In today's dynamic marketing landscape, it’s paramount that we understand which investments are truly driving results. Marketing Mix Modeling (MMM) empowers us to measure each marketing channel’s return on investment (ROI) and behavior, enabling data-driven decisions that maximize impact.
MMM allows us to attribute our KPIs (revenue, conversions, etc) using only higher-level date and channel aggregated cost data and without requiring log-level or individual user tracking data. This is especially beneficial in the world of cookie deprecation, user tracking limitations, query parameter removal and other privacy restrictions.
How an MMM works: A Simplified Example
To demonstrate how a basic MMM may work, let’s review the following example:
Suppose we have the marketing data below, and our goal is to estimate the ROI of each of the three channels (Paid Search, Radio and Billboard).
| Month | Paid Search Cost | Radio Cost | Billboard Cost | Total Revenue |
| January | $30 | $15 | $5 | $163.50 |
| February | $40 | $25 | $8 | $208.70 |
| March | $50 | $20 | $12 | $238.80 |
| April | $20 | $30 | $7 | $152.80 |
| May | $50 | $0 | $0 | $200 |
| June | $0 | $0 | $0 | $50 |
| July | $45 | $5 | 20 | $219 |
We quickly notice that without spending any money in June, we generated a baseline of $50. We can now naively attribute any additional revenue to our marketing spend. In May, for example, the $50 invested in Paid Search generated $150 above the baseline or an ROI of 200%. After analyzing all the data, we can assign an ROI to each channel with some level of confidence.
Going from initial total revenue:
To revenue grouped by each channel:
Overview and Building on the Basics
As you can see, even a basic MMM implementation can be very valuable. It provides the effectiveness of each channel without stitching individual user journeys.
Advanced MMM solutions, like Google Meridian, provide a holistic view of marketing performance. They consider factors like geographic impact, the delayed effect of campaigns and diminishing returns to give us a clear picture of what's working and where to invest. Additionally, when we add control variables and prior business knowledge to the mix, we can positively impact the accuracy of MMM. Meridian, in particular, also allows us to take advantage of additional channel metrics that directly impact the KPI, such as reach and frequency.
Budget Optimization
With the model trained, we can now utilize MMM to optimize our marketing spend. On the most basic level, that would mean investing more in channels with higher ROI and limiting the amount we spend on channels with lower or even negative ROI. The problem, however, is far more complex than that.
Modern MMM solutions consider saturation (diminishing results) and lagging effects (adstock) when providing these suggestions. The marginal ROI of seeing a single billboard on a road section compared to none is likely much bigger than seeing two instead of one. Investing in a Super Bowl ad today will likely have long-lasting effects. These solutions take the bigger picture into account and allow you to optimize based on a fixed budget, target ROI or minimal marginal ROI.
Real-life MMM Data Preparation
On the one hand, the data structure for an MMM is generally pretty straightforward; we’re looking at the aggregated spend by channel along with the total value of the KPI for the observed business divided into equal-sized periods (day, week, month). On the other hand, we are potentially working with channels where we cannot easily access cost data (or automate access), we may not have data for expected periods, etc.
As we work on our model’s complexity, we need to consider control variables and population scaling, and have a solid understanding of priors, correlations and other factors outside our base (channel, cost, KPI, date) data.
That’s where an experienced team like ours can help you extract the right data and implement your solution properly.
Why Meridian?
Meridian is an MMM solution built by Google. Its open-source nature provides transparency and accessibility, empowering businesses of all sizes to leverage its capabilities. It also does not require any type of SaaS subscription and can run completely in your environment. Technically, it takes a Bayesian approach that allows business stakeholders to incorporate existing knowledge into the model. Meridian also comes with geo-based hierarchical modeling, incorporates reach and frequency, and provides budget optimization functionalities and access to GQV, paid search, frequency and other Google-related data for organizations.
As a certified Meridian partner, our team has extensive knowledge of how Meridian works and provides comprehensive services to help you effectively implement and utilize the tool to achieve the highest ROI. Want to learn more? Please contact us.