By mid-2024, Google (the last major holdout) will begin phasing out third-party cookies from Chrome. Opt-in, first-party cookies will remain for direct tracking of campaign results but will lack the multi-touch attribution of digital impressions leading up to the tracked engagement. Soon all the cookies will have crumbled, and marketing attribution models based on their usage will be truly dead. Ideally, your marketing team will have already transitioned to other forms of measurement and modeling well before the death of Chrome’s cookies.  One long-term solution to your woes for measuring marketing contribution will be the tried-and-true method of MMM, or media mix modeling. What is Media Mix Modeling?Also known as marketing mix modeling, MMM is a statistical method of measuring the performance of diverse paid, organic, and earned media on driving key outcomes, such as sales or revenue.  The top-down analysis afforded by MMM allows your marketing team to adjust and optimize the mix of marketing, i.e., what kinds of ads get placed where and when, and make meaningful predictions about their outcomes. Most importantly for the future without cookies, MMM doesn’t rely on tracking individual users or their behavior.

How Does MMM Work?

MMM works by analyzing large data sets (ideally at least two years’ worth of weekly data) consisting of marketing activity, pricing, trade or sales team activity, competitive activity, economic trends, and other external factors that can positively or negatively influence your sales outcomes over time. The analysis assesses how each of your media channels, offers, and promotions contribute to total sales in the context of these other factors outside of your control. This is essentially detecting how certain changes in impression levels and the media mix change the sales outcomes in the weeks and months that follow. With non-marketing and external factors included, MMM can calculate marketing contribution even when sales are declining or are partially offset by competitive activities. A good MMM compares any number of variables against each other in order to adequately track the strength and effectiveness of each in contributing to the final goal. These variables could be sales, revenue, new customers, engagement, or some other metric that you’d like to measure.

Why is MMM Important in a Cookie-less World?

MMM is important for a big-picture perspective in a post-cookie world and actually offers advantages over MTA (Multi-Touch Attribution) modeling in the current world with cookies. Attribution models added insights beyond last-touch attribution, but never fully captured the impact of non-digital media or external factors. While bottom-up alternatives to MTA may emerge, MMM remains the best and most well-understood top-down attribution method. The flexibility of a strong marketing mix model is unparalleled in terms of a holistic, top-down, controlled viewpoint. As technology improves, so too will the ability to find data points with which to build a robust model.

How to Use MMM

The first and most important step is to gather a comprehensive data set to build the model. Like other statistical models, the key behind a strong prediction is finding the right mix of quality data with as little noise as possible. If your company doesn’t have at least a couple years’ worth of data to begin with, the validity of your model may be limited (but can still offer directional insights).  To start with, gather as much information as possible as it relates to:

• Marketing Activities: Promotions, sales, and all marketing activities. You’ll need a breakdown by week (monthly works but is less predictive) of impressions and spend for all marketing activities.

• Non-Marketing Activities: price changes, product availability, distribution and product add/drops that can change sales outcomes independent of marketing. The model not only quantifies these impacts on sales but also better predicts marketing contribution with more complete data.

• Sales: Unit/dollar volume sold (including the percentage on promotion).

• Market Conditions (or External Factors): Economic trends such as inflation or consumer sentiment, competitive spend (if available from 3rd party sources), and other exogenous factors (e.g., COVID, etc.).

Naturally, gathering and combing this information will be easier if your company has already invested in high-quality data storage and organization. What if you didn’t?  Then what?  The effort for the first model may require a lot of searching and manual aggregation, but this does get easier over time. If you want to keep the process automated after the first time you gather this data, you’ll need to find a way to create a database with pipelines that automatically update as new information becomes available. Often not all data can be automated this way, but there are other methods such as through a centralized database (typically tied to dashboards).

Modeling Data

A marketing mix model is going to attempt to take many independent variables and quantify how much each contributes to sales or revenue (or some other dependent variable).  For example, you may have ten (or more) independent variables like money spent on social media advertising, programmatic display, paid search, influencers, PR, TV, sampling, email marketing, and print ads, a concurrent sales promotion, and rising inflation. Each of these variables will have some impact on the dependent variable, which you could say is sales.  An MMM using these variables would attempt to explain how much impact each of the three variables had on the final outcome.

Comparing Variables

The beauty of using a statistical model to analyze data is that you can leverage the results to create simulation or optimization tools. These tools allow you to input media plans and vary scenarios to compare and forecast the outcomes.  For example, your analysis of the historical data might reveal that your social media had an outsized effect on the sales during the period it was running, whereas your paid search had minute impact. If you model the sale with a different media mix, how would that change the outcome? A robust model can give you insights like these into market forces you weren’t even aware of.

Real World Example: Evivo Infant Probiotic

Evivo infant probiotic (a parade category product) launched as a DTC brand in July 2017 and went up against leading companies such as Gerber, Enfamil, and Culturelle for market share.  After two years of significant marketing spend for a startup, the company needed more sizeable sales volumes and to reduce marketing spend. Evivo turned to E29 for media mix modeling. We used our model to uncover key marketing insights that enabled a 50% reduction in overall spend when reallocated between channels to maintain/grow sales volume more effectively than the previous media plan. All without the use of a single cookie.

Modeling Success

As third-party cookies disappear from browser histories around the planet, marketing mix modeling that has been in place and evolving for decades is likely to have a resurgence that will drive marketing into the future. Start now by organizing your data storage, and determining your variables, and, if you need it, E29 can help you with the all-important modeling that can build your digital marketing strategy in a post-cookie world. Contact us today to see how E29 can help with your marketing mix model.