This guest post was written by AdTech veteran Dan Elddine
Imagine a world where most existing industry measurement solutions are either ineffective or sit within siloed walled gardens who are incentivized to showcase their performance in the best possible light, disregarding brands’ marketing activity across other partners and channels. This is the challenge modern marketers face as they slowly start to recognize measurement as a significant casualty in the post-cookie era.
As many in digital advertising have been busy anguishing over the latest Adalytics report on the proliferation of Made for Advertising (MFA) inventory or relentlessly firing shots across the bow at Google and the Privacy Sandbox in not-so-veiled attempts to tout their own alternative solutions (or hide real fear about how they’ll function/perform in a post-cookie landscape), marketers face more questions than answers about how to measure campaign effectiveness moving forward. Multi-Touch Attribution models have long failed to deliver on their promises and will only worsen without scaled, cross-channel identifiers. The shine of Retail Media Networks are starting to patina as brands learn that they’re some of the biggest culprits of MFA inventory arbitrage, come with heavy pay-to-play service price tags, and have mostly siloed, self-favoring measurement solutions.
Simply stated, lower-funnel attribution-based measurement is in for turbulent times. And while marketers wait for alternatives to surface, Sandbox applications to improve, and cookieless modeling techniques to be refined, they are turning to older, tried-and-true methods. Hence the resurgence of Marketing Mix Modeling (MMM).
In what’s perhaps an acknowledge of the limitations inherent in post-cookie measurement, Google (like Meta before them) introduced Meridian, their open-source MMM solution which they claim as innovate, transparent and actionable. That said, it’s important to recognize that MMMs, regardless of the provider/solution, come with their own set of challenges. That said, by 1) blending MMM’s traditional methods with the breadth and depth of today’s data and 2) taking a practical approach in incorporating MMM results into their everyday advertising operations, modern marketers can deploy MMM as an updated, essential tool in their toolkit (even if it’s been there all along).
A Quick MMM Primer
What is MMM?
Decades ago, marketers needed a way to quantitatively understand the effects of their paid and owned marketing efforts on sales across channels and timeframes. As a result, they worked with econometricians to design the MMM methodology. Econometrics, the broader field from which MMM is derived, applies statistical and mathematical theories to economics to test hypotheses and forecast future trends. It enables the integration of theory with data to give empirical content to economic relations. MMM applies a similar approach to marketing by (on a high level) taking inputs like channel-level spends and using methodologies such as regression or time series analysis to assess their impact on sales or conversions. The results provide insight into which channels positively contribute to sales and which do not. From there, marketers can adjust strategies and tactics such as channel-level investment accordingly. Figure 1 illustrates the MMM process in action. The core principles of MMMs have largely remained unchanged since their creation.
Figure 1 – MMM process (abstracted)
Common Challenges
In practice, MMM is not without its caveats. For one, MMM generally tracks spends and sales in aggregate over extended time periods (e.g. quarters).This makes it difficult to discern nuances such as different creative concepts or if the majority of one’s TV ads were delivered to empty rooms while your target audience was grabbing snacks from the fridge. In addition, looking at aggregate channel-level spend doesn’t account for price variances across or within channels. For example, spending $100 may get you one 30-second ad on primetime television, ten 15-second ads on late night reruns, or several thousand Instagram sponsored ads. Even novice marketers can tell that these aren’t equally impactful forms of advertising.
Second, MMM can’t account for channel-level tactics such as targeting (or lack thereof). As such, it can often aggregate channel activity too broadly – e.g. lumping all digital advertising together rather than isolating the differing impact of display vs online video ads. This leaves marketers with at best, more questions than answers and at worst, the wrong inference on what is and isn’t driving results.
Lastly, MMM incorporates time series analysis. This technique, broadly defined, involves monitoring the impact of various inputs on outputs across different time periods, inherently requiring time to identify trends (ie. quarters to years). This analysis of “slow moving” data, together with MMM’s frequent recommendation for significant adjustments in channel-level budget allocations, can be likened to navigating across three freeway lanes while relying solely on the rearview mirror.
Recent Renaissance
Despite these challenges, MMM’s have recently experienced a resurgence with the advent of modernized approaches by relative newcomers in the space like Google and Meta as well as more comprehensive platforms from stalwarts like Neustar. Other providers like Measured are banking on automation to incorporate results from other measurement solutions such as incrementality testing into its models to better link MMM to real-world media contributions and improve efficacy. Overall, these organizations can supply models with newer, “faster-moving” data sources (i.e. data that can be surfaced in hours/days vs weeks/months/quarters) with more granularity within existing media channels. This is particularly true of digital channels where more granular reporting has long been a differentiating benefit relative to offline channels. These additions not only help improve the utility of MMMs, but also serve as a bridge between MMM’s inherent “big picture” nature and more immediate action-driven measurement methodologies.
