The Ecomm Analyst

New customer ROAS is attributed revenue from first-time buyers divided by ad spend. It answers a narrower question than blended ROAS, which is whether your advertising is bringing in people you did not already have. For most brands running paid acquisition it is the more honest number, and yes, you should be looking at it. Whether you should optimize toward it depends on how much of your revenue comes from repeat purchase, and that answer is different for a coffee subscription than for a mattress company.

What blended ROAS hides

A campaign posting 5.0 blended might be running 5.0 against new customers, in which case acquisition is genuinely working. Or it might be running 2.0 against new customers and 12.0 against people already on your email list who were reached by a retargeting ad on their way to buy something they had already decided to buy. Both average out the same. Only one of them is growth.

The second case is more common than most operators expect, and it gets worse as brands mature. The bigger your existing customer base, the more of your paid traffic consists of people who already know you. Retargeting and broad campaigns both sweep them up. Your blended ROAS drifts upward year over year while new customer acquisition quietly stalls, and the dashboard reports this as improving performance.

Reading the two together

The gap between blended and new customer ROAS is the diagnostic. A small gap means most of what you are buying is genuinely new, which is typical of a young brand or a well-separated prospecting campaign. A large gap means a lot of your attributed revenue is coming from existing customers, and you should be asking how much of that you were going to get without paying for it.

I watch that gap over time rather than at a point. When it widens month over month with spend flat, something has shifted in delivery toward warmer audiences. That is usually the platform doing exactly what it was told, since a bid strategy optimizing for conversions will reliably find the people most likely to convert, and the people most likely to convert are the ones who already bought from you.

ThoughtMetric, which sponsors this blog, exposes new customer ROAS and returning customer ROAS as separate metrics alongside blended ROAS, which makes the comparison a reporting question rather than a data export and a spreadsheet. It also breaks out cost per new customer order specifically, which is the version of CAC I find most useful for campaign-level decisions. Pricing is pageview-based starting at $99 per month for 50,000 pageviews with every feature at every tier, and there is a two-week free trial. Any tool that segments first-time from repeat purchase will get you there, but a surprising number of dashboards still do not.

When optimizing toward it is right

If repeat purchase drives a large share of your revenue, first-order economics are the wrong constraint. A subscription brand or a consumable can rationally accept a new customer ROAS well below one, because the second and third orders carry the return. In that world you optimize for new customers acquired at a cost you can defend against lifetime value, and new customer ROAS is close to the primary metric.

If you sell something people buy once every few years, this reverses. Repeat revenue is not coming to rescue a bad first order, so new customer ROAS needs to clear break-even more or less on its own, and there is less distance between it and the blended figure anyway.

The mistake is applying the subscription logic to a durable goods brand. I have watched a team accept a new customer ROAS of 0.8 on the argument that lifetime value would cover it, when the median customer’s lifetime value was one order plus a slow trickle of accessory purchases. The repeat rate has to be measured, not assumed, and it should be measured before it becomes the justification for a spend level.

The failure mode

Optimizing hard toward new customer ROAS pushes budget out of retargeting and toward cold prospecting, which is broadly the right direction. Pushed far enough it starves campaigns that were doing real work. Retargeting is not pure waste. Some share of it genuinely rescues abandoned carts that would not have returned.

The way to find out how much is a holdout rather than an attribution model. Turn retargeting off in one region for two or three weeks and watch total orders, not attributed ones. Whatever revenue survives the blackout was never incremental. That test costs almost nothing and settles an argument that attribution data alone cannot.

My practical setup is blended ROAS for the business-level view, new customer ROAS for judging acquisition, and cost per new customer order for campaign decisions. Three numbers, and the gap between the first two tells you most of what is going on. On why channel-level CAC stopped being part of that set, I wrote about it in why I stopped looking at channel-level CAC.