How Much Should You Invest in Marketing (And to Know If It’s Working)?

When Laura, CMO of a major FMCG brand, sat down with the board to defend her budget, she was clear:

“We need more media investment if we want to grow at the pace the business demands.”

But then came the question.

The usual one.

“How do you know that ‘spend’ is working?”

Silence.

A few Google Analytics charts.

A reach report.

And a last-click attribution table.

Nothing truly convincing.

Every year, marketing teams face the same uncomfortable question:

What’s the ideal marketing budget?

Spending too little can stall growth. Spending too much without returns burns resources.

And even harder: how do you really know if that investment is working?

For years, we’ve relied on platform metrics, dashboards, last-click attribution—or simple gut feeling. But in a context of data fragmentation, the disappearance of cookies, and increasing pressure to justify every cent, those tools are no longer enough.

The good news? Today, there are models that do let us answer this with precision. One of them is making a strong comeback: Marketing Mix Modeling (MMM).

The Budget Dilemma: Intuition vs. Data.

Many marketing budgets are still built on inertia: last year’s plan copied over, slightly adjusted per campaign, or split across channels based on visibility or cost-per-click. Sometimes even based on internal politics.

But the real return on that investment—the actual impact on sales, leads, or business—rarely gets measured with certainty.

And without that certainty, everything turns into a negotiation: between departments, between agencies, between expectations.

What used to pass as a “reasonable estimate” doesn’t fly anymore. Marketing teams now need to prove, with data, that what they’re doing makes sense. That they’re investing in what works. And that they can demonstrate it.

As Harvard Business Review noted in its article  “Why CMOs Never Last”, marketing leaders are under growing pressure to show tangible results and link investment to business growth. Without solid data, trust erodes. In 2024, CMS Wire reported that marketing budgets had dropped to just 7.7% of revenue—the lowest since the pandemic—and that 73% of CMOs feel they must do more with less. The pressure is clear: justify every euro.

Moreover, as Think with Google noted in January 2025’s “How to unlock the hidden 50% of your marketing ROI,” up to half of your marketing ROI might be invisible if you only measure short-term performance. To uncover that value, you need more comprehensive models—like MMM.

In this new reality, Marketing Mix Modeling is re-emerging as a tool that brings real evidence of what works—and why.

Marketing Mix Modeling, Made Simple.

El Marketing Mix Modeling (MMM) es un modelo estadístico que analiza el Marketing Mix Modeling (MMM) is a statistical model that analyzes the individual impact of each channel on business results: sales, leads, awareness, or other key KPIs.

Unlike digital attribution models, MMM isn’t limited to clicks or user journeys. It also includes traditional media like TV, radio, or in-store promotions, along with external factors such as pricing, competition, or seasonality.

Its biggest strength: it can simulate future scenarios.
What happens if I double my investment in social media? What if I cut search spend entirely? How would a promotion affect each channel’s performance?

MMM doesn’t just answer what worked. It also predicts what could work better.

The difference from other models is clear:

• Doesn’t rely on cookies or user-level tracking
• Works for both digital and traditional media
• Includes external variables like seasonality, price, or competitors
• Can simulate “what-if” scenarios

How Does an MMM Model Work?

Although it may sound complex, a Marketing Mix model follows a simple logic based on analyzing multiple variables. Here are the key steps:

1. Data Collection. Gather historical data on media (digital and traditional), business outcomes (sales, leads, brand awareness), and external factors like price, promotions, seasonality, or competitor activity.
2. Data Preparation & Cleaning. Standardize and clean the data to ensure consistency. This includes removing biases, unifying formats, and correcting inconsistencies.
3. Statistical Modeling. Apply multivariable regression models to identify the impact of each channel on the results. At this stage, you find patterns and determine which channels are generating the most ROI.
4. Model Validation. Test the model with new data to verify predictive accuracy and avoid overfitting. Aim for robust models, not just the ones that best match past data.
5. Scenario Simulation. Once calibrated, the model can answer questions like: What if I cut paid media by 20%? What if I invest more in retail media? Where’s the point of diminishing returns for a channel?
6. Visualization & Activation. Turn results into clear, actionable dashboards to help business and marketing teams make fast decisions.

