Marginal Returns & Incrementality: Why Your Best-Performing Channel May Not Deserve the Next €1

 

From Historical Performance to Marginal Value, Incremental Growth and Better Paid Media Budget Allocation

Performance marketing gives us an enormous amount of data for deciding where to invest.

ROAS, CPA, conversion rate, revenue, new customers and contribution margin can all help us understand how campaigns are performing. When one channel consistently produces stronger numbers than another, increasing its budget can appear to be the obvious decision.

But there is an important difference between two questions:

How has this channel performed?

and

What is likely to happen if I give this channel more money?

Those questions sound similar. They are not.

A channel can have the strongest ROAS in the media plan while the additional budget being added to it produces progressively weaker returns. A campaign can also report substantial attributed revenue even though some of those customers would have purchased without seeing the advertising.

This is where marginal returns and incrementality become useful.

I use them to answer different parts of the same investment problem.

Marginal returns help me understand what additional spend produces.

Incrementality helps me understand how much additional business advertising actually creates.

Put the two together, and the budget conversation becomes much more useful:

Where is the next €1 most likely to create additional business value?

What Are Marginal Returns in Paid Media?

Start with a simple eCommerce campaign.

Suppose I spend:

€100,000

and generate:

€500,000 in attributed revenue

The campaign has produced a 5x ROAS.

Now I increase the budget by €20,000.

Total spend becomes:

€120,000

and total attributed revenue increases to:

€540,000

The overall campaign still looks healthy:

€540,000 ÷ €120,000 = 4.5x ROAS

But I am interested in something else.

What did the additional €20,000 produce?

Revenue increased from €500,000 to €540,000.

So:

Additional Spend: €20,000

Additional Revenue: €40,000

Marginal ROAS: 2x

That is very different from looking at the 4.5x blended ROAS.

Metric	Before Increase	After Increase	Additional Investment
Spend	€100K	€120K	€20K
Revenue	€500K	€540K	€40K
ROAS	5.0x	4.5x	2.0x marginal

 

The original €100,000 performed very well.

The next €20,000 still generated revenue, but at a considerably lower rate.

That is the basic idea behind marginal returns.

When I am deciding where to place new budget, the historical average remains useful, but I also want to understand what happened to the latest increments of investment.

Why Do Marginal Returns Change as Spend Increases?

Paid media rarely scales in a perfectly straight line.

If a campaign generates €500,000 from €100,000, I cannot automatically assume that €200,000 will generate €1 million.

There are several practical reasons.

In paid search, the first portion of budget might capture highly relevant queries with strong commercial intent. Additional budget may require expanding into broader queries or increasingly competitive auctions.

In paid social, the initial investment might reach the strongest audience segments efficiently. As spend grows, reach expands, frequency changes and the campaign may need to find conversions among increasingly broader audiences.

In programmatic advertising, additional investment can mean expanding inventory, audiences, geographies, formats or frequency.

The mechanics differ by channel, but the underlying investment problem is similar.

Imagine the following simplified example:

Spend	Revenue	Blended ROAS
€50K	€350K	7.0x
€100K	€550K	5.5x
€150K	€645K	4.3x
€200K	€720K	3.6x

The channel is still generating substantial revenue at €200,000.

But look at each additional €50,000.

€0 → €50K: +€350K revenue

€50K → €100K: +€200K

€100K → €150K: +€95K

€150K → €200K: +€75K

The average ROAS tells me how the total investment performed.

The marginal view shows me how the economics changed as I kept adding money.

That difference becomes extremely important when I have another €50,000 available and several channels competing for it.

What Is Incrementality?

Marginal returns deal with additional investment.

Incrementality deals with a different question:

What would have happened without the advertising?

Imagine an eCommerce campaign reports:

1,000 purchases

It is tempting to interpret those 1,000 purchases as the result of advertising.

But customers do not disappear when advertising disappears.

Some may already know the brand.

Some may have visited the website organically.

Some may regularly purchase from the business.

Some may have searched specifically for the brand.

Some may have been influenced by another channel.

