Why every business needs a Media Mix Model

Introduction

Global Advertising spend for 2024 is forecasted to reach US$752.8B. With this much money being invested in advertising, how do companies ensure that they’re allocating their funds appropriately? Every business is likely to tackle this problem differently. However, with so much data at our fingertips, budget allocation shouldn’t be a guessing game. Relying on heuristics and internal opinions without considering a data-driven approach could be the difference between long-term success and recklessly setting money on fire.

Large advertising providers like Google and Meta dedicate a lot of time and energy to providing us with the tools and insights to make informed decisions based on channel performance. Is this helpful in the budget allocation process? These giants invest billions in R&D to provide us with user-friendly analytics, data-driven Multi-Touch Attribution, cross-platform tracking and “Smart” bidding, just to name a few. Surely, they’re the experts and we should use their tools to allocate our budgets in the most effective way. But what if they aren’t THE experts.

This is where the world of MMMs could be the secret weapon that your business needs.

Platform Attribution vs. MMM

It’s important to clarify what we mean when we talk about “platform attribution” and “MMM” in this discussion. The distinction between these two approaches is not black and white. For the purpose of this discussion, we’re defining the terms as follows:

  • Platform Attribution: The reports found in Google Analytics, Meta Ads Manager, AppsFlyer or similar that aim to attribute conversions or revenue between different media channels. These reports are usually limited to digital channels. On the simplest end of the spectrum, we have first- and last-click attribution models. On the opposite end, we have “smart” or “data-driven” attribution models where buzzwords like machine learning and artificial intelligence are commonly used. These reports rely heavily on click-tracking, floodlights and browser cookies to report on impressions, clicks, cost, CTR, conversion rates and ROI broken down by channels, campaigns, geographic regions and creative executions to name a few.

  • MMM: Media Mix Models (or Marketing Mix Models). Analytical models that analyze spend per channel and resulting patterns in conversions or revenue to explain the relationship between the inputs and the outputs of all advertising investment over a period of time. They can include data from both the digital and non-digital sources and also allow businesses to incorporate internal expertise and experience into the analysis.

The shortcomings of Platform Attribution

Platform attribution is an important tool for media specialists to use. When used in the right context, it can highlight weaknesses in keywords, bidding strategy, SEO, creative performance and many other areas. When allocating budget between channels, this usefulness is diminished for a few reasons:

  • Limited field of view: Consider a person viewing an and on the Facebook App and then opening their favourite browser to search for your website, click on a Google Search ad and purchase a product. Even with everything set up perfectly, Google has no idea about the Facebook Ad impression and Facebook has no idea about the Google Search Ad. In reality, with Meta floodlights configured according to best practice, both platforms will claim responsibility for the conversion. Just like that, with only two digital platforms and a single sale, you have conflicting sources of data and neither of them tell the full story. This problem gets exponentially worse when you add offline advertising, word of mouth, cross-browser activity and a host of other real-life factors into the mix. Even so, each of these platforms will overwhelm you with dashboards and metrics to use when making your investment decisions.

  • Reliance on Cookies: This is usually the first point raised in most conversations about the value of MMM. Digital platforms use cookies to essentially tag users as they browse the internet. For years, this has allowed companies to identify browsers and track impressions, clicks, conversions and other activity and information between different websites. As privacy concerns and regulations have grown, the use of cookies has become increasingly restricted and this type of cross-website tracking (third-party cookies) is being phased out. Without being able to track exactly which ads a user saw and/or clicked on before visiting your website, companies like Google and Meta will have a very difficult time reliably connecting the dots. Sure, new techniques and technologies are being developed to overcome this but they will only be more opaque and less reliable than cookie-based tracking was in the first place.

  • Black boxes: As we stated earlier, the media giants invest billions in R&D. They measure and store unimaginable amounts of data (or signals) and employ armies of data-scientists, economists and researchers to develop their technologies. As a user of their products, this makes them reliable and cutting-edge. It gives these enterprises a level of credibility that only a crisp white lab coat can command. However, these platforms generally only leave us with the end result and possibly some recommendations on how to use or interpret it. We have no idea what inputs were used and, even if we employed our own army of scientists, we wouldn’t be able to recreate their results. This poses a large risk when we need to translate these results into actions for our own business. It also means that our only option is to blindly follow their recommendations and then trust those same methods to measure effectiveness.

MMM: The solution

A Media Mix Model is not a silver bullet. Instead, it’s an additional tool that advertisers should use when evaluating the effectiveness of their budget allocation. In contrast to platform attribution, a MMM is ideal for a few reasons:

  • Holistic view: While MMMs don’t provide the granular insights that these platforms claim to offer, they do consider the full picture of what’s actually going on. A well-built MMM includes every media channel or event, whether offline or online, and analyses it’s impact on the desired business outcome. Aside from direct investments in media, MMMs can consider other factors like changes in price, inflation, GDP, weather fluctuations or any factor that may be relevant to your business.

  • Privacy-first (no cookies required): Since MMMs require only the inputs and the outputs, there is no need for cookies or any other advanced user tracking. This means, even in a cookie-less world, advertisers can gain insights on the performance of their media channels and make decisions on how best to optimize their budget allocations.

  • Explainable relationships: The outputs of a MMM will not only tell you how each media channel has performed but will also explain why. Modern MMMs report on 3 main elements per media channel: The magnitude of the channel’s impact, the saturation curve per channel and the delayed effects of each channel

    This is not only useful for explaining certain decisions stakeholders within the business but it also allows advertisers to sense-check the outputs of the model. For example if Google Search ads have more of a delayed impact than a billboard campaign, this may immediately flag concerns around the accuracy of the model based on prior experience. In addition, knowing how saturated each media channel is for your particular business is crucial when deciding where spend needs to be increased or decreased while maintaining maximum impact.

Conclusion

In a data-rich world, budget allocation is crucial for a successful advertiser. While individual platforms use sophisticated technologies to report on performance at a granular level, they are not the best tool for the job. A well built MMM gives the advertiser a clear view of the whole situation by considering all inputs into the system and allowing for internal business expertise and prior experience. It does so without requiring any privacy-infringing data and without obscuring the underlying mechanics of the relationship between investment and outcome.