The Dangerous Seduction of High ROAS
Every digital marketer has experienced the rush of checking a dashboard and seeing a Return on Ad Spend (ROAS) that looks like a statistical anomaly. A 10x, 15x, or even 20x return suggests that for every dollar you put into the machine, twenty dollars in revenue are pouring out the other side. In many boardrooms, this is the cue to open the champagne and double the budget.
However, for the sophisticated growth marketer, a high ROAS is not always a cause for celebration. Sometimes, it is a warning sign. The fundamental question isn’t just “What is the return?” but rather, “Would this revenue have existed without the ad?”
The gap between reported performance and actual incremental growth is where millions of marketing dollars are lost every year. When an ecommerce company hires a PPC agency, the honeymoon period usually consists of high conversion volumes and a healthy ROAS. On the surface, the strategy is a resounding success. But if you look closer, you might find that the campaign is simply standing in front of a door that was already open. If those conversions would have occurred anyway via direct visits or organic search, the paid campaigns are merely taxing the business rather than growing it.
The eBay Experiment: A Lesson in Causal Lift
To understand the limitations of ROAS, we must look at one of the most famous case studies in the history of paid search: the eBay experiment. In 2013, researchers from the University of California, Berkeley, teamed up with eBay to analyze the effectiveness of the company’s massive spend on branded search terms.
At the time, eBay was spending millions of dollars bidding on its own name. Their internal metrics showed a massive ROAS. However, the researchers conducted a controlled experiment: they turned off paid search ads for the keyword “eBay” in specific geographic regions while keeping them active in others. The results were startling. In the regions where the ads were turned off, organic traffic picked up nearly 100% of the lost clicks. The revenue remained almost identical.
The conclusion was clear: eBay was paying for traffic it already owned. Despite this evidence, many brands continue to spend heavily on brand keywords. Sometimes this is a defensive move to prevent competitors from poaching the top spot, but often it is a “safe” way to inflate reported ROAS. Platforms love these campaigns because they provide high-confidence conversions, but from a business growth perspective, they represent zero incremental value.
The Black-Box Trap: Performance Max and Advantage+
As digital advertising moves toward total automation, the difficulty of measuring true growth has intensified. Modern advertising tools, such as Google’s Performance Max (P-Max) and Meta’s Advantage+, are essentially black boxes. They use machine learning to find the users most likely to convert, but they don’t necessarily prioritize finding *new* customers.
Algorithms are designed to achieve the goal you set for them. If you tell an algorithm to maximize ROAS, it will find the path of least resistance to a conversion. Often, this path leads straight to your existing customers. Automation thrives on “safe” signals, which often results in the following:
- Brand Search Cannibalization: Algorithms bid aggressively on your brand name because those users are the most likely to buy.
- Aggressive Retargeting: The system serves ads to users who have already added items to their cart and were seconds away from checking out.
- Reporting Bias: Platforms claim credit for any user who saw an ad and eventually purchased, even if the ad had no influence on their decision.
Without a way to measure incrementality, automation simply amplifies these non-incremental signals. You may see your ROAS climb, but your total business revenue remains stagnant. You aren’t scaling your business; you are scaling your platform spend.
Incrementality: Measuring Causal Impact
Incrementality is the gold standard for measuring marketing effectiveness. It refers to the “causal lift” created by a specific campaign. In simpler terms, it answers the question: “What changed because this campaign existed?”
While platform attribution tells you which channel was the last touchpoint before a sale, incrementality tells you if the sale would have happened in a world where that channel was turned off. This is a much more useful lens for budget allocation. A channel can have a fantastic in-platform ROAS and still generate a weak incremental impact if it is merely harvesting demand rather than creating it.
Think of it this way: Attribution is like a scoreboard in a basketball game. It tells you who took the last shot. Incrementality is like an advanced scouting report. It tells you how much better the team performs when a specific player is on the court versus when they are on the bench. If the team scores the same amount of points regardless of whether that player is playing, that player’s “incrementality” is zero, regardless of how many shots they take.
The Difference Between Demand Generation and Demand Capture
To master incrementality, you must distinguish between campaigns that create new demand and those that capture existing demand. High-funnel activities, such as YouTube awareness ads or social media prospecting, often have lower reported ROAS because they are introducing people to the brand for the first time. However, their incrementality is often very high because they are moving people who would never have considered your brand into the sales funnel.