Suppose a properly designed experiment compares an exposed group with an appropriate control group and estimates that, without the advertising, approximately 700 purchases would still have occurred.

Observed purchases:

1,000

Estimated purchases without advertising:

700

Estimated incremental purchases:

300

Those 300 purchases represent the estimated additional outcome caused by the advertising.

That is incrementality in practical terms.

Attributed Performance and Incremental Performance Can Tell Different Stories

Consider a customer who already buys the same skincare product every six weeks.

They see a retargeting ad two days before their usual purchase, click it and buy.

Depending on the attribution methodology, the advertising platform may receive credit for the transaction.

That attribution can be perfectly valid according to the reporting rules.

But the business question is different:

Would that customer have purchased anyway?

Now consider another person who has never purchased from the brand.

They see an ad, discover the product, compare it with alternatives and eventually become a customer.

The reported transaction may look similar inside a dashboard.

Its incremental value could be very different.

This is why I would not treat:

Attributed Revenue

and

Incremental Revenue

as interchangeable concepts.

Suppose a campaign reports:

€500,000 attributed revenue

An experiment estimates that the campaign generated:

€200,000 incremental revenue

The €500,000 still tells me something useful about conversions associated with the campaign.

The €200,000 answers a different question about the additional revenue the advertising appears to have created.

For investment decisions, I want to understand both.

Marginal Returns and Incrementality Answer Different Questions

These concepts are sometimes discussed together, but I find it useful to keep their jobs separate.

Concept	Question
Blended Performance	How has the total investment performed?
Marginal Return	What did the additional investment produce?
Incrementality	What happened because of the advertising?
Marginal Incremental Value	What additional business value is the next unit of investment creating?

This is where the budget conversation becomes more interesting.

Suppose a channel has historically produced excellent performance.

That does not tell me how efficiently it can absorb another €100,000.

And if it can absorb that money while producing more attributed conversions, I still want to understand how much of that additional performance represents genuinely incremental business.

The two concepts therefore meet at the point where I make the investment decision.

Why the Best-Performing Channel May Not Deserve the Next €1

Imagine a hypothetical eCommerce advertiser running three channels.

Channel	Current Monthly Spend	Blended ROAS	Marginal ROAS
Google Ads	€200K	6.5x	2.4x
Meta	€150K	4.8x	3.7x
Programmatic Advertising	€75K	3.5x	3.1x

If I look only at blended ROAS, Google Ads appears comfortably ahead.

If management gives me another €50,000, it might therefore seem logical to put the money into Google Ads.

But the marginal numbers tell me something important.

At its current level of investment, additional Google Ads spend is producing 2.4x.

Additional Meta spend is producing 3.7x.

Additional programmatic advertising spend is producing 3.1x.

That does not mean I should automatically move the next €50,000 to Meta.

I still need to understand incrementality, margins, customer quality, campaign objectives, channel interactions and how much additional spend each channel can absorb.

But I now have a much better question to investigate.

Google Ads can simultaneously be:

the strongest historical performer

and

a weaker destination for the next budget increment.

There is no contradiction between those two statements.

They measure different things.

Now Add Incrementality

Suppose we run appropriate incrementality studies and discover another layer.

The advertiser has strong existing brand demand.

A meaningful share of conversions reported through search comes from customers actively looking for the brand.

Meta is reaching a broader mix of existing and prospective customers.

Programmatic advertising is contributing earlier in the journey and has a different attribution profile.

Now the simple ROAS ranking becomes even less useful for incremental budget allocation.

I would want to build a view closer to this:

Channel	Blended Performance	Marginal Performance	Incrementality Evidence	Scaling Headroom
Google Ads	Strong	Lower at current spend	Mixed by campaign type	Moderate
Meta	Strong	Stronger on recent increments	Positive test evidence	Higher
Programmatic Advertising	Moderate	Competitive	Requires appropriate testing	Moderate

These are hypothetical values, but they illustrate the decision process.

I am no longer asking which platform dashboard shows the biggest number.

I am asking:

What is each channel contributing at its current level of investment, and what is likely to happen if I add more money?

