC Internet Advertising Revenue Report for Full Year 2025 came out on 16 April 2026. It puts US internet ad revenue near 300 billion dollars, up 13.9 percent. It is compiled from reported company revenue rather than from a survey of marketers. That is why it survives a CFO conversation. More money went into the system. It bought fewer first time customers.

What this costs inside the account

The practical damage is misallocation, and it compounds quietly. A blended report rewards ads that harvest demand you already created. Retargeting, brand search, offers sent to your own list. Those ads report beautifully because the buyer was already on the way. Prospecting creative looks weak next to them on every dashboard. So it gets cut first, in the exact year when new customer cost was rising fastest. Then the blended number holds for a quarter or two, because the repeat base is still paying out. Most 8 and 9 figure accounts I look at are living inside that lag right now. The cliff arrives later, when the cohorts you stopped buying fail to show up.

Do this on Monday

You do not need a new tool or a new vendor for any of this. You need one split, and an hour.

• Split the last 12 months of revenue into first time and returning. Shopify tags this natively and most platforms will too.
• Divide paid spend by first time orders, month by month. That is your first time CAC.
• Chart it against the same months last year. Your own trend matters more than any vertical median.
• Rank your creative by first time orders instead of blended ROAS, and see which ads change places.

If first time CAC is climbing while blended sits still, you have your answer before you touch a single campaign. The ads worth protecting are often the ones that look mediocre in a blended view. Run the split first. Open the account second.