Every entrepreneurial kid with a lemonade stand knows that if they want to create loyal neighborhood customers, their lemonade has to be good. But the real question is: What convinced their customers to buy that first cup?
Brands of all sizes and verticals need to understand how exactly their customers found their brand and what enticed them to buy. And there’s a tool that helps you decode this puzzle: marketing mix models (MMM). In this expert guide, Celina Wong, CEO at Data Culture and former Head of Data for TULA Skincare, walks us through how adding an MMM to your marketing stack can accelerate your growth stage.
What is a marketing mix model?
A marketing mix model is a statistical technique used to quantify the impact of various marketing activities on KPIs like revenue, ROAS, or cost per acquisition. It helps companies understand which elements of their marketing mix, especially ad spend, most effectively drive business outcomes.
Celina explains it like this: An MMM cracks the code on what made the kid’s lemonade stand take off. It assesses all the variables at play to reveal what exactly led to the conversion. For instance, the MMM may reveal that the hand-crafted sign caught people’s attention and that weather was also a factor — it was an exceptionally hot day. In other words, the poster created awareness, but the hot day drove business.
As a result, the kid changes their marketing and business strategy: setting up shop on cooler days may not be worthwhile, but they should still leave their lemonade sign out to spread the word.
However, unlike an entrepreneurial kid running a lemonade stand, brands spending millions in ad dollars can’t afford to rely on theory alone, which is where a modern MMM comes in.
How a modern MMM differs from historical methods
Modern MMM is the next evolution of marketing insights for e-commerce and media attribution measurement, tracking high volumes of digital touchpoints. Traditionally, MMM required lots of data but only accounted for limited variables, such as last-click attribution. However, MMM was forced to evolve as brick-and-mortar stores set up online shops, which required complex models to track online and in-person touchpoints simultaneously.
Modern MMM analyze recent marketing behavior to leverage your brand’s historical first-party data to predict future results. Using a Bayesian statistics model, the sets of data points are ordered on a timeline where time is the independent variable. This establishes a baseline for attribution and marketing channel performance that updates as you spend into existing or new channels.
The result? Modern MMM reveal daily ROAS information and saturation points that your brand can use to fine-tune marketing spend and other variables that affect conversions. In contrast, traditional MMM could only tell you about the current moment — the last interaction before conversion. With today’s MMM, you can assess signals like revenue against spending on marketing daily, by channels as well as campaigns. Doing so demonstrates which tactics produce more revenue, helping you allocate spend for revenue optimization.
How MMM compare to other forms of measurement
Because other forms of attribution, like platform, last click, or pixel-based MTAs, measure only a partial picture of your marketing spend, they can’t help you capture unbiased attribution across channels. In contrast to these single, pixel-dependent solutions, MMM use future-proof, first-party data that is collected by the brand and can be used to tell the full picture of how marketing spend impacts downstream performance metrics like Revenue, ROAS, and CAC.
Platform attribution
Platform attribution credits sales and conversions to the different marketing platforms and channels in your marketing mix. Celina says this can be an imperfect way to understand your marketing spend because the platforms want your brand to spend more money with them.
The reported numbers are often biased toward the platform, artificially inflating your conversions for that channel. “If your internal numbers don’t match what the platform claims, that may be why,” Celina adds. For this reason, many brands lose valuable time to data validation — because they’re tracking platform attribution manually.
Last-click attribution
Last-click attribution credits the conversion to the last marketing touchpoint or channel that the customer interacted with. While this measurement is an important variable — the customer’s last interaction with you — it may create gaps in your marketing insights. Last-click attribution fails to assess all the other steps on the customer’s path to conversion.
“Last-click attribution also biases your bottom of funnel by ignoring the middle and top,” Celina says. Only investing in one-third of your funnel doesn’t make sense. She adds, “Don’t forget the fact that if you just keep investing at the bottom, then you’re just starting to tighten that funnel up top.” In other words, if you don’t fund the mechanisms that lead the customer to the bottom of your funnel, you’ll eventually cut off the flow of customers.
Multi-touch attribution
To understand which aspects of your marketing have the highest impact on sales, you need to credit all the touchpoints a customer encounters before converting. Multi-touch attribution can pinpoint the incremental contribution of each customer interaction across a customer’s digital journey using click-based data, but only an MMM can help you uncover what caused a person to purchase both online and in person.
Celina offers an example: If your lemonade stand has a physical location as well as an e-commerce storefront, the MMM can help tell you who saw your signs in real life but converted online. This changes the calculation for the value of the physical stand — now you know it’s driving awareness for your e-commerce arm.
How backtesting improves forecasting confidence by reality-checking your MMM
The MMM is a key component of your tech stack that reveals just how many factors impact marketing, from seasonality and promotional content to different channels and in-person interactions. These insights are another value-add that sharpens your marketing IQ.
However, the outcomes from different vendors can vary drastically. And most models aren’t perfect right out of the box. That’s where backtesting (also known as holdout forecasting) comes in.
“The first time you run a model, it’s not going to be right,” Celina explains. “In fact, you should make sure you check your numbers. Then be prepared to fine-tune your MMM and keep adding variables until you’ve attained accuracy.”
How to backtest your MMM
To backtest your MMM, set it back in time by a few months and have it “predict” the past — i.e., use the data you already have to model outcomes. If your MMM predictions match or are similar to those historical results, you know it’s calibrated. If there’s too much variance between the MMM’s results and past data, consider tweaking or adding variables, like seasonality.
Forecast daily for accuracy
Once you’ve calibrated your MMM, make sure to check its daily forecasting accuracy because that’s where it will make a difference in your spend. “Even if your weekly forecasts are accurate, your daily forecasts can be inaccurate,” Celina says. Why? Because the longer the time period your MMM is assessing, the more aggregated the score it produces.
This follows a fundamental statistical tenant: averages regress to the mean. That’s why weekly or monthly accuracy scores aren’t very specific. “They’re averages over longer periods of time,” Celina says. “Instead of capturing the variability during those time periods, they aggregate volatility into big buckets of time.”
Instead, Celina suggests daily forecasting because, “The point of the MMM is to get highly accurate, specific results. Daily forecasting tells you how different from average you are and lets you make the daily tweaks that improve your average over time.”
When to invest in an MMM
So, when is the perfect time to invest in an MMM? Celina suggests when your business is ready to grow and you want to safeguard its future. In her words, “Don’t put all your ad spend eggs in one marketing basket. Instead, use an MMM to diversify.”
As a safety mechanism, MMM can help you optimize ad spend and grow your top-of-funnel. But Celina says it’s not the only use. She uses MMM to triangulate the knowledge she already has by validating her intuition and the other tools in her stack. The inevitable outcome? Improved ROAS thanks to incrementality.
Ready to try MMM for yourself?
Prescient is the leading MMM for e-commerce brands, offering daily channel and campaign insights. We help brands and agencies like HexClad, Backbone, Catalina Crunch, and Good American supercharge their marketing stacks and optimize ad spend without the heavy lift or added expense of other solutions.
Book a demo to find out more.
You may also like:
Understanding and optimizing customer acquisition cost (CAC)
Harness the power of Amazon with the new Prescient dashboard feature
