Incremental ROAS vs Attributed ROAS: How to Find Wasted Marketing Spend

A very high marketing ROAS can be a warning sign.

That sounds counterintuitive. If a channel is reporting an 8x or 10x return, why would you question it?

Because attributed ROAS and incremental ROAS answer fundamentally different questions.

Attributed ROAS asks how much revenue a channel can claim according to an attribution model. Incremental ROAS asks how much additional revenue was actually caused by the marketing investment.

That difference matters enormously when allocating budget.

A paid search campaign can legitimately receive credit for a conversion from someone who was already planning to buy. A retargeting campaign can show an exceptional ROAS because it deliberately targets people who have already visited the website. A conversion campaign can become increasingly efficient at identifying people who were going to purchase anyway.

The attribution system is not necessarily broken when this happens. It may be doing exactly what it was designed to do.

The mistake is taking attributed revenue and treating it as incremental revenue.

That is how marketing teams can report improving ROAS while revenue growth slows, marginal returns collapse and an increasing share of the budget goes towards harvesting demand that already exists.

What is attributed ROAS?

Attributed ROAS measures the revenue credited to marketing activity relative to the amount spent.

If a campaign spends £100,000 and the attribution system credits it with £500,000 of revenue, the attributed ROAS is 5.

The important word is credited.

The attribution model has rules for deciding which marketing interactions receive credit for a conversion. Those rules can be simple, such as last-click attribution, or considerably more sophisticated.

But the underlying problem remains.

Observing an interaction before a conversion does not establish that the interaction caused the conversion.

Someone searches for your brand, clicks a paid ad and buys a £2,000 product. The paid search campaign receives the conversion.

That tells you something useful about how the customer reached the website.

It does not tell you whether the customer would still have purchased if the ad had not existed.

That second question is the one budget allocation actually cares about.

What is incremental ROAS?

Incremental ROAS attempts to measure the additional revenue caused by the marketing intervention.

Instead of asking which channel should receive credit for a conversion, the question becomes:

Would this revenue have happened anyway?

If spending £100,000 generates £500,000 of attributed revenue but only £150,000 of that revenue would disappear without the advertising, the commercial picture changes considerably.

Attributed ROAS is 5.

Incremental ROAS is 1.5.

The difference between those numbers is not academic. It determines whether the next £100,000 should stay in the channel, move somewhere else or return to the business as margin.

This is why incrementality testing matters. The objective is to construct a credible counterfactual and compare what happened with advertising against what would probably have happened without it.

Attribution distributes credit.

Incrementality attempts to establish causality.

Why the channels closest to the sale often report the highest ROAS

Most attribution systems naturally reward marketing activity that sits close to the conversion.

Brand search is the obvious example.

Imagine someone returning home. Every time they arrive, they walk through the front door. If you only observed the final part of that journey, you could conclude that the door was responsible for bringing them home.

It was not.

The door captured the arrival. It did not create the intention to return.

Brand search can play a similar role.

Someone sees advertising, hears about the company from a friend, passes one of its stores, watches a review and becomes increasingly interested in the product. Two weeks later, they search the brand name on Google, click the paid result and purchase.

Paid search is easy to observe, so the conversion gets attributed there.

But most of the work that created the demand happened before the search.

Retargeting creates a similar measurement problem. The reason somebody enters a retargeting audience is usually that they have already demonstrated some form of intent. They visited the website, browsed a product, added something to their basket or interacted with the brand.

Retargeting deliberately selects people with a higher probability of purchasing.

A high conversion rate among those users therefore tells us very little about what would have happened without the retargeting campaign.

The important question is not whether retargeted customers convert.

It is whether more customers convert because they were retargeted.

Why attributed ROAS can stay high while incremental ROAS collapses

This becomes more important as spend increases.

Early investment in a performance channel can be highly incremental. There is available demand to capture and the campaign reaches customers who genuinely need the additional advertising exposure to convert.

As budget increases, however, the highest-value opportunities are progressively exhausted.

The platform needs somewhere to put the next pound.

It expands audiences, increases frequency, bids more aggressively and finds people increasingly close to purchasing already.

Spend continues rising, but the pool of available high-intent demand does not necessarily rise with it.

Eventually the marginal return begins to flatten.

This is what we call the Death Zone.

The channel is still producing conversions. The attribution dashboard still reports revenue. The average ROAS across the entire campaign can still look perfectly healthy.

