Media Mix Optimization: How Diminishing Return Curves Turn MMM into Budget Decisions

• Media mix optimization transforms Marketing Mix Modeling from historical analysis into forward-looking budget allocation decisions
• The critical shift is from Average ROI (what happened) to Marginal ROI (what the next dollar will return)
• Diminishing return curves mathematically capture how each channel saturates at higher spend levels
• Optimal allocation occurs when marginal ROI is equalized across all channels; the equimarginal principle
• Models must be validated with incrementality testing to ensure curves reflect the causal relationship between Media and Sales, and not just historical correlation

You’re staring at a spreadsheet showing that your brand spent $127 million on marketing last year. Search drove $43M in attributed revenue. Social claimed $38M. TV says it influenced $52M. Your analytics dashboard shows a blended ROAS of 3.2x.

But here’s the question keeping you up at night: If you had to cut 20% of that budget tomorrow, which channels would you touch? And if the board gave you an extra $10M, where would you put it?

Most marketing teams answer this question with gut instinct, historical momentum, or by trusting whatever their ad platforms report. The problem? Your ad platforms have every incentive to tell you to spend more with them. And historical allocation assumes that what worked at $127M will work at $150M, which is almost never true.

This is where media mix optimization transforms marketing mix modeling (MMM) from an interesting analytical exercise into the most powerful budget planning tool in your arsenal. The secret lies in understanding one elegant concept: diminishing return curves.

Media mix optimization is the process of determining the mathematically optimal allocation of your marketing budget across channels to maximize a target outcome—usually revenue, new customers, or profit; while adhering to real-world constraints.

It answers three questions that traditional attribution cannot:

1. Where is my next dollar going to deliver the most incremental return?
2. At what spend level does each channel start to plateau?
3. If I must reduce budget, which channel can afford to give up spend with minimal impact?

Marketing Mix Modeling tells you “what happened.” Media mix optimization tells you “what to do next.” This transition happens through diminishing return curves.

Most marketing dashboards report Average ROAS or Average ROI for each channel. This metric is dangerously misleading for budget allocation.

The Pizza Analogy

Imagine you are hungry:

• The first slice of pizza gives you immense satisfaction (high marginal utility)
• The fourth slice is okay, but you’re getting full (lower marginal utility)
• The eighth slice makes you sick (negative marginal utility)

If you calculated your “average satisfaction” after 8 slices, it might still look positive because the first few slices were so good. But the decision to eat the 8th slice was a bad one.

Marketing is the same. A channel might have a fantastic average ROI because the first $100k performed incredibly well, masking the fact that the last $50k you spent was wasted.

The Three Blind Spots of Average ROI

1. Over-investment in saturated channels: Channels like Google Search or retargeting often show high average ROI because they capture intent—but they saturate quickly
2. Underfunding upper-funnel channels: Brand-building channels like CTV or YouTube may have lower average ROI but massive untapped potential
3. The “justification wall”: You can’t justify budget increases with past averages—you need to show the future marginal return

The biggest mistake marketers make in media budget allocation is confusing two fundamentally different metrics:

Marginal response should be looked as “the ‘next dollar response’… the foundation of understanding how nonlinear budget allocation works.”

A Practical Example

Your blended ROI looks decent (3.67x overall). But notice the marginal ROI misalignment:

• The next $10k in social will return $40,000 (4.0x)
• The next $10k in search will return $25,000 (2.5x)

You should reallocate $10k from search to social. Even though search has higher average ROI, social has higher marginal ROI, meaning it’s currently underinvested relative to its potential.

A diminishing return curve (also called a response curve or saturation curve) is a mathematical function that models the non-linear relationship between your marketing spend and its resulting output.

The Three Phases of Marketing Response

1. Increasing Returns Phase: At low spend levels, each additional dollar delivers more than the previous one; you’re reaching highly receptive audiences
2. Linear Returns Phase: This is where spending increases produce proportional revenue gains
3. Diminishing Returns Phase: Saturation sets in. You’re hitting frequency caps, exhausting audiences, and facing auction pressure. Each additional dollar delivers less incremental value

Why Every Channel Saturates

This is true for every marketing channel:

• Search: Your first $10k buys high-intent keywords; your 100th thousand targets broader, less-qualified queries
• Social: Initial campaigns target best lookalikes; expansion reaches progressively colder prospects
• TV: First impressions reach engaged prime-time audiences; the 20th frequency impression hits diminishing attention
• Display: Initial retargeting converts ready buyers; scaled prospecting has lower conversion rates

Diminishing returns isn’t a bug; it’s a fundamental feature of all marketing channels.

How MMM Captures Diminishing Returns: The Two Critical Transformations

Modern MMM doesn’t assume a linear relationship between spend and sales. Instead, it uses two key transformations:

1. Adstock: Capturing Carryover Effects

Advertising effects don’t vanish immediately. A TV ad viewed today might influence a purchase next week. Adstock modeling captures this carryover effect.

Practical implication: Adstock prevents you from undervaluing channels whose impact unfolds over weeks, not hours.

