Blended ROAS is the most comfortable number in paid search. It looks healthy, it trends in the right direction, and it quietly rewards spending on customers you would have won anyway.
It is also the single most common reason we see accounts that look brilliant in the platform and mediocre in the P&L. The dashboard says four pounds back for every pound in. The finance team asks why revenue has been flat for three quarters. Both are telling the truth. They are just measuring different things.
TL;DR: Blended ROAS averages two different businesses into one flattering number: cheap revenue from people who already know you, and expensive revenue from people who don't. Smart Bidding chases the cheap kind, the average holds, and acquisition quietly shrinks. The escape is splitting the number: new-customer ROAS or CAC for decisions, blended for context, margin fed into both.
How the trap forms
Blended ROAS mixes two fundamentally different kinds of revenue into one average.
Returning customers already know you. They search your brand name, click the first ad they see, and convert at a rate a cold prospect never will. Every pound spent on them looks spectacular, because most of that revenue would have arrived anyway through an organic listing, an email, or a typed-in URL.
New customers are the opposite. They are expensive, sceptical and slow to convert. They are also the only thing that actually grows the business.
Mix the two together and the returning customers flatter the average. Smart Bidding notices exactly what you have told it to notice: the cheapest conversions live close to home. So the algorithm leans further into brand terms, remarketing lists and past purchasers, the blended number holds or even improves, and the share of budget doing real acquisition work shrinks month after month. Nobody made a bad decision. The target did it for you.
The uncomfortable part is that this is the system working as designed. Automated bidding is exceptionally good at hitting the number it is given. Hand it a blended target and it will find the shortest route there, and the shortest route always runs through people who were already going to buy. The algorithm is not wrong. The instruction is.
Two accounts, same headline number
Here is the trap made visible. Both of these accounts report a blended ROAS of 4.0. The figures are illustrative, but the pattern is one we see constantly:
| Account A | Account B | |
|---|---|---|
| Blended ROAS | 4.0 | 4.0 |
| Share of spend on brand and remarketing | 70% | 40% |
| Returning-customer ROAS | 7.5 | 5.5 |
| New-customer ROAS | 1.5 | 3.0 |
| Customer file | Shrinking | Growing |
On the dashboard they are identical. In reality, Account A is liquidating its warm audience to fund a flattering average, while Account B is buying genuine growth at a sustainable rate. A blended target cannot tell them apart. That is the whole problem.
The tell-tale symptoms
You do not need a forensic audit to spot the trap. It shows up in patterns:
- ROAS is stable or improving while total revenue growth is flat.
- Brand and remarketing quietly consume a growing share of spend.
- Customer file growth has slowed, even though the platforms report record efficiency.
- Platform-reported ROAS climbs while your blended marketing efficiency ratio, total revenue over total spend, stays put. The gap between those two lines is usually double-counted warm revenue.
- Every attempt to scale spend "breaks" the target, so budgets get pulled back.
That last one matters most. When your target is blended, scaling always looks like it is failing, because incremental spend can only go to colder, more expensive audiences. The average falls, the alarm goes off, and the account retreats to the warm core again. The target becomes a ceiling on growth.
The numbers that replace it
Escaping the trap is not about abandoning ROAS. As we argued in Dear ROAS, we need to talk, it remains a useful diagnostic. The change is splitting it into numbers that mean something:
- New customer ROAS or CAC. What you actually pay to acquire someone who has never bought from you, measured against what a new customer is worth. This is the growth number.
- Returning and brand efficiency. Measured separately, with honest questions about incrementality. Some of it is worth paying for. All of it should be challenged.
- Margin-adjusted value. Revenue is not profit. A blended ROAS of 4 on a 20% margin product is a very different business from the same number at 60%. Feed margin into the measurement or the platforms will optimise for turnover, not profit. Getting this data flowing cleanly is measurement work, not media work, and it is usually the missing foundation.
Prove incrementality, don't argue about it
"Would we have got that sale anyway?" is the question that stalls every brand-spend conversation, and it does not need to be a debate. It is testable.
The cleanest version is a holdout: pause brand ads in a set of comparable regions for a few weeks, keep them running everywhere else, and measure what actually happens to total sales in the paused regions, not just paid sales. Organic and direct capture some of the demand. The share they do not capture is your genuine incrementality, and it becomes the honest price of brand protection. Sometimes the test says brand spend is defending real revenue against competitors camped on your name. Sometimes it says you are paying a toll on traffic you already owned. Either answer is worth more than another quarter of assumption.
How to escape, step by step
- Separate new from returning at the tracking level. Pass a new-versus-returning flag into your conversion data, from your order system or CRM rather than a cookie that forgets everyone after ninety days. Without this, nothing else on the list is possible.
- Set targets against new customer acquisition and margin. Agree with finance what a new customer is worth, then set the acquisition target from that, not from last quarter's blended average.
- Restructure so budgets can be steered. Brand, remarketing and prospecting need enough separation that you can control how much goes to each job, rather than letting one target arbitrate everything.
- Feed real value back into the platforms. Margin, predicted lifetime value, new-customer bonuses through Google's new-customer acquisition goal. Smart Bidding is exceptionally good at hitting whatever number you give it, so give it the right number.
- Report both numbers, every month. The blended figure for context, the new customer figure for decisions. The moment the two diverge, you have learned something important.
What changes when you do
The first effect is usually uncomfortable: the account looks worse on paper. New customer ROAS is always lower than the blended number, and the first honest report tends to prompt hard questions about what the budget has actually been buying. Expect that conversation in month one, budget reallocation in month two, and the first genuinely comparable acquisition numbers by month three. The trap took quarters to form. It takes about one to unwind.
The second effect is the point. Budgets stop retreating to the warm core, scaling decisions get made against a number that reflects real growth, and the gap between what the dashboard says and what the business feels starts to close. It is also what compounding looks like from the inside: our longest-running accounts, like six years of paid search growth at Joma Jewellery, are built on acquisition measured honestly year after year, not on a blended average defended quarter to quarter.
Growth that pays for itself only happens when you measure the growth, not the average.
If your headline ROAS has looked great for a year while the business has grown slower than it should, the trap is probably already sprung. The way out starts with one question: how much of that number is new?
