Editorial
The Gist
- Marketing measurement gets an upgrade. Google Meridian introduces a programmatic, open-source approach to marketing mix modeling for deeper performance insights.
- Customization meets control. Marketers can tailor budget analysis and channel evaluation through a single Python-based modeling framework with advanced visual outputs.
- Baselines still dominate. Despite advanced modeling, most revenue continues to come from baseline performance, not marketing campaigns — highlighting the need for strategic investment.
Necessity is often called the mother of invention. Necessity has never been more called upon than within the digital marketing space.
As data has been increasingly needed to manage customer experiences, marketers have needed better measurement that spans a variety of data sources and workflow. Analysts have long worked with the three Ms — marketing mix model — while facing attribution measurement changes. Until now, most data analytics dashboards have had limits when providing customizable options.
Marketers now have a new solution to these challenges: Google Meridian. This open-source analysis tool offers a programmatic solution for crafting marketing mix model dashboards. The advantage is that marketers gain budget guidance, improving budget planning decisions and overall customer experiences with media.
Table of Contents
- The Basics of Marketing Mix Models
- Where to Set Up Your Marketing Mix Model Analysis
- Meridian’s Place in the Martech world
The Basics of Marketing Mix Models
Marketing mix models (MMMs) are a statistical method for estimating the impact of marketing tactics on sales. They can be used to optimize advertising campaigns and promotional tactics. The term “marketing mix model” dates back to 1949, but MMMs have become increasingly popular with the rise of digital marketing.
How Marketing Mix Models Work
Marketing mix models are crafted in the same way as other statistical data models. The steps resemble those for a regression. However, the scope of the data analyzed is the biggest difference between the two. Regressions are a general statistical modeling technique that uses data to estimate the scale of different variables that contribute to an output. They apply to many domains, with some consideration to decisions with assumptions, such as whether to establish a linear or logistical relationship among the variables.
In contrast, marketing mix models assign the dependent variables to either sales metrics or a percentage of the market share. They incorporate business constraints and realities, such as budget limitations, channel capacity constraints, minimum effective spending levels, and time constraints. Marketing mix models are specifically designed to incorporate campaign conditions, focusing on how different marketing channels and activities drive business outcomes.
Marketing Mix vs. Attribution
The marketing mix model also differs from attribution. Many times marketers think of attribution when they turn to market mix modeling. The model naturally encourages discussions about attribution when campaign data sources are mentioned.
However, whereas attribution is a nuanced look at a channel, MMM is a high-level view of campaign activity and response on a channel. For example, for attribution, you are examining which sites to segments are worthwhile for general investment.
Related Article: How to Get Attribution in Analytics Right
Where to Set Up Your Marketing Mix Model Analysis
Getting Started With Meridian
To access Meridian, you have to set up a Google Colab account. Google Colab is an online notebook platform used by data analysts to create data models and reports. Think of it as an advanced cloud version of a word processor that hosts programming code alongside the text and you have the general idea of its value.
Google offers Meridian as a Python module. This arrangement also means a bit of developer chops is needed to set up Meridian, but the steps are not extensive. The model must be built, then you must assess the model output and debug as needed.
Feeding Data Into the Model
Analysts upload their media data using standard Python techniques, ranging from reading a file to importing data through an API. Once the data is added, you begin to create your marketing mix model.
Meridian allows two choices of model variables to represent marketing media – media data and media spend. Media data is data that represents message exposure. The data must display an exposure metric per channel within a period, such as impressions per day. The Media spend is monetary data containing the media spending per channel and period. The media data and media spend must have the same dimensions, with no negative values in the data. The model uses media spend only if media data is not immediately available.
The next step is to map the columns to their data type so that Meridian interprets them when modeling is applied. Meridian can perform hierarchical modeling, so mapping columns to data type and parameters that identify media and calculate metrics.
Medium then creates a model, applying a Bayesian and Markov Chain algorithm to the data and then configuring the model parameters.
Meridian’s hierarchical modeling allows marketers to select data segments within one single model. This is good for maintaining one analysis model while having the ability to quickly craft a custom analysis for each segment.
Let’s say you have a regional retailer who needs channel activity as it relates to sales in a few states, both individually and region-wise. You can create an unpooled model – one that displays a model for each state – and a pooled model – a model that “averages” all regions. In this approach, you have a prediction model dedicated to channel changes related to a region. The model “learn” using data brought together while keeping the data and parameters for a model for a specific state separately.