2. Saturation: Modeling Diminishing Returns

The Hill function is a common mathematical form for saturation, borrowed from pharmacology:

​

Where:

• ec = the half-saturation point (spend level where you reach 50% of maximum response)
• slope = controls whether the curve is concave or S-shaped

Once you have response curves for each channel, optimization becomes a constrained calculus problem.

The Equimarginal Principle

In the simplest case (no other constraints), the optimal solution follows the equimarginal principle: invest more in the option with higher marginal return until marginal returns equalize.

In marketing terms:

Reallocate budget from channels with lower marginal ROI to channels with higher marginal ROI until marginal ROI converges (or you hit constraints).

Let’s walk through a case study to see media mix optimization in action.

The Setup

A DTC ecommerce brand spends $50M annually across five channels:

Total Revenue: $122.1M | Blended ROI: 2.44x

What the MMM Reveals

After fitting saturation curves to 2 years of weekly data:

Key Insight: Search and Display are over-saturated with marginal ROI well below average. TV and Podcast have strong marginal returns but are underfunded.

The Optimized Allocation

Projected Outcome:

• Same $50M budget
• Projected Revenue: $138.4M (up from $122.1M)
• New Blended ROI: 2.77x (up from 2.44x)
• Incremental Revenue Gain: $16.3M (+13.4%) with zero additional spend

Step 1: Build Your Channel-Specific Response Curves

Your MMM must generate a unique diminishing returns curve for each major channel. Ensure your model captures both adstock (lag) and saturation effects.

Step 2: Define Your Objective and Constraints

Before shifting budgets, clarify your goal:

• Maximize total revenue?
• Maximize profit (revenue minus media cost)?
• Acquire specific number of new customers at target CAC?
• Achieve minimum overall ROAS?

Define real-world constraints:

• Total budget ceiling
• Minimum/maximum channel caps (contractual obligations, brand safety)
• Operational limits (creative capacity, inventory availability)

Step 3: Run Scenario Simulations

Use your response curves to simulate thousands of potential budget allocations. Modern platforms enable “what-if” scenarios:

• Budget cut scenarios: “What if we only have $100M instead of $127M?”
• Budget increase scenarios: “We have an extra $10M—where should it go?”
• Channel blackout scenarios: “What if we pause TV for Q3?”

Step 4: Validate with Incrementality Testing

Critical step: Do not blindly trust the model. MMM can suffer from:

• Omitted variable bias
• Endogeneity (you increased spend because sales were up)
• Collinearity (channels move together)

Run incrementality tests (geo holdout tests, conversion lift studies) to check if MMM estimates match real-world causality. Meridian explicitly supports incorporating ROI priors from experiments.

Step 5: Implement, Monitor, and Iterate

Media mix optimization is not “set and forget”:

Build Your Business Case

Step 1: Identify the Problem

• Current measurement gaps or biases
• Budget misallocation risk
• Lost revenue due to inaccurate attribution

Example:

“We spend $10M on marketing across Google, Meta, TikTok, and TV. GA4 only sees Google touchpoints. We’re making budget decisions based on incomplete data, likely misallocating $1-2M annually.”

Step 2: Quantify the Opportunity

• Potential efficiency gains (10-20% is typical)
• Revenue impact of better budget allocation
• Cost of status quo (wasted spend, missed opportunities)

Example:

“If we improve marketing efficiency by 15%, that’s $1.5M in incremental revenue. The cost to achieve this is $500K/year. Payback is 4 months, with ongoing value generation.”

Step 3: Address Finance and Leadership

• Finance: ROI, payback period, NPV
• Leadership: Strategic advantage, growth impact, competitive differentiation

Example:

“This investment pays for itself in 4 months and generates $1.5M in incremental revenue annually. Competitors using advanced measurement are gaining market share. This is a competitive necessity.”

Step 4: Show the Risk of Inaction

• Cost of status quo (wasted spend each quarter)
• Privacy risk (GA4 becoming less reliable)
• Competitive risk (competitors using better measurement)

Example:

“Every quarter we delay costs us ~$375K in potential wasted spend. Privacy changes are making GA4 less reliable. Competitors using advanced measurement are outpacing us. We need to act now.”

Do I have to choose between GA4 and Measured?
No. GA4 and Measured serve different purposes. GA4 is great for website analytics and tactical optimization. Measured complements it by providing strategic, causal measurement for marketing budget decisions. Use both.

Isn’t GA4 good enough for marketing measurement?
GA4 is good for website analytics and tactical optimization. But it falls short for strategic marketing measurement, budget optimization, and omnichannel attribution. If you’re making strategic budget decisions, GA4 alone is insufficient.

Why is Measured expensive if GA4 is free?
GA4 is website analytics; Measured is specialized marketing measurement. You’re paying for incrementality testing, Media Mix Modeling, budget optimization, and expert support, all capabilities GA4 doesn’t have. The ROI (15-20% efficiency gain) justifies the cost.

Can I use GA4’s attribution models instead of Measured?
GA4’s attribution models are better than last-click, but they’re still correlative and biased toward Google channels. Measured’s causal approach (incrementality testing) is fundamentally superior.