Conversely, bottom-of-funnel activities like branded search and retargeting often have astronomical ROAS but low incrementality. They are simply capturing the demand that was created by your high-funnel activities, your brand reputation, or word-of-mouth.
The Hidden Metric: Marginal ROAS
Even if you prove that a channel is incremental, you still need to know how much to spend on it. This is where Marginal ROAS comes into play. Marginal ROAS measures the return on the *next* dollar of spend, rather than the average return across the entire budget.
Every marketing channel is subject to the law of diminishing returns. The first $1,000 you spend usually targets your “low-hanging fruit”—your most loyal customers and highest-intent prospects. As you increase spend to $10,000 or $100,000, you are forced to target audiences that are less familiar with your brand or less likely to convert immediately. Consequently, the efficiency of those extra dollars drops.
A Practical Example of Marginal ROAS
Imagine your current spend is $10,000, and it generates $50,000 in revenue. Your average ROAS is 500%. You decide to scale the budget by adding another $5,000. This additional spend only generates an extra $5,000 in revenue.
Your new dashboard metrics would look like this:
- Total Spend: $15,000
- Total Revenue: $55,000
- New Average ROAS: 366%
On the surface, a 366% ROAS still looks pretty good. However, your Marginal ROAS for that last $5,000 was 100%. You essentially traded $1 for $1. Once you factor in product costs, shipping, and overhead, you actually lost money on that expansion. The “average” ROAS hides the fact that the last $5,000 you spent was entirely wasted. This is the trap of scaling based on averages.
How to Test for True Incrementality
Moving away from a ROAS-centric view requires a commitment to testing. You don’t need a massive data science team to start measuring incrementality; you just need a structured approach to experimentation.
1. Geo-Split Testing
This is one of the most reliable ways to measure lift. You divide your target markets into two comparable geographic groups—for example, two sets of mid-sized cities with similar demographics and historical sales patterns. In the “test” group, you run your ads as usual. In the “control” group, you turn the ads off or significantly reduce the spend. The difference in total revenue between these two regions over a set period reveals the true incremental lift of your advertising.
2. Intentional Holdout Groups
Many platforms now offer built-in conversion lift or search lift tools. These tools allow you to create a “holdout group”—a small percentage of users (usually 5-10%) who are intentionally prevented from seeing your ads. By comparing the conversion rate of the group that saw the ads to the group that didn’t, you can calculate the exact percentage of sales that were driven by the ad versus those that would have happened anyway.
3. Brand Spend “Pulse” Testing
If you suspect your branded search spend is non-incremental, try a “pulse” test. Turn off branded search for 48 to 72 hours and monitor your organic search traffic for those same keywords. If organic traffic spikes to fill the gap, your branded ads were likely cannibalizing “free” clicks. If total traffic drops significantly, the ads were providing a necessary defensive or navigational function.
Shifting the Culture: From Reporting to Capital Allocation
The real challenge in moving beyond ROAS is not technical; it is cultural. Most marketing teams are incentivized to report “good news.” High ROAS is the ultimate “good news” metric. Admitting that a high-ROAS campaign isn’t actually driving growth feels like a failure, but it is actually the first step toward genuine efficiency.
The role of the modern marketing leader is shifting from a reporter of performance to an allocator of capital. When you view your marketing budget as an investment portfolio, your priorities change:
- ROAS tells you who is taking the credit for the sales.
- Incrementality tells you which investments actually moved the needle.
- Marginal ROAS tells you where the next dollar of capital should be deployed.
By focusing on these three pillars, you move marketing out of the realm of “discretionary spending” and into the realm of “growth engineering.” You stop chasing vanity metrics and start chasing business outcomes that actually show up on the balance sheet.
Conclusion: The Path to Smart Scaling
In a world of automated bidding and black-box algorithms, a great ROAS is easy to achieve, but real growth is harder than ever to find. If your dashboard says you’re winning but your bank account says you’re plateauing, it’s time to stop looking at averages and start looking at incrementality.
Don’t be afraid to turn things off. Don’t be afraid to challenge the platform’s version of the truth. The goal of marketing is not to have a pretty dashboard; it is to create revenue that wouldn’t have existed otherwise. Focus on the next dollar, test your assumptions, and remember: if a metric looks too good to be true, it’s probably not driving the growth you think it is.