From Outputs to Insights
When marketers are ready to share the model findings, Meridian offers a way to create a report based upon the model output. A Python module called summarizier creates a class for your model. The output can then be placed in reports issued in the HTML format. The reports contain model performance and other visuals to help you understand what drove your revenue.
The channel contributions report reveals what drove your revenue. In this sample image, Channel_4 and Channel_3 drove the most overall incremental revenue above the baseline and organic channel activity. This would imply that those two channels were effective to increase customer interest to produce sales and that their performances are worth examining for further sales growth opportunities.
Additional reports include response curves that depict the relationship between marketing spend and the resulting incremental revenue. These curves reveal the optimal spend for each channel that maximizes the total incremental revenue. There are even optimization scenarios based on your data available, so you can see what changes ion sales lift could potentially occur.
Once you are done with the analysis, you can save the model for future use, avoiding repetitive model runs and saving time and computational resources. After the model object is saved, you can load it at a later stage to continue the analysis or visualizations without having to re-run the model.
Steps to Set Up and Use Google Meridian
Like outlined in the graphic above, this table outlines the key phases marketers and analysts follow when implementing Google Meridian for marketing mix modeling and analysis.
| Step | Action | Purpose |
|---|---|---|
| 1. Set up Google Colab | Create a Google Colab account to access the cloud-based notebook interface. | Provides an environment to run Python code and build the model. |
| 2. Install Meridian module | Install Google Meridian as a Python package within your Colab notebook. | Enables access to Meridian’s modeling functions and configuration tools. |
| 3. Load and prepare data | Import media spend or media exposure data using standard Python techniques (file upload, API call). | Provides the foundational dataset for the marketing mix model. |
| 4. Map variables | Identify and tag columns by data type, including channels, spend, exposure, etc. | Ensures correct interpretation by Meridian for hierarchical modeling. |
| 5. Build the model | Use Meridian’s Bayesian and Markov Chain algorithm to generate your model. | Establishes relationships between spend and outcomes across marketing channels. |
| 6. Segment and customize | Apply pooled or unpooled modeling based on region, product, or audience. | Allows side-by-side comparison and personalization of insights per segment. |
| 7. Generate output and reports | Use the summarizer module to create HTML-format reports with channel contributions and performance visuals. | Visualizes ROI, lift curves, and optimal spend recommendations for stakeholders. |
Meridian’s Place in the Martech world
Traditional digital attribution methods are becoming less reliable as customer engagement and campaign messaging occur on various platforms. Many platforms have introduced challenges to measuring user activity across different touchpoints, making it harder to understand the customer journey. Meridian offers a new way to measure the impact of marketing efforts.
Meridian’s design also reflects a growing BI trend – Analytics as Code. Analytics as Code is a workflow in which analysts insert programming logic to craft advanced analysis from analytic solutions. The benefit is offering analysts more options to manage cardinality – the number of features within a dataset that represents real world activity.
Having either too few or too many features can skew model results, indicating imaginary trends and causing erroneous conclusions. The use of machine learning, punctuated by AI adoption, has transformed how analysts must plan their analysis. Solutions like Meridian can provide the right support for campaign strategy based on that analysis.
Webinar
Sep
25
CMS Briefing: A Live Look at What’s Next in AI-Driven Platforms
Learn how leading organizations are using AI‑driven tools to publish faster, personalize smarter and stay secure.
Register
Webinar
Sep
25
Call Spoofing Trends Financial Institutions Can't Afford to Ignore
Don't let robocalls sabotage your customer relationships. Learn how to secure your voice channel.
Register
Webinar
On demand
Ready or Not: How Data-First Organizations Are Unlocking Agentforce Potential
Learn how to cut through the noise, activate Agentforce and build a Salesforce AI strategy that actually delivers.
Watch Now
Webinar
On demand
CX Reality Check: What Your Customers Are Really Thinking
Register now and get the insights you need to not just meet rising expectations, but exceed them.
Watch Now
Webinar
On demand
Content Leaders Collective: What a Good CCMS Actually Looks Like
Discover how content leaders transform clunky systems into growth engines — and why your competitors are ahead.
Watch Now
Webinar
On demand
From Hype to High-Impact CX Strategies That Actually Scale
Turn buzzworthy AI and outsourcing trends into measurable CX wins with fresh 2025 data.
Watch Now
Marketers seeking updates to their measurement stack will find Google Meridian to be a suitable middle ground between custom dashboards that require a bit of syntax to create and ease in managing sales lift from planned marketing channels.
Learn how you can join our contributor community.
