How Much Should You Spend on Marketing? Why Marginal ROAS Matters More Than Average ROAS

A marketing channel generates $500,000 in attributed revenue from $100,000 in advertising spend.

Its return on ad spend is 5.0.

Management is considering an obvious next step:

“If the channel produces five dollars of revenue for every dollar spent, why not increase the budget?”

Because the next $100,000 may behave very differently from the first.

As marketing investment increases, organizations often reach less responsive audiences, pay more for additional impressions or clicks, increase advertising frequency, and exhaust the highest-value opportunities first.

The result is diminishing returns.

This creates an important distinction between two questions:

“How productive has our advertising spend been on average?”

and

“How productive will the next dollar of advertising spend be?”

The first question is related to average return on ad spend.

The second is a marginal decision.

When businesses are deciding whether to increase, decrease, or reallocate marketing budgets, the marginal question is often the more useful one.

Average ROAS Describes the Past

A simplified ROAS calculation is:

ROAS = Revenue Attributed to Advertising / Advertising Spend

Suppose a paid-search campaign spends $100,000 and receives credit for $500,000 in revenue.

ROAS is:

$500,000 / $100,000 = 5.0

That is useful descriptive information.

But it does not establish that another $100,000 of spending will generate another $500,000 of revenue.

Average ROAS summarizes the relationship across the budget already spent.

A budget decision concerns spending that has not happened yet.

That difference is fundamental.

Why Marketing Returns Can Decline as Spending Increases

Imagine a business beginning a campaign with a relatively small budget.

The advertising system can initially prioritize users who are easiest or cheapest to reach.

As the budget expands, several things may happen.

The campaign may need to reach audiences with lower purchase intent.

Advertising frequency may increase among people who have already seen the campaign.

Auctions may become more expensive as the campaign attempts to capture additional inventory.

The organization may also begin exhausting highly efficient keywords, audiences, geographic areas, or placements.

This creates a response curve.

Early investment may generate substantial additional outcomes.

Later investment can still generate value, but each additional dollar may produce less than the previous one.

That is the basic intuition behind diminishing marginal returns.

A Worked Example: A 5.0 ROAS Channel That Should Not Automatically Receive More Budget

Consider the following simplified spending pattern.

At $100,000 of spending, the channel looks extremely attractive.

Average ROAS is 5.0.

But moving from $200,000 to $250,000 generates only $75,000 of additional associated revenue for $50,000 of additional spending.

The marginal ROAS of that increment is:

$75,000 / $50,000 = 1.50

The historical average is still 3.30 at the $250,000 spending level.

Yet the most recent budget increment performs very differently from the average.

This is why using average ROAS alone for budget expansion can be misleading.

The important question is not simply whether a channel has historically performed well.

It is how much additional value another unit of investment is expected to create. Organizations can also compare this analysis with paid media performance analytics to connect channel reporting with allocation decisions.

Revenue Is Not Profit

Even marginal ROAS is incomplete without business economics.

Suppose the company's gross contribution margin before advertising is 40%.

An additional $50,000 of advertising produces $75,000 in incremental revenue.

The associated contribution before advertising would be:

$75,000 × 40% = $30,000

But the additional advertising costs $50,000.

Under these simplified assumptions:

Incremental contribution after advertising = $30,000 − $50,000 = −$20,000

The campaign can therefore generate additional revenue while reducing contribution.

This is an important reason why maximizing revenue and maximizing profit are not the same optimization problem.

Businesses need to understand the economic outcome they actually want to improve.

Marginal ROAS and Incremental ROAS Are Related but Not Identical

These concepts are easy to confuse.

Marginal ROAS asks how much additional revenue is associated with an additional unit of spending.

Incremental ROAS attempts to estimate revenue that would not have occurred without the marketing intervention.

A marginal calculation can still be based on attributed revenue.

If so, it does not automatically establish causality.

For example, increasing branded-search spending might coincide with additional attributed revenue even though some customers would have purchased through organic search without the additional ads.

That is why experiments and the distinction between attribution and incrementality can be valuable.

Ideally, businesses want to understand the incremental response curve: how much additional business outcome is caused by each additional level of investment.

That is considerably more informative for allocation decisions than attribution alone.

Budget Allocation Is a Portfolio Problem

Businesses rarely decide whether to fund only one channel.

They allocate a limited budget across several alternatives.

Suppose a company has an additional $100,000 available.

Its current estimates for the next $25,000 increment are:

If management ranks channels using average ROAS, Paid Search appears strongest.

But the estimated next increment performs better in Paid Social.

That does not mean the company should automatically move all available budget to Paid Social.

Additional spending there may eventually reduce its marginal return as well.

Instead, budget allocation becomes an iterative question:

“Where does the next unit of investment create the greatest expected value?”

As money is allocated, the marginal returns of the alternatives can change.

This is much closer to an optimization problem than a dashboard-ranking exercise.

Marketing Mix Modeling Can Estimate Response Curves

Marketing mix modeling can help organizations estimate how business outcomes relate to different levels of marketing investment over time.

Rather than assuming a linear relationship such as:

10% more spending → 10% more sales

a model can represent saturation and diminishing returns.