The Response Curve: What Scaling Actually Looks Like

A useful way to visualize marginal returns is through a response curve.

Put media spend on the horizontal axis and business outcomes on the vertical axis.

At lower levels of investment, additional spend may produce substantial gains.

As investment increases, the curve can begin to flatten.

Conceptually:

Spend ↑

Business Outcome ↑

but eventually:

Additional Business Outcome per Additional € ↓

This matters because media planning sometimes assumes scaling is approximately linear.

A campaign currently producing 5x ROAS receives another €100,000, and the forecast assumes roughly another €500,000 in revenue.

That may happen.

But it should not be assumed simply because the historical average was 5x.

The more useful forecasting question is:

Where are we currently sitting on the response curve?

A channel with lower historical ROAS but substantial scaling headroom may produce more value from additional investment than a historically stronger channel already operating much further along its curve.

How Can We Measure Incrementality?

Incrementality needs some form of counterfactual: an estimate of what would have happened without the advertising intervention.

The appropriate method depends on the channel, business, geography, available data and scale.

Randomized Holdouts

A portion of an eligible population is withheld from advertising and compared with the exposed population.

If the exposed group converts at a meaningfully higher rate, the difference can help estimate incremental lift.

For example:

Control group: 10,000 purchases

Exposed group, adjusted to comparable population: 11,500 purchases

Estimated lift:

+15%

The experimental design, statistical confidence and population comparability matter, but the business idea is straightforward.

Conversion Lift Experiments

Some advertising environments support controlled experiments designed specifically to estimate the conversions caused by advertising.

Rather than relying only on attributed conversions, I can examine the difference between exposed and control populations.

Geo Experiments

Different geographic areas can be used as test and control markets.

Imagine increasing Meta investment by 50% across a carefully selected group of German regions while maintaining comparable control regions.

If sales develop differently between test and control after accounting for baseline differences, that can provide evidence of incremental impact.

Matched Markets

Markets with similar historical behaviour can be paired, with advertising changed in one group and maintained differently in another.

This can be useful when individual-level randomization is unavailable or unsuitable.

Econometric and Model-Based Approaches

For businesses with substantial historical data across channels, markets and time periods, modelling can help estimate the relationship between advertising investment and business outcomes.

These approaches can be particularly useful for understanding broader media portfolios, although the quality of the result depends heavily on the quality of the data and methodology.

There is no single method I would force onto every campaign.

The measurement design should match the business question.

Marginal Incremental Economics

This is where the two ideas become especially useful together.

Imagine I add €50,000 to a channel.

The platform reports:

€150,000 additional attributed revenue

Marginal ROAS:

3x

Useful.

Now suppose an incrementality study suggests that approximately €90,000 of the additional revenue was actually incremental.

The picture becomes:

€50K Additional Spend

→ €150K Additional Attributed Revenue

→ €90K Estimated Incremental Revenue

Then I bring in the business economics.

Suppose gross margin on those sales is 50%.

€90K incremental revenue

→ €45K incremental gross margin

At that point, the investment decision looks very different from simply celebrating a 3x marginal attributed ROAS.

Depending on the business, I may also need to consider:

New Customer Rate | Average Order Value | Contribution Margin | Repeat Purchase | Customer Lifetime Value | Returns | Discounts | Variable Costs

The analysis moves progressively closer to what the business actually earns.

Additional Spend → Marginal Outcome → Incremental Outcome → Revenue → Margin → Business Value

That is the chain I ultimately care about.

A Hypothetical German eCommerce Media Plan

Consider a German eCommerce retailer investing €500,000 per month across paid media.

Its current allocation is:

Channel	Monthly Spend
Google Ads	€220K
Meta	€160K
Programmatic Advertising	€70K
CTV	€50K
Total	€500K

The business approves another €100,000 per month for growth.

The easiest approach would be to rank the channels by historical ROAS and give most of the additional budget to the strongest performer.

I would want to go further.

Google Ads

Historical performance is excellent.

But I would separate brand, non-brand, Shopping and other campaign types and examine how marginal performance changes as investment increases.