How long does it take to see ROI from Measured?
Payback is typically 4-6 months. After that, you’re generating $1-1.5M in annual incremental revenue.

Do I need both incrementality testing and MMM, or just one?
Both. Incrementality testing proves causal impact (ground truth). MMM provides strategic, omnichannel insights. Together, they create a complete measurement system. GA4 has neither.

What if our company is small or early-stage?
If you’re spending less than $100K/month on marketing, GA4 may be sufficient. Once you’re spending $500K+/month or running omnichannel campaigns, investing in Measured becomes ROI-positive.

How does Measured handle privacy better than GA4?
Measured uses aggregated, non-user-level data and doesn’t rely on cookies. GA4 relies on user-level tracking, which is increasingly restricted by privacy laws and consumer privacy settings.

Pitfall 1: Optimizing on Extrapolation

Response curves are most reliable where you have data density. Be careful of curves that may require extrapolation above/below historical ranges, and “user discretion is required.”

Mitigation: Impose spend bounds close to historical ranges; validate large reallocations with experiments before locking in.

Pitfall 2: Treating Optimizer Output as Single Truth

MMM is estimated; curves have uncertainty; therefore the “optimal mix” has uncertainty too. Google Research notes that the optimal media mix can have large variance.

Mitigation: Plan with intervals (best-case/base/worst-case); use risk-adjusted objectives.

Pitfall 3: Over-Optimizing for Short Term

Focusing solely on immediate revenue can starve long-term brand equity.

Mitigation: Incorporate brand lift studies and long-term value metrics; run longer attribution windows (8-12 weeks).

Pitfall 4: Data Quality Issues

Garbage in, garbage out. Inconsistent spend data or delayed revenue data produces unreliable curves.

Mitigation: Audit your data pipeline; ensure spend is reported consistently (net media cost); flag outliers.

Platforms like Measured combine MMM with incrementality testing for validation, plus scenario planning for budget optimization.

Pros: Faster setup, managed infrastructure, ongoing support Cons: Less model transparency, higher cost

One of the biggest criticisms of MMM: “How do I know it’s not just fancy curve-fitting?”

This is where incrementality testing becomes essential. It measures causal lift through controlled experiments.

How Incrementality Validates MMM

Scenario: Your MMM says Facebook has a 3.2x marginal ROI at current spend.

Validation Test:

1. Run a scale test: Randomly assign DMAs to “test” (continue ads) and “control” (pause ads)
2. Measure the sales difference between test and control
3. Calculate true incremental ROI

Outcomes:

• Test confirms 3.2x → Model is well-calibrated
• Test shows 1.8x → Model overestimates (recalibrate saturation curve)
• Test shows 4.5x → Model underestimates (increase recommended spend)

This closed-loop validation separates causal marketing from traditional analytics.

What is the difference between MMM and media mix optimization?

MMM is the statistical analysis that measures historical performance—the coefficients. Media mix optimization is the application of that analysis to predict future outcomes and recommend specific budget changes.

How often should you perform media mix optimization?

Best practice is quarterly for strategic planning, monthly for tactical adjustments, and weekly monitoring for real-time signals. Modern “Continuous MMM” enables more frequent optimization.

Why is marginal ROI better than ROAS for budgeting?

ROAS is an average metric across all historical spend. Marginal ROI measures the efficiency of additional spend. Using ROAS can lead to overspending in saturated channels, while marginal ROI identifies the point of diminishing returns.

What’s the difference between Average ROI and Marginal ROI?

Average ROI shows how past spend performed—it’s great for reporting. Marginal ROI shows how the next dollar will perform—it’s essential for budget allocation.

Can I use media mix optimization if I only have 12 months of data?

It’s possible but risky. With limited data, you may not have enough variation to reliably estimate saturation curves. Consider supplementing with incrementality tests to validate and calibrate the model.

Does media mix optimization work for brand awareness channels?

Yes, but the “return” metric must be defined correctly. If you optimize purely for immediate sales, the model may defund brand channels. Advanced models account for base sales and long-term brand equity effects.

Diminishing returns curves represent the crucial bridge between Marketing Mix Modeling analysis and media budget allocation decisions. By moving beyond average ROAS to marginal ROI, marketers transform from historians documenting past performance to strategists shaping future outcomes.

The most successful marketing organizations don’t just build MMM models—they build optimization engines that continuously balance spending across channels to maximize marginal returns. They understand that media mix optimization isn’t a quarterly planning exercise but a daily discipline of measurement, testing, and adjustment.

The law of diminishing returns sounds like bad news; after all, it means channels get less effective as you scale. But here’s the counterintuitive truth:

Diminishing returns are the reason media mix optimization works.

If channels scaled linearly forever, there would be no optimization problem; you’d put all money in the highest-ROI channel. But because every channel saturates, the optimal strategy is to spread budget across channels until marginal ROI equalizes.

That curve isn’t a limitation; it’s a map. And marketing mix modeling gives you the compass to navigate it.

Ready to transform your MMM from historical analysis to forward-looking optimization? Measured combines marketing mix modeling with geo-based incrementality testing to deliver budget recommendations you can actually trust.