Deciding the best way to spend your finite marketing budget should be one of the foremost concerns of any marketing professional. Unfortunately, it’s not always easy to know the optimal allocation for these funds. With dozens of potential channels for your content, from Internet ads to radio, how can you determine which of them will provide the largest number of customers or the greatest return on investment?
To address these issues of sourcing and attribution, marketers have come up with tactics such as marketing mix modeling (MMM), which is alternatively known as media mix modeling. Over the past several decades, major companies such as AT&T and Kraft Heinz have successfully implemented MMM within their business.
Using MMM and similar tools is indispensable for any marketer that wants to better understand their customer base. In this article, we’ll discuss everything you need to know about marketing mix modeling: what it is, why it’s important, how it works, and a comparison with other similar marketing techniques.
Marketing mix modeling (MMM) or media mix modeling is a statistical technique for measuring the impact of different elements and channels in a marketing campaign.
Specifically, MMM uses multiple linear regression, which attempts to find a mathematical relationship between a dependent variable and two or more independent variables. Multiple linear regression assumes that there is a linear relationship between these variables, and therefore that they can fit in a linear equation.
In MMM, the volume of sales is modeled as the dependent variable, while the independent variables represent certain elements of the marketing process.
Choosing the right independent variables to model in MMM is as much an art as it is a science. The goal of MMM is to understand how changes to a marketing campaign, as represented by changes to these independent variables, will increase or decrease sales.
The term “mix” in MMM refers to the combination of factors that are often referred to as the “four Ps.”
In general, the choices of independent variables in MMM are pulled from these four categories. For example, marketers using MMM may try to determine if a product has more sales when it is advertised online or via billboard ads.
Too many marketers allow themselves to be governed by whims and gut feelings about what has worked in the past to decide how to run their next campaign. While these ideas can be beneficial, they can also be wildly inaccurate and overlook many important factors.
Fortunately, marketing mix modeling provides the data that marketers need in order to back up their assumptions. With MMM, the objective is to mathematically model the cause-and-effect relationships between different factors in a marketing campaign. In turn, this allows marketers to calculate the return on investment for each of their channels to decide how much budget and effort to dedicate to them.
MMM also enables users to distinguish between incremental and baseline sales. “Baseline” sales are those that your company would have even without running any future marketing campaigns. These sales are usually a result of brand equity among loyal customers who have already been using your products. “Incremental” sales are those that you can trace back to marketing activities like TV advertisements and special offers.
A third benefit of MMM is that it can help shatter assumptions about why a marketing campaign succeeded or failed. In particular, MMM separates the drivers of a campaign’s outcome into both internal factors (such as advertising) and external factors (such as seasonality and macroeconomics). No matter how good a promotion you’re offering for swimsuits, for example, your campaign will be less successful if it runs during winter than during summer.
Beyond analyzing historical data, MMM can also help you perform “what-if analyses” to estimate what effect certain changes will have on a marketing campaign or your company’s profitability. If you’re thinking about taking a certain action, MMM will provide the data and insights you need to confirm or deny your idea.
Because it is a fundamentally statistical technique, MMM requires some degree of mathematical expertise. Data science environments like R and Python can handle MMM modeling. In organizations that are large enough to hire them, data scientists and IT professionals will collaborate with the marketing team in order to run MMM tests and simulations.
MMM models may fall prey to either underfitting or overfitting. With underfitting, the model is unable to find the general trend in the data; with overfitting, the model adheres too closely to the initial data it’s seen and cannot generalize well. In cases such as these, it can help to use data regularization techniques such as lasso regression and ridge regression.
As noted above, media mix modeling can help you answer questions such as:
In the right situation, MMM can be an irreplaceable tool in any marketer’s kit.
For example, the travel and hospitality industry is highly dependent on the season, as tourists want to visit a location without worrying about inclement weather (but also perhaps without running into any other tourists). When running a marketing campaign in winter, MMM can help travel and hospitality companies separate concerns about seasonality from concerns about the campaign itself.
While highly common among savvy marketing departments, MMM is not the only technique for understanding the countless variables at play within a marketing campaign. In this section, we’ll discuss one popular alternative: multi-touch attribution (MTA).
During a marketing campaign, there are multiple “touch points” between the customer gaining awareness of your product and finally converting. For example, the customer’s first touch point with the product may have been via a Facebook ad, followed by visiting your website and signing up to your mailing list. In the end, the customer makes a purchase after clicking on a link in your email that tells them about an ongoing sales event.
Traditional “first touch” and “last touch” models attempt to measure the effectiveness of different channels by determining where customers had their first or last contact with your brand. However, this model is overly simplistic, and it fails to account for the fact that customers may interact with your company multiple times before making their first purchase. Marketing content such as blog posts, white papers, and customer reviews can all be invaluable in moving the customer along the buyer’s journey, yet the “first touch” and “last touch” models would not capture their influence.
MTA is an alternative to MMM that seeks to track this full history of customer interactions, so that all channels involved in the conversion can be properly credited. This is different from MMM, which would typically give outsized credit to only a single one of these channels.
Different MTA models may assign different interactions different influences: for example, the linear model distributes credit equally among all channels, while the time decay model gives more credit to interactions that happened immediately before conversion.
In practice, MTA is better suited for digital marketing efforts, in which it’s significantly easier to track the customer’s full history of interactions with your company. This kind of behavioral monitoring for your customers is often not possible with MMM, depending on how you’ve modeled the problem.
Thanks to MMM, the days are gone when marketing departments had to adopt a “spray and pray” attitude, blasting customers with a campaign across multiple channels and hoping that something will stick. Such a strategy doesn’t just waste time and money; it also demonstrates a fundamental lack of understanding about customers’ desires and the product’s own unique selling proposition.
Careful analysis using MMM lets marketing departments better understand which factors are most influential for their campaigns, and which channels demonstrate the highest return on investment. With these data and insights in hand, marketing professionals will find it much easier to win approval for their ideas and initiatives from higher-ups.
Looking to unlock the hidden potential of your sales funnel from start to finish? Understand what you’re doing right and wrong by speaking with a qualified, knowledgeable partner for funnel optimization and analytics such as ironFocus. To learn more about how to improve your marketing strategy, check out the ironFocus blog for the latest news and updates.