But the last portion of the budget may be generating almost no incremental growth.

This distinction between average and marginal return is critical.

Imagine the first £1 million of spend generates substantial incremental revenue. The next £500,000 creates less. Another £500,000 creates less again.

The dashboard can average those results together and report an apparently attractive ROAS across the full £2 million.

That does not mean the next £100,000 is a good investment.

Capital allocation happens at the margin.

The business needs to understand what the next pound is likely to return, not simply what the average pound returned historically.

Why an exceptionally high ROAS deserves investigation

Marketing teams are trained to investigate poorly performing campaigns.

I think unusually high attributed ROAS deserves scrutiny too.

A 15x or 20x attributed return might mean you have discovered an extraordinary marketing opportunity.

It might also mean the campaign has positioned itself extremely close to demand that already exists.

Brand search is a classic example. Retargeting can be another. Affiliate programs, shopping campaigns and heavily optimised conversion campaigns can face similar issues depending on how they are configured and measured.

The closer the marketing activity gets to an existing purchase decision, the easier it becomes to claim the conversion.

So when I see an exceptional attributed return, I want to know:

Who is being targeted?

How much intent did they already demonstrate?

What would happen if the advertising disappeared?

Would another channel, particularly organic search or direct traffic, absorb part of the reported revenue?

Has revenue grown as spend increased, or have attributed conversions simply moved between channels?

What does the marginal response curve look like?

The goal is not to distrust every high-ROAS campaign.

It is to understand what the number is actually measuring before allocating more capital to it.

How attributed ROI targets can make the problem worse

This is where a measurement problem becomes an incentive problem.

Leadership understandably wants accountability from marketing. One of the easiest ways to create that accountability is to set an ROI target.

The marketing team responds rationally.

If their objective is to increase attributed ROI, they move budget towards the activity most capable of producing attributed conversions.

More brand search.

More retargeting.

More high-intent audiences.

More conversion optimisation.

Less activity whose effect appears earlier in the journey and is harder to attribute directly.

The dashboard improves.

But the marketing system becomes increasingly concentrated on people who already know the company and are already close to buying.

Eventually there is less new demand entering the system.

The company responds by optimising harder against the remaining demand.

This can create a nasty feedback loop.

Demand creation weakens. Performance campaigns have a smaller pool to harvest. Marginal acquisition costs increase. The response is to move even more money towards the channels reporting the clearest short-term ROI.

The measurement framework is now encouraging the business to solve a demand problem with more demand capture.

The individual marketers involved are not behaving irrationally. They are responding to the incentives created by the numbers they are being asked to maximise.

That is why measurement design matters beyond analytics.

What you measure changes where the money goes.

How to identify marketing spend that is no longer incremental

You do not always need to begin with a complicated causal model.

Start by looking at what you can observe.

Plot spend against revenue over time.

Then look further up the funnel.

What happened to qualified website sessions, product interactions, pricing-page activity, store-locator usage, branded demand or other meaningful indicators of customer intent as spend increased?

If conversion spend rises substantially while these signals flatten, that is worth investigating.

Look at the spend-response curve by channel and campaign. You are trying to identify the point where additional investment stops producing a proportional change in business behaviour.

This is not yet proof of causality.

It is a hypothesis.

That distinction matters.

Observed data can tell you where something looks wrong. Incrementality testing can then tell you whether the revenue would actually have disappeared without the spend.

The sequence should be:

Observe the problem. Form the hypothesis. Test the decision.

Not every question requires an experiment. But if you are considering moving millions of pounds because you believe a large part of a channel is non-incremental, that is exactly the kind of decision where causal evidence becomes valuable.

What happens when you reduce spend?

One of the most useful tests is often surprisingly simple conceptually.

Reduce the activity you believe has entered diminishing returns and observe what happens to total business outcomes.

Do not simply look at the conversions reported by the platform being cut. Of course those will fall.

Look at total revenue.

Look at total conversions.

Look at organic substitution.

Look at direct traffic.

Look at the rest of the marketing mix.

If a brand-search campaign reports £2 million of revenue and you reduce it, some of those conversions may simply move to organic search.

The platform lost attributed revenue.

The business did not lose revenue.

That difference is exactly what you are trying to identify.

In one example from our own work, the diminishing-return curve suggested the conversion budget was roughly 30% overextended. That portion was reduced and topline revenue did not move.

The point is not that every company has exactly 30% of its budget to cut.

