Meridian MMM refers to Google’s newly-introduced open-source media mix modeling solution, aimed at enabling businesses to better analyze and improve the effectiveness of their marketing activities using ML-powered statistical analysis.
What is MMM?
In brief, Marketing Mix Modeling, or short: MMM, is a method to measure and analyze the effectiveness of businesses’ digital marketing activities using statistical modeling techniques.
The key objectives of using MMM include measurement of the true effectiveness of the company’s marketing and advertising activities (media channels, overall marketing spend), as reflected in ROAS, and the ability to better forecast and optimize/enhance business outcomes in future.
Building a quality model usually requires the input of clean and granular historical data regarding the company’s sales, marketing channels with relevant activities, along with any other available marketing and advertising performance metrics.
One of the pillars of using MMM tools lies in incrementality testing, which implies the evaluation of the real impact of a specific advertising campaign or any other marketing activity, i.e. measurement of the extra (incremental) value, generated due to this specific intervention, – by comparing at least one so-to-speak “control group” with the “treatment group.”
The incrementality test results can further be used to calibrate media mix models, mitigate revenue risks and optimize marketing spend, accordingly.
What is Meridian?
Designed as the substitute to the formerly introduced (and soon to be deprecated) LightweightMMM, Meridian empowers digital companies to extract valuable insights from their business data to better analyze the effectiveness of their marketing tactics and predict their future impact on essential revenue outcomes, hence helping them to make the more informed business decisions and allocate their marketing budgets more effectively.
Key Meridian MMM Features
Even though Meridian is still in its beta, hence not available for the general public testing, the already released documentation enables pinpointing some of its key distinctive features.
These, namely, include:
- use of the Bayesian inference;
- use of geo-level hierarchical regression model;
- ability to incorporate ROI priors, along with incrementality test insights and some industry benchmark data for model calibration;
- ability to add search volume data in the model to measure paid search advertising performance;
- ability to input frequency and reach data, as well as the so-called time-variable intercept (e.g. data on seasonal sales fluctuations, etc.), and more.
Potential Drawbacks
As the experts admit, some of the noticeable drawbacks of Meridian lie in its limitations.
First, these concern the supposed inability to include time-variable parameters for various media channels in the model (i.e. Instagram advertising campaigns are assumed to bring the same results over time, which is unreal), which can potentially distort the obtained results.
Second, similarly to other MMM solutions on the market, Meridian still lacks comprehensiveness of the implementation by marketing executives, who don’t have the sufficient level of expertise in data science and ML engineering. This, obviously, means that the technical adoption of this MMM will still require extra operational resources, hence financial investment from businesses, wishing to take advantage of its use.