What used to be limited to large corporations and consulting firms can now be implemented using open-source tools like Robyn (Meta) or Meridian (Google), making MMM more agile and accessible.

Robyn & Meridian: The Engines Behind the Shift.

MMM’s resurgence hasn’t been accidental. Meta and Google have both released open-source models to support adoption across companies of all sizes.

Robyn (Meta).

An open-source tool developed by Meta that automates many complex MMM tasks, including:

• Automatically testing dozens of models
• Measuring delayed and saturation effects by channel
• Powerful visualizations and comparative reports

Robyn is ideal for companies with analytical maturity but is flexible enough to adapt to various needs. You can access the official documentation here.

Meridian (Google).

Launched in January 2025, Meridian is Google’s answer to the need for more accessible models.

• Lighter and easier to implement
• Designed for teams with smaller data sets and no advanced infrastructure
• Aims to democratize MMM with a more user-friendly approach

Documentation is available on the Google Developers site.

Why It Matters More Than Ever.

Because the context has changed.

For years, marketers worked with accurate data, third-party cookies, and attribution tools that—while limited—let you follow users through digital journeys.

But that’s ending.

Cookies have an expiration date. Chromium-based browsers are now blocking third-party cookies by default. This shift toward privacy-first environments has made user tracking fragmented and unreliable. Meanwhile, the pressure to justify spend has only increased.

In this new scenario, MMM is back in the spotlight. Not only for its statistical rigor, but because it doesn’t rely on personal identifiers—and it lets us measure the impact of channels we previously could only guess about.

Meta released Robyn. Google launched Meridian. The major platforms are backing MMM’s comeback.

And this time, they’re doing it with accessible tools for everyone.

The message is clear: If you want to make confident decisions, you need models that see the whole picture—not just the last click.

Discover Good Rebels Mix: A Solution for Predicting Sales Impact.

Back to Laura. When she returned to the board to present her marketing plan, she wasn’t carrying just a couple of dashboards or reach reports.

She had a model.
Scenarios.
Clear answers.

At Good Rebels, we’ve developed Good Rebels Mix, a solution based on Robyn (Meta) and Meridian (Google) designed to help marketing teams make clearer decisions, rooted in data—without added complexity.

With Good Rebels Mix, you can:

• Measure the real ROI of every channel, including traditional media
• Identify when an investment becomes inefficient
• Simulate future scenarios and plan ahead
• Visualize results in clear, business-focused reports

Good Rebels Mix helps CMOs and marketers move from intuition to insight—without friction. Knowing how much to invest in marketing (and whether it’s working) is no longer a mystery.

With the right models, you can make decisions with confidence, justify every euro, and anticipate your campaigns’ impact.

Because the new language of marketing isn’t just creative.
It’s statistical.
Predictive. Actionable.

If you want to start making data-driven decisions—not intuition-based ones—let’s talk.

FAQ About MMM.

How much data do I need to apply MMM?
It depends on the model, but generally at least 1–2 years of weekly data per channel and business variable is recommended. Huge data volumes aren’t essential, but consistency is.

How long until I see results?
Between 3–6 weeks from data collection to a validated model with simulated scenarios. Implementation is faster if your data is already organized.

Do I need a technical team to use it?
No. With tools like Good Rebels Mix (based on Robyn and Meridian), the approach is low-code and designed for marketers—not just analysts or engineers.

Does this replace digital attribution?
Not necessarily—it complements it. Attribution tracks the user journey; MMM measures the aggregate impact of each channel. They’re different but compatible approaches.

Does it work if I don’t run offline campaigns?
Yes. While MMM is especially useful for including traditional media, it also works for digital-only channels and online variables. In fact, many companies use it this way.