While each provider takes a slightly different approach – largely based on the data they have direct access to or are able to stitch together via other technology partnerships (think clean rooms and identity resolution providers) – the general direction MMMs are heading in is one that 1) produces more granular insights 2) drives faster and more frequent actionability and 3) mitigates privacy concerns and the degradation of third party identifiers.
Looking at the three partners mentioned above, we can see how this is taking shape. Neustar introduced new tools back in 2021 to help marketers bridge their MMM solution with their other products/features such as multi-touch attribution, audience augmentation and identity resolution. That same year, Meta launched Robyn, an open-source MMM that ingests data from other tools such as Facebook Lift, geo-based measurement solutions and ML-based modeling. And just recently, Google aims to incorporate reach, frequency and incrementality results alongside their search data to drive their own MMM outputs and emphasize its role in measuring advertising effectiveness moving forward.
The Path Forward
With these new MMM providers and competing modern methodologies gaining steam, it can be difficult to figure out where to begin. However, whether you’re starting out on your MMM journey or revisiting its role in the wake of measurement gaps left by cookie deprecation, here are ten ways marketers can maximize the utility of MMMs in the evolving marketing data paradigm:
- Start small – While it’s tempting to jump into the deep end and incorporate advanced metrics and calibrations, it’s better to first ensure every marketing channel utilized is accurately accounted for in your MMM.
- Increase channel-level data granularity – As most platforms can report on nuances such as placement type, media environment, or creative variation, it makes sense to incorporate these details into your MMM. Econometric methodologies can identify how./if the additional data granularity impacts MMM outcomes, all without substantial resources or model training requirements.
- Incorporate relevant, optimizable variables – Adding stalwart metrics like reach and frequency to MMM is an effective way to integrate media planning aspects into the model as most platforms can optimize towards these metrics. In addition, it’s worth exploring the addition of emerging metrics such as attention or media quality indices. While their direct, isolated effect on sales may be minimal, they can provide valuable insights to inform campaign investment and optimization.
- Be aware of model biases – Intended or not, MMMs may favor certain channels, time periods and/or other factors. This extends beyond MMMs and to all third-party measurement products. For example, a MMM provided by a search engine giant may demonstrate bias towards search, whereas a closed-loop attribution model from a retail media network will likely show favor towards its own data and inventory. As such, it’s important to be aware of these nuances when interpreting the outputs of such models.
- Demand transparency – Having access to the underlying code, source data, and other foundational elements of a model should be table stakes, particularly if you’re investing in an expensive MMM from a non-media affiliated provider.
- Require media partner participation – Establish clear expectations with media partners that your campaign data will be used to inform and action on your MMM. DSPs, social platforms, RMNs, etc. should each be prepared to supply data (privacy protected of course) to support your MMM efforts even if the model is associated with a competitor (e.g. Amazon and Google, Google and The Trade Desk, Meta and Google).
- Seek customization – Beyond the ability to modify independent variables, most businesses operate across a variety of sales channels and thus may benefit from MMMs capable of modeling across a range of dependent variables. This can include direct sales from a single point of distribution, or indirect sales indicators such as online interactions. While these variables might not paint the holistic picture typically associated with MMM, they can provide quicker unbiased reads on actionable performance.
- Use MMM to inform top-down strategies… – Each campaign planning cycle should incorporate the latest MMM readouts along with creative approaches, data insights and media partner strategies. This integration can serve as a foundational step for the effective allocation of campaign budget.
- …As well as bottoms-up approaches – In addition, variables that enhance MMM outcomes should be used to inform campaign decisions. This might range from simple adjustments, like reallocating budget away from underperforming partners to more sophisticated techniques such as optimizing towards novel signals (e.g. attention, media quality).
- Embed MMM into your operations – Like any measurement solution, MMM is only as effective as its real-world application, requiring discipline and time to integrate into one’s operating process. As such, we recommend adopting a practical approach by establishing a cycle that includes:
- Reviewing MMM outputs across strategy, planning, activation and analytics teams, and sharing relevant findings with agency and media partners as needed.
- Jointly understanding and agreeing on how MMM insights are to be applied, whether that be current campaign optimizations or future cross-channel investment strategies.
- Monitoring the effectiveness of implemented actions in future MMM readouts. This will help foster the cross-functional ownership that’s needed for the ongoing evolution of modern marketing measurement.
A Solid Data Foundation Remains Key
The deprecation of third-party identifiers will significantly impact existing industry measurement solutions in both the short and long term. However, with this shift comes the opportunity to develop new tools, or in the case of MMM, refresh existing ones with richer, more granular, and more actionable data. As with any model however, the principle of “garbage in / garbage out” applies, underscoring the importance of a robust data strategy that addresses data infrastructure, capture, governance, storage and deployment. Thus, specialist partners who are adept at providing data and technology solutions while also understanding the practical application of such services for brands will become increasingly valuable to marketers moving forward.