It is that the attributed performance of a channel does not tell you where its incremental return reaches zero.

You have to measure that separately.

Brand search is a useful place to start

Brand search is particularly interesting because the mechanism is relatively intuitive.

If nobody else is bidding aggressively on your brand terms and the organic result already dominates the page, how much incremental value is the paid advertisement creating?

The answer is not automatically zero.

Competitors may be bidding on the terms. Paid listings may protect important traffic. Search layouts vary. Organic rankings can change. Different audiences can behave differently.

That is why the decision should be based on evidence rather than ideology.

Start by assessing competitive pressure.

Who is bidding against you?

What is their impression share?

How much branded organic traffic are you already receiving?

Then test incrementality.

Depending on the business and available data, that might involve a geographic holdout or another controlled reduction in paid brand activity.

If paid clicks fall sharply while total conversions remain largely unchanged because organic absorbs the demand, you have learned something commercially useful.

The purpose is not to prove that brand search is bad.

The purpose is to establish how much you actually need to spend defending existing demand.

Retargeting deserves the same treatment

Retargeting frequently reports impressive attributed returns because its audience selection is doing much of the work before the advertisement is served.

Someone who visited a £3,000 product page yesterday is already very different from a random member of the population.

Serving them an advertisement and observing that they later purchased does not establish that the advertisement caused the purchase.

Again, the question is counterfactual.

How many of those customers would have returned anyway?

A properly designed incrementality test can compare outcomes between eligible users or markets that receive the intervention and an appropriate control.

If the attributed ROAS is 12 but the incremental ROAS is 2, the campaign may still be worth running.

But it should be funded based on the 2, not the 12.

That distinction is the entire point.

Attribution is still useful

None of this means attribution should disappear.

Attribution can be useful operationally.

It helps teams understand observable customer paths, compare campaign activity, diagnose tracking problems, inspect creative interactions and understand how users move through parts of the digital journey.

The problem begins when the question changes from:

Where did we observe the conversion?

to:

What caused the conversion?

Those require different evidence.

You cannot turn attribution into incrementality by adding more touchpoints.

A more sophisticated map of the observed journey is still a map of the observed journey.

Some of the most important influences on purchase happen outside what the tracking system can see at all. Brand memories, previous product experiences, offline exposure, recommendations, competitor perceptions and conversations with other people may all contribute to the final decision.

That is why attribution should be treated as one measurement lens rather than the mechanism that decides the entire marketing budget.

The five questions I would ask before increasing spend on a high-ROAS channel

Before allocating more money because a channel reports an exceptional ROAS, I would want clear answers to five questions.

1. How close is the audience to purchasing already?

The more intent the audience has demonstrated before exposure, the more carefully attributed performance should be interpreted.

2. Is revenue growing with spend?

Do not only inspect attributed conversions. Look at the relationship between channel investment and total business outcomes.

3. What is the marginal return?

Average ROAS tells you how the historical budget performed as a whole. The budget decision concerns the next tranche of spend.

4. What happens when the activity is reduced?

If attributed conversions disappear but total sales do not, another part of the system was probably capturing demand that already existed.

5. What causal evidence do we have?

When the allocation decision is large enough to matter, use an incrementality test to estimate what would have happened without the intervention.

These questions change the conversation from reporting performance to allocating capital.

Incremental ROAS is ultimately a capital allocation metric

The objective of marketing measurement is not to produce the highest possible ROAS on a dashboard.

It is to decide where the next pound should go.

Sometimes the answer will be to increase performance spend because the incremental response remains strong.

Sometimes it will be to hold investment because the return is attractive but approaching saturation.

Sometimes it will be to reduce spend because the final portion is capturing customers who would have purchased anyway.

And sometimes the best decision will be to take money out of exhausted demand capture and invest it in creating more future demand.

That is the part attributed ROAS cannot tell you by itself.

A channel can be excellent at capturing existing demand and still be a poor destination for the next pound.

The most dangerous assumption is that because marketing can claim a sale, it created the sale.

It is not about which channel deserves the credit.

It is about which investment created the growth.

Growth Dynamics helps B2C brands understand where marketing spend is genuinely incremental, where marginal returns have flattened, and where capital should be reallocated. We combine observed business data with incrementality testing and broader marketing measurement to produce an evidenced invest, divest or hold decision.

Because a dashboard telling you where revenue was attributed is useful.

Knowing where your next dollar should go is considerably more valuable.