How to use MMM to increase your Marketing ROI

Why I'm Writing This Post

I'm Gabriele Franco, the founder of Hybrida Marketing, a growth marketing agency that last year alone managed 12 million dollars of advertising spend for our clients.

In order to optimize our work, we developed an internal technology that is now called Cassandra.

Marketing Is Changing

I've been working in marketing since 2016 and in April 2021 we faced an enormous change in our field.

We faced a new "restriction on cookie policy" that ruined the accuracy of almost all the tracking systems we were using.

Why Is That?

Let me explain what happened.

Google Analytics, Google Ads, Facebook Ads, TikTok Ads, and Bing all use a measurement system called "Multitouch Attribution." This system uses third-party cookies to measure user behavior on our site and discover from which source or channel each user came from.

This system has been incredibly useful for almost 10 years, but recently it's facing big inaccuracies. Tracking systems fail to attribute correctly up to 30% of transactions due to the restriction of the cookie policy of April 2021.

This leads the advertising platforms to raise marginal costs (CPO, CPL, CPA) and reduce the average contribution margin. All marketers making decisions based on data now feel stuck with inaccurate data that doesn't describe reality.

Introducing a Solution: Marketing Mix Modeling

In April 2021, all our clients faced a stop sign for growth because we could not continue making decisions on inaccurate data. We started searching for a measurement solution that did not use cookie technology.

We found out about Marketing Mix Modeling (MMM).

"MMM are statistical techniques applied on historical data to understand the marginal contribution that each single factor has with respect to our sales."

MMM are basically tools that help you study your investment efficiency and discover new opportunities.

How MMM Can Be Applied in Your Organization

Studying this market, we tested all the possible applications of MMM and selected the 3 most important ones.

1. Solving the Attribution Problem

Marketing budget planning is a complex activity based on the economic returns of each past investment. We are now facing big problems in attributing digital investments — and we never had a method to measure offline investment attributions.

MMMs are able to discover the contribution of all media investments, including offline ones. You can see through a dashboard how many conversions each channel has contributed:

• The spend share: the % of how much we spent on each channel

• The transaction share: the % of transactions attributed to that channel

• The ROI estimation on that investment

2. How Can I Predict Next Month's Sales?

What happens to my sales if I double down on my investments this month? What if I lower them by 35%?

If you are in marketing, you have asked yourself this question at least once. The answer is more complex than you think — you should consider tens of factors that influence your sales, process them all simultaneously, and produce a prediction. It is simply impossible for a human being to do this manually.

Marketing Mix Modeling helps you do that by considering the following factors:

• Competitors: Promotions, discounts, advertising investments

• Offline Media: Investments in PR, influencers, TV, OOH, Radio

• Digital Media: Investments in Meta Ads, Google Ads, TikTok Ads, Outbrain

• Price Variation and Promotion: Measure price elasticity of demand

• Seasonality: Discover how much the demand for your product changes each month

• External Factors: Weather, inflation, oil prices, and other factors that influence consumer behavior

3. How Can I Maximize My Multichannel Marketing ROI?

Much like a financial portfolio, your returns are based on how you allocate and distribute your investments. When you have 6 possible channels and more than 15 campaign types, how do you decide what's the right amount of budget for every campaign?

Marketing Mix Modeling, instead of using best practices, learns from all your historical data what the diminishing returns of each channel are. It then uses this information to tell you how to distribute your budget to maximize your ROI.

For example: the model might suggest increasing TV spend, search spend, and print spend while lowering OOH spend and Facebook spend. This new distribution, according to Cassandra, would lead to an increase of 8.85% in revenue — and the model is also able to predict exactly how much incremental revenue you would get by following this course of action.

Are MMMs Expensive?

If you already know about MMM, you know that they are usually really expensive. They are promoted by big consulting firms like McKinsey, EY, and Nielsen. To give you an idea, the average listing price for this service is:

• $50,000

• 3–6 months for delivery

• Output: one model and one suggestion

That's incredibly expensive for a mid-size company to invest in just to receive one suggestion that can realistically be applied only once.

That's why we've developed Cassandra: a solution for all companies that invest more than $50,000/month in advertising to have a customized Marketing Mix Model that is constantly updated on new data — at a fraction of that price.

Case Study: +13% Transactions with MMM

A new client sold shipping services for e-commerce businesses. Their monthly advertising budget was approximately €52,000/month, and they had 3 years of historical data. Our goal: generate growth through data-driven budget allocation.

We tried using Google Analytics attribution but couldn't learn much — most of the transactions came from the "Direct" channel. There was a clear attribution problem.

We developed a Marketing Mix Model that suggested lowering the budget on Facebook Ads and allocating it to Google Search and Google Video.

According to the model, this action would generate 5.5% incremental sales compared to the previous period. We executed it for two weeks and noticed a 5.75% increase in transactions.

We then refreshed the model on the new data and received a new allocation suggestion: lower Facebook Ads further and reallocate to Google Discovery. The prediction was 6.6% incremental sales — after execution, we generated over 7.6%.

Summary of Results

In one month and one week, by modifying budget distribution only:

• +13.36% in the number of transactions

• +42.21% in profit

• +22% in ROAS

• -7% CAC (thanks to a geo-lift experiment)

How to Implement This in Your Organization

In Cassandra, we have developed a solution composed of 3 steps:

Step 1: In 2 weeks you receive a personalized MMM for your company with a money-back guarantee if you don't like the outcome.

Step 2: In 2 weeks we build a data infrastructure to centralize all your data daily and refresh the model every two weeks.

Step 3: 2 hours of monthly consultancy to transform model insights into experiments that will drive better ROI over time.

Try Cassandra for free or book a demo to see if and how we can help you.