ROAS vs NC-ROAS: which bar should grade your creative?

Two media buyers can look at the same ad account and disagree about whether it's healthy, and both can be right, because they're using different bars. One is looking at blended ROAS: all revenue over all spend. The other is looking at NC-ROAS, new-customer ROAS: revenue from first-time buyers only, over the same spend. Same account, same month, two verdicts. Which bar you grade against isn't a reporting detail. It decides which creatives get called winners, which get killed, and what your next round of concepts looks like.

What each metric measures, and what each one hides

Blended ROAS is the simple one: total attributed revenue divided by total ad spend. It counts everything the ads touched, including a loyal customer who saw a retargeting ad on the way to a purchase she was probably going to make anyway. That's its weakness dressed up as a strength. Returning customers already know the brand, convert more easily, and cost less to reach, so any creative that leans on them looks efficient. Blended ROAS rewards harvesting demand you already own.

NC-ROAS keeps only the revenue from first-time buyers and divides by spend. It answers a narrower and more expensive question: are these ads bringing in people who've never bought from us? That's the question growth actually depends on, because a brand that only re-sells to its existing list is shrinking in slow motion. The cost of that narrowness is that NC-ROAS ignores real money. A creative that reactivates lapsed customers at great margin scores a flat zero on a strict new-customer bar, and that's a real distortion too.

So neither metric is the honest one and the other a lie. They're two different windows on the same account, and each has a blind spot exactly where the other one sees clearly.

A worked example (the numbers are made up)

Here's a deliberately simple, invented scenario, not industry data, just arithmetic you can redo on a napkin. Say a skincare brand spends $10,000 on paid social in a month and the ads get credit for $40,000 in revenue. Blended ROAS: $40,000 over $10,000, a 4.0x. Most dashboards call that a good month and move on.

Now split the revenue. Of that $40,000, suppose $15,000 came from first-time buyers and $25,000 from people who'd bought before. NC-ROAS: $15,000 over $10,000, a 1.5x. Same spend, same ads, and the number just fell from "comfortably profitable" to "roughly break-even on first purchase, depending on margin." Neither number is wrong. One says the machine is efficient. The other says the machine is barely acquiring.

Push it one level down, to the creative level, and the story gets sharper. Suppose one video in that account drove $12,000 of revenue on $2,000 of spend, a 6.0x, but $11,000 of it came from repeat buyers because the ad mostly reached warm audiences. Its NC-ROAS is 0.5x. Meanwhile a scrappier ad did $6,000 on $2,500, a modest 2.4x blended, but $5,000 of that was first-time buyers: a 2.0x NC-ROAS. Grade by blended and you brief more of the first ad. Grade by NC-ROAS and the second ad is your best acquisition creative, and your next concepts should look like it. Same data. Opposite creative direction.

Why blended ROAS flatters, structurally

The flattery isn't a bug in your dashboard, it's built into how paid social delivery works. Platforms optimize toward whoever is most likely to convert, and the people most likely to convert are the ones who already trust the brand. Left alone, spend drifts warm. Retargeting and brand-search dollars pick up conversions that were already in motion, and blended ROAS books them all as ad-driven revenue. The stronger a brand's retention, the more this compounds: a great repeat-purchase business can post a beautiful blended number during a quarter when it's acquiring almost nobody. The chart looks like growth. The customer file says otherwise.

Blended ROAS grades the harvest. NC-ROAS grades the planting. Confuse them for long enough and you'll have a very efficient account attached to a shrinking brand.

When each bar is the right one

The right bar depends on the economics of the brand, not on which metric sounds more sophisticated. Two brands can make opposite choices and both be correct.

• LTV-rich brands (subscriptions, consumables, strong repeat rates): NC-ROAS, or new-customer CPA, is usually the honest bar. These brands can afford a first order that only breaks even, or even loses a little, because the customer keeps paying for months. Grading their creative on blended ROAS punishes exactly the ads doing the hard, valuable work of acquisition.
• Cashflow-constrained brands (thin margins, no outside capital, long inventory cycles): blended ROAS or first-order profitability is often the right bar, because a payback that arrives in month six doesn't make payroll in month one. A strict NC-ROAS target that tolerates unprofitable first orders can be technically smart and practically fatal.
• One-and-done purchase brands (mattresses, some high-ticket goods): the gap between the two metrics is small, since almost every buyer is a first-time buyer. Blended ROAS is fine, and NC-ROAS adds ceremony without much new information.
• Agencies running both kinds of client: the bar has to be set per brand. Applying one house metric across a portfolio guarantees you're grading at least some of those brands against the wrong economics.

Notice what's constant across all four cases: the metric follows the money model. The question is never "which metric is better," it's "what does a new customer eventually pay this brand, and how long can this brand wait to collect it?" Answer that, and the bar picks itself.

The caveat both metrics share: neither proves incrementality

It's tempting to treat NC-ROAS as the pure metric and blended as the compromised one. Don't. Both are attribution metrics, which means both count revenue the ads touched, not revenue the ads caused. A first-time buyer who clicked your ad might have found the brand through a friend last week and would've bought regardless; NC-ROAS still books her as ad-driven acquisition. NC-ROAS fixes the who question, is this buyer new, but not the why question, did the ad make it happen. The honest tools for the why question are holdout and geo tests, where you withhold ads from a group and measure the gap. Most teams can't run those continuously, which is fine, as long as you hold both ROAS flavors with appropriate looseness: they're directional reads on relative creative performance, not proofs of causation. An ad that wins on your chosen bar has won against that bar. That's a real, useful fact. It's just not the same fact as "this ad created this revenue from nothing."

Pick the bar per brand, then grade everything against it

Here's the practical failure mode we see most: the bar gets chosen by the tool instead of by the brand. A dashboard defaults to blended ROAS, so every client gets graded on blended ROAS, including the subscription brand whose whole model depends on cheap acquisition and patient payback. The winners' list comes out warm-audience-flavored, the next creative round chases those "winners," and the account slowly optimizes toward its own customer list.

This is why AgentMark doesn't have an opinion about which metric is correct. It has an opinion about who gets to decide: the brand does. You set the primary KPI per brand, blended ROAS, NC-ROAS, cost per new customer, CPA, whatever matches the economics, and every creative is graded Winning, Losing, or Fatiguing against that bar. Flip the KPI and the whole account re-scores: the pattern grid rebuilds, the winners' list changes, and the concepts you generate next are grounded in what wins against the bar you actually chose. The 6.0x retargeting darling from the example above stops being a winner the moment the brand says acquisition is the job. Nothing about the data changed. The question did, and the grading followed it. That's the discipline worth keeping whether you use our software or a spreadsheet: decide the bar first, per brand, out loud. Every creative verdict downstream inherits that choice.