A conceptual response curve might rise quickly at first and gradually flatten.

That curve can help estimate where additional spending becomes less productive.

However, the results depend on model specification, data quality, variation in historical spending, control variables, and assumptions.

An estimated response curve should therefore be treated as analytical evidence rather than unquestionable truth.

Where possible, experimental evidence can strengthen or calibrate conclusions derived from observational models.

Experiments Can Measure What Happens When Spend Changes

One way to learn about marginal returns is to deliberately create variation.

For example, comparable geographic markets could receive different advertising intensities.

If carefully designed, the resulting differences can provide evidence about how outcomes respond to changes in spending.

The question becomes:

“What additional outcome occurred when investment increased from one level to another?”

Repeated across several spending levels, this can begin revealing the shape of the response curve.

Experiments are not always practical.

They may require sufficient sample size, geographic comparability, operational coordination, and willingness to temporarily alter marketing activity.

But when major budget decisions are at stake, learning the response curve can have substantial value.

Historical Data Needs Variation to Identify Diminishing Returns

Suppose a company has spent almost exactly $100,000 per month on a channel for three years.

The dataset contains many months, but very little variation in spending.

It may be difficult to estimate what would happen at $50,000 or $200,000 because the organization has rarely observed those conditions.

More rows do not automatically mean more information about every business question.

To estimate a response curve, the data needs meaningful variation in the variable whose effect is being studied.

This is an important consideration when evaluating whether historical marketing data can support budget optimization. Reliable conclusions also depend on consistent data definitions and validated measurement pipelines.

Constraints Make the Real Problem More Interesting

Even if one channel appears to have the highest expected marginal return, businesses face constraints.

A search campaign may have limited search volume.

A social campaign may require additional creative production.

A channel may have contractual minimums.

Management may want to maintain market presence across multiple channels.

Some campaigns may support brand objectives that are difficult to capture through short-term conversion metrics.

There may also be uncertainty around every return estimate.

The real allocation problem therefore looks more like:

Maximize expected business value

subject to:

• total marketing budget
• channel minimums and maximums
• audience capacity
• operational capacity
• risk tolerance
• strategic requirements

This is where marketing analytics intersects directly with decision modeling and optimization.

Scenario Analysis Makes Budget Recommendations More Honest

A budget model should not pretend its estimates are perfectly known.

Suppose Paid Social's expected marginal ROAS is estimated at 3.1.

Management could examine several scenarios:

• Conservative: 2.3
• Base: 3.1
• Optimistic: 3.7

The same analysis can be performed across competing channels.

If Paid Social remains the preferred allocation under a wide range of assumptions, the recommendation is relatively robust.

If a small change causes another channel to become preferable, management should understand that uncertainty before reallocating millions of dollars.

Sensitivity analysis and scenario-based decision modeling therefore turn a point estimate into a decision range.

The Best Budget Is Not Necessarily the Entire Available Budget

Marketing teams are often given a budget and implicitly encouraged to spend it.

But an economically rational allocation can sometimes leave money unspent.

If every remaining marketing opportunity has an expected marginal contribution below its cost or below the organization's required return, additional spending may destroy value.

The analytical question should therefore not always be:

“How should we spend the entire budget?”

It may instead be:

“How much of the available budget should we spend, and where?”

That distinction changes marketing planning from budget distribution to capital allocation.

A Better Marketing Budget Workflow

A practical analytical process can be structured as:

1. Define the business objective
   Revenue, contribution margin, profit, customer acquisition, or another clearly defined outcome.
2. Establish consistent measurement
   Align campaign spend, outcomes, attribution definitions, and financial data.
3. Separate descriptive attribution from causal evidence
   Understand what receives credit and what can reasonably be interpreted as incremental.
4. Estimate response curves
   Use experiments, historical modeling, MMM, or a combination appropriate to the available data.
5. Translate revenue into economics
   Incorporate margin, customer value, variable costs, and other relevant financial consequences.
6. Estimate marginal returns
   Evaluate the expected value of additional investment rather than relying only on historical averages.
7. Apply constraints
   Include budget limits, capacity, channel saturation, and strategic requirements.
8. Test uncertainty
   Use scenarios or sensitivity analysis to determine whether recommendations remain stable.
9. Allocate incrementally
   Direct the next unit of budget toward the highest-value available opportunity and reassess as marginal returns change.

This workflow connects marketing measurement directly to business decision-making. A KPI framework with clear definitions and decision rules helps make that connection operational.

From Campaign Reporting to Marketing Investment Decisions

Marketing dashboards are good at answering:

“What happened?”

Attribution systems can help answer:

“Which observed touchpoints received credit?”

Experiments can help answer:

“What did marketing actually change?”

Response curves can help answer:

“How does that impact change as spending increases?”

And decision models can help answer:

“Given our budget and constraints, what should we do next?”

That final question is ultimately the reason marketing analytics exists.

A channel with excellent historical ROAS is not automatically the best place for the next dollar.

The better question is:

“Which available investment is expected to create the greatest additional business value now?”

That moves marketing analytics from performance reporting toward resource allocation.