If brand demand is already heavily captured, additional spend there may have limited headroom.

Non-brand and Shopping may show a different response curve.

The channel should not be treated as one number.

Meta

Suppose recent budget increases have maintained relatively strong marginal acquisition economics.

Incrementality testing also indicates that the platform is generating additional purchases rather than simply receiving attribution for existing demand.

That gives me evidence to test further scaling.

Programmatic Advertising

Suppose reported ROAS is lower because the channel operates across a broader part of the customer journey.

I would examine incremental reach, audience strategy, downstream customer behaviour and appropriate experimental evidence before deciding whether the lower attributed ROAS means the channel deserves less investment.

CTV

Direct platform attribution may provide an incomplete picture of its role.

A geo experiment or another suitable measurement approach could help evaluate whether additional CTV investment changes total business outcomes within exposed markets.

Now the additional €100,000 becomes a portfolio allocation problem.

I might test portions of the incremental budget across the channels with the strongest combination of:

Marginal Return

Incrementality Evidence

Customer Economics

Scaling Headroom

and then observe how the response curves change.

The allocation is not predetermined by whichever platform currently reports the highest ROAS.

From Channel Performance to Portfolio Performance

Customers do not organize their behaviour according to our advertising platforms.

Someone might:

See a CTV ad

→ encounter a Meta campaign

→ visit the website directly

→ leave

→ conduct a Google search three days later

→ click a Shopping ad

→ purchase

Several platforms may report or influence parts of that journey.

If I optimize every channel independently around the conversions it claims, I can end up with several platforms competing for credit around the same underlying demand.

This is why marginal returns and incrementality become particularly useful at the portfolio level.

The objective is not to determine which channel can claim the largest share of the customer's journey.

I want to understand how changing investment across the portfolio changes the business outcome.

That changes the conversation from:

“Which channel has the highest ROAS?”

to:

“What happens to total business performance when I increase, decrease or redistribute media investment?”

My Next €1 Investment Framework

When additional budget becomes available, this is the sequence I would work through:

Current Investment

↓

Blended Performance

↓

Marginal Return

↓

Incremental Contribution

↓

Revenue & Margin

↓

Scaling Headroom

↓

Cross-Channel Opportunity

↓

Next €1

Each layer answers a different question.

Historical performance tells me where I have been.

Marginal performance tells me what happened as investment changed.

Incrementality helps estimate what advertising actually added.

Business economics tells me whether that additional outcome created enough commercial value.

Scaling headroom tells me whether the opportunity can absorb more investment.

Cross-channel analysis tells me whether another part of the portfolio offers a stronger opportunity.

Only then do I reach the next budget decision.

What I Would Monitor After Moving the Budget

Reallocation is not the end of the process.

It creates new information.

If I increase a channel from €100,000 to €130,000, I want to understand how that additional €30,000 behaved.

If performance remains strong, there may be more headroom.

If marginal returns deteriorate quickly, another channel may become more attractive.

I would therefore continue monitoring a combination of:

Marginal ROAS | Incremental Revenue | Incremental CAC | New Customer Acquisition | Contribution Margin | Reach | Frequency | Audience Saturation | Response Curve | Customer Value

The exact metrics will depend on the business.

The important point is that the next allocation decision should use what I learned from the previous one.

Budget allocation becomes iterative:

Invest → Measure → Test → Learn → Reallocate → Measure Again

Closing Thoughts

The channel with the highest ROAS can be an excellent channel.

It can also be approaching the point where additional investment produces much weaker returns.

A channel with lower historical performance may still have substantial headroom and generate stronger returns from the next budget increment.

And attributed performance alone cannot tell me how much of the observed business would have happened without advertising.

This is why I find marginal returns and incrementality so useful together.

Marginal returns tell me what happened when I added more money.

Incrementality helps me understand what additional business the advertising created.

Business economics tells me whether that additional value was worth buying.

Once I have those pieces, the budget conversation becomes much more useful.

The question is no longer only:

Which channel performed best?

The question I want to answer is:

Where is the next €1 most likely to create incremental business value?