Your ROAS is lying to you — and that's the problem

Your ROAS looks good. Your growth has stalled. If that's where you are right now, the instinct is to assume a performance problem — wrong creative, wrong channels, under-investment in what's working. So you push harder, test more, optimise the campaigns that are already performing. And nothing fundamentally changes.

The problem is not the performance. It's the measurement.

ROAS — Return on Ad Spend — is the metric most paid media operations are built around. For good reason: it's intuitive, it creates clear accountability, and in the right context it does exactly what it's designed to do. But at a certain point in a brand's growth journey, ROAS stops reflecting business health and starts distorting it. Continuing to optimise toward it doesn't just stop being useful — it actively defunds the investments that would unlock the next phase of growth. This is what that looks like, and what to do instead.

Why ROAS worked — for a while

ROAS is not inherently wrong. In the early stages of building a paid media account it does exactly what it's designed to do.

When a brand is new to paid media, ROAS creates accountability where there wasn't any. It identifies which campaigns convert, which creative connects, which channels earn their budget. When there's genuine demand in the market — audiences who know the brand, people actively searching for what it sells — ROAS is the right lens for capturing that demand efficiently. It rewards precision and penalises waste.

This works well when organic momentum is strong, when the brand is growing and audiences are fresh, and when the lowest-hanging demand simply needs harvesting. ROAS is a metric designed for that specific moment. The problem is what happens next.

Where ROAS runs out of road

Once a brand has harvested the available demand pool — reached the warm audiences, retargeted the engaged visitors, optimised the bestseller campaigns — ROAS-led optimisation has nowhere useful to go. The algorithm finds the path of least resistance, which is always existing customers and audiences that already know the brand. Frequency rises. CTR falls. New customer acquisition slows. Full-price sell-through suffers because the audiences being reached are increasingly familiar, and familiar audiences are more price-sensitive.

The brand is running hard to stay still.

At this point ROAS stops being a floor and starts being a ceiling. Not because the campaigns are poorly managed — technically, they're performing. The problem is structural: the metric optimises for capturing existing demand while the business needs to create new demand. It can't reward what it can't see. Brand building, new audience development, upper-funnel investment — these are largely invisible to last-click attribution.

"A missed opportunity does not appear on a balance sheet."

Rory Sutherland, Vice Chairman of Ogilvy and author of Alchemy, argues that an obsession with measurable ROI produces an efficiency-driven race to the bottom rather than an opportunity-driven race to the top. The observation above is the one that lands most directly here. The new customers never acquired, the brand consideration never built, the markets never entered — none of these register in a ROAS dashboard as a problem. They are invisible losses.

If you haven't read Alchemy, it's worth your time — one of the few books about marketing and human behaviour that will genuinely change how you see the industry.

There is a second layer to this, compounding the first. ROAS doesn't just fail to capture the full picture — it can actively mislead, often without anyone intending it to.

Attribution windows are one of the most underappreciated sources of distortion. A 7-day click, 1-day view window tells a very different story to a 1-day click window — and the difference isn't academic. Wider windows allow campaigns to claim credit for conversions that were already in motion, driven by organic search, direct traffic, or other channel activity. The ROAS number improves. Nothing in the business has changed.

This is rarely deliberate — platform defaults favour wider windows, and without a rigorous attribution audit most brands never question them. Post-iOS 14, the problem has deepened further: modelled conversions now account for a significant proportion of reported Meta results. In plain terms, a meaningful share of what Meta reports as conversions is now an estimate, not a measurement. The gap between what platforms report and what's actually happening in the business has never been wider. The right response isn't to distrust all data. It's to stop treating any single platform's self-reported numbers as the primary view of business health.

What gets defunded

The commercial consequences of ROAS as the primary lens follow a predictable pattern.

Brand campaigns — the ones that build desire, justify premium pricing, and attract genuinely new audiences — cannot win a budget argument against a retargeting carousel of bestsellers. The numbers don't support it. Upper-funnel activity looks expensive and inefficient on a 7-day attribution window. New channels get deprioritised because they don't convert immediately. The result is a media operation that becomes progressively narrower: concentrated on a single platform, increasingly reliant on discounting to sustain conversion rates, and effective at talking to existing audiences while quietly failing to build new ones.

The technology investment often tells a different story. Brands that experienced strong early growth frequently invested in the infrastructure for the next phase — CRM and lifecycle marketing platforms, attribution tools, first-party data infrastructure, international fulfilment and market expansion capabilities. The ambition was real. But the measurement framework didn't keep pace. It was still rewarding demand capture while the business needed demand creation. The tech was forward-looking. The measurement was not.

A better framework: MER and POAS

The solution is not to abandon measurement rigour. It's to adopt a framework that reflects what the business is actually trying to do.

Two metrics do this well, and neither requires a data science team to implement.

Media Efficiency Ratio (MER) is the simplest and most useful: total revenue divided by total media spend.

Unlike ROAS, MER captures the halo effects of brand activity that last-click attribution misses entirely. A YouTube awareness campaign that looks expensive in isolation may be warming audiences who convert through search three weeks later. ROAS misses that contribution. MER sees it.

More importantly, MER creates permission to invest in the things that ROAS systematically defunds. If total revenue grows faster than total spend, MER improves — regardless of how that growth is attributed across individual channels. Brand investment, upper-funnel campaigns, new channel exploration: all of these become defensible when MER is the primary lens.

Profit on Ad Spend (POAS) connects media investment directly to profitability rather than revenue. Where MER measures blended efficiency at scale, POAS asks whether that efficiency is actually producing margin. A campaign running 10x ROAS on orders with a 20% contribution margin delivers a POAS of 2x — meaning $2 of contribution margin retained for every $1 of media spend. That is the number worth tracking. Campaigns inflating ROAS through heavy discounting, or optimising toward low-margin product lines, become immediately visible and accountable.

Together, MER and POAS don't just change the scorecard. They change what the business is allowed to invest in.

This is the empirical case behind Binet and Field's research into brand and activation investment — the evidence that sustained growth requires both long-term brand building and short-term performance activity in roughly a 60/40 ratio. A measurement framework that can't see brand contribution will always underinvest in it. Every time.

What this looks like in practice

With MER as the primary lens, brand investment becomes defensible — evaluated on its contribution to total business revenue over time, not its 7-day ROAS. Upper-funnel channels — YouTube, programmatic, editorial — stop being unaccountable costs and become comparable investments. New channel expansion can be assessed on its contribution to the total system rather than on isolated ROAS.

Incrementality testing — geo holdouts, ghost bidding, brand versus non-brand search splits — becomes the tool that connects this framework to real decisions. It answers the question that ROAS cannot: are these campaigns generating conversions that wouldn't have happened otherwise? A geo holdout test, for example, simply means running your campaigns in some regions while going dark in others — the difference in sales performance between those regions tells you what your media is actually generating, not just claiming credit for.

The goal is not to abandon efficiency. It's to measure efficiency in a way that reflects the full value of media to the business — not just the value that last-click attribution can see.

Your practical first step

You can run this calculation today, without an agency and without a complex data project.

Those questions tell you more about the health of your media operation than any campaign dashboard.

To go further, apply your blended contribution margin to the revenue figure. Now you're measuring what the media investment is actually doing for the business — not what the platforms are claiming credit for.

If this calculation reveals a gap between your reported ROAS and your actual growth trajectory, that gap is worth understanding. It's almost always a measurement problem, not a performance problem. Measurement problems are solvable. And the solution is simpler than most agencies will tell you.