The benefits and limitations of Marketing Mix Models
During my recent discussions with Mojo clients, I’ve heard the words “Marketing Mix Models” pop up more often than they used to. These models are often produced in-house to understand which activities drive sales and profit in a given campaign. At their most basic level, you can think of Marketing Mix Models like this: they show how a variable (a marketing or sales activity, for example) is related to an outcome (sales, profit or both).
At Mojo, we act as a strategic measurement partner. As such, my data science team is constantly working to complement and supplement the work of in-house analytics teams—deriving more granular insights than they may have the resources to produce, and translating these into optimizations that drive brand growth. My recent discussions about Marketing Mix Models led me to dive deeper into how these are being used in today’s marketing landscape, and how they fit into the work we’re doing at Mojo.
Here’s what I found: Although Marketing Mix Models are useful tools, they don’t improve campaign performance on their own. Where many brands struggle, and justifiably so, is translating these models into an actual marketing plan with specific vendors, calls to action, targets, campaign cadence and a multitude of other variables.
Here’s a quick primer on Marketing Mix Models, and how a strategic measurement partner can help transform analyses into optimized performance.
What Are Marketing Mix Models?
Marketing Mix Models are market research tools used to provide a deep understanding of multiple drivers of sales and profits. Originally used in Consumer-Packaged Goods, Marketing Mix Models are now implemented in industries like pharma to evaluate a broad range of marketing campaign components. (Think: advertising, promotion, media weight levels, called-on targets and beyond). Brands often use these analyses to help make specific marketing decisions such as optimal spend, determine tradeoffs and guide strategic planning.
These approaches use regression models to determine how “x” inputs (independent variables) relate to an outcome (a dependent variable like sales, profit or both). Once the model and independent variables have been validated, data teams can manipulate the variables to determine the net effects on sales and profit.
For pharma, it’s especially important that your data is specific to individual brands, not the company as a whole.
The Benefits (and Challenges) of Marketing Mix Models
In the hands of skilled data scientists, Marketing Mix Models can identify key drivers of sales, profit and performance. They can also enable brands to explore high-level “what if” scenarios that compare marketing tactics; project sales activity during a specific period; and assess budget allocation across various channels.
But just like every analytics tool, Marketing Mix Models have their drawbacks. These models are designed to say how much to spend in each channel, not how or with which vendor. Because they determine “what” but not “why,” these models tend to make numerous assumptions.
Some additional limitations include:
- Significant cost and time required
- Lack of measurement standards and transparency: It’s often difficult to get details on how models are created or the measures they use
- Messy data can affect validity, as is the case with any analytics tool
- Difficult to obtain accurate detailed inputs (for example, the number of samples given to each HCP)
- Advertising content is difficult to quantify
- The non-linear effect: A 10% investment does not always lead to a 10% increase in conversions
- Final models are not stable and can be a recipe for disaster
On another note: Marketing Mix Models are most often used by advertisers to determine the best media allocation across media types. However, research shows that marketers should take these recommendations with a grain of salt.
Test-Control Design and Bridging the Gap
Test-control design is still the gold standard in data science. It can be directly tested, has far fewer assumptions than Marketing Mix Models and, most importantly, is directly causal. Mojo can help brands implement test and control design, which is an effective way to “pressure test” the assumptions associated with Marketing Mix Models.
At the end of the day, Marketing Mix Models won’t boost campaign performance on their own. A strategic measurement partner like Mojo can help brands bridge this gap. Through rigorous measurement and insights, we can determine how to implement recommendations for channel spend—all the way down to vendor, cadence and segmentation.
Like most things in marketing, Marketing Mix Models shouldn’t operate in a silo. When paired with test-control design and rigorous measurement insights, these analyses can help brands determine optimal channel spend—and translate that into optimized campaign performance.