Marketing mix modelling (MMM) is here to stay for good. With privacy changes occurring globally, and third-party cookies on browsers, which were used to measure conversion, being phased out, businesses are turning to MMM for privacy-safe ways to measure ROI. A tried-and-true method, it assesses the impact of marketing without relying on personal information.
However, not all MMMs are equal. Using the wrong model for your business can lead to biassed decisions that undermine your ROI.
In fact, new research by Hakuhodo DY Group and dentsu respectively, in collaboration with Google, finds that the use of MMM which doesn’t align with one’s business operations can cause as high as a 10X margin of error on calculations of customer conversion due to media investment.
For example, one study assumed that only brand advertising and performance campaigns affected the results of an online-offline business. Based on this, simulated data was generated and a simple marketing mix model, which disregarded any synergy between brand advertising and search marketing, was applied. As a result, the estimated brand advertising effectiveness was between 3X and 10X lower than the true value.
To help you identify a model that’ll let you make accurate marketing investment decisions and grow your bottom line, we pull back the curtain on MMM here to help you get the most out of it.
Things to get right about MMM
For a topic that’s trending in the industry, getting into the “what” of MMM can seem like an unnecessary introduction. But once you know how it works, you can avoid being misguided and use it well. So here are three things to get right about MMM.
What sets MMM apart: It’s a statistical method that analyses data from many sources to provide a comprehensive view of how your marketing mix and activities affect sales. It considers things ranging from the products promoted and their distribution setup to the marketing channels used, online and offline, and the results, both immediate and in the longer term.
What makes MMM tick: It’s rigorous and so requires a wide range of data, including marketing measurements such as ad placement volume and costs, and performance data like the number of leads generated and the products and services sold. It also considers external factors like competitive and macroeconomic trends and seasonal fluctuations.
What benefits MMM offers: It shows how your various marketing efforts and media channels contribute to your marketing goals. This lets you identify which activities and channels are most effective at driving your goals such as sales. It also empowers you to make informed decisions, especially in times of macroeconomic uncertainty, about the best way to allocate your marketing budgets and investments.
What people get wrong about MMM
It is a model, not a fixed formula. This means the type of statistical solutions used to create a model, and the assumptions that underpin those choices will affect one’s analysis, investment decisions, and returns.
In other words, a model can be biassed if it doesn’t mirror all your marketing activities and how your business actually operates — from the myriad sales channels it uses to its advertising structure, and the synergy between its various media investments. Our research with Hakuhodo DY Group and dentsu, respectively, validated this when we ran simulations and compared the values from the models with the actual business results.
Your MMM is biassed if it doesn’t mirror all your marketing activities and how your business actually operates.
Additionally, people may be mistaken about the capability of MMM. While it is a powerful tool, it is not a one-size-fits-all measurement solution. It should be part of your full stack of measurement solutions and used together with granular attribution solutions like Google Analytics 4 to evaluate your marketing investment accurately.
How to find your MMM match
To get your hands on a model that matches your business reality and lets you accurately measure ROI to make the right investment decisions, here are four things you can do.
Define the purpose and premise of your analysis. This includes who’s using it and what questions they want answered. With this, you can pinpoint the data that needs to be collected, how it’s to be cleaned, and importantly, what kind of a model needs to be built.
Determine your customer’s purchase behaviour. This helps you ensure your model precisely reflects your marketing and business activities, from the touchpoints where shoppers make contact with you, to what they’re looking for, and how and where they shop, whether online and/or offline.
Tap the right pool of experts. Their expertise and diversity of views will help ensure your model is built to purpose and robust. Look for people both inside and outside your company who have a breadth of knowledge, not only in the field of machine learning, but also in statistics, econometrics, and causal inference. Have them evaluate the structure of your model and help you identify any potential blindspots.
Thoroughly evaluate the build of the model. Assess the model you’re using, whether it’s from a provider or built in-house, because even the smallest, unintended bias can lead to significant errors in measurement. Check the assumptions that underpin your model, and see if it is compatible with your business objectives and marketing activities.
To help you in your evaluation, consider asking questions like:
If you need help with the evaluation, you can always tap external experts who are familiar with MMM. Google is working with advertising companies and industry stakeholders to develop MMM data that is granular and accurate, and to come up with methodology innovations that contribute to MMM best practices. We also provide bespoke MMM-based consultations to help businesses make better decisions about their marketing investments.
MMM is a powerful tool. And when used correctly, it can help you measure, analyse, and improve your marketing performance to achieve your business goals.
Contributor: Masaki Yoshida, Video Solutions Expert
