Measurability no longer has to limit your media choices

Yes, Apple’s App Tracking Transparency framework has killed Facebook’s ability to track users on iOS devices. Yes, GDPR, CCPA, Firefox, Safari and even Chrome are in the middle of killing the cookie and shattering attribution as we know it. But not being able to tie every conversion to a touchpoint; to an individual ID, doesn’t make a channel ‘unmeasurable’.

How relevant is that ID-based measurability anyway if it only looks at whether the media touchpoint happened – rather than how incremental it was?

Shift to marketing mix modeling

The end of the cookie is pressuring marketers to add more tools to their measurement toolbox – arguably a change for the better when those tools should have been there already. Two of the most important tools for every marketer are testing and marketing mix modeling (MMM).

Also called econometrics, media mix modeling, or holistic marketing modeling, MMM applies statistical analysis to different media channels to measure how their inputs correlate with business results. A strong model will also take into account relevant business inputs and economic factors beyond marketing, like product pricing, or inflation.

The case for using marketing mix modeling is stronger than ever: by definition, MMM is privacy-safe, and holistically measures incrementality. Most, if not all ID-based tracking methods, have uncertain fates; even the solutions deemed ‘privacy safe’ now may not be in the future.

MMM requires no cookies or any kind of session-based or user-based tracking. The only required inputs are marketing activity and business results, making it fully ‘future-proof’.

MMM for incrementality

ID-based attribution solutions are also typically limited in scope. Even the most thorough marketer will find it difficult to track all marketing channels with a single ID solution, let alone account for external factors. In contrast, a holistic marketing mix model can measure the impact of virtually every channel and variable on business performance: events (like a pandemic), macroeconomic trends (like a recession), press (like a bad news story), and more.

Modeling performance with econometrics also solves perhaps the most underrated flaw of most current attribution models: ID-based attribution will show whether a user had an impression or click before converting, but not how much that impression affected the user’s likelihood to convert compared to another channel.

Using statistical regression, MMM will show whether each channel or tactic was incremental, helping marketers avoid investing in ‘spray and pray’ tactics that drive impressions but not incremental conversions.

With such a strong case behind it, marketers should be chomping at the bit to use MMM. But many brands we talk to often assume that MMM is either too slow or too expensive to be worthwhile. The good news: marketers have more ways to start using MMM than ever before, with a smorgasbord of options across a spectrum of lower to higher support, complexity, and maturity.

Homegrown MMM models

Example: Robyn from Meta, Lightweight MMM from Google.

Homegrown models might be the best fit for businesses with a strong internal analytics team (or integrated consultants) and a relatively simple channel mix.

Teams leverage open-source code, input their brand’s data, and build an entirely custom model. While the source code is free, this approach requires proficiency in, at the very least, the basics of coding and econometric modeling, as well as time to run multiple iterations and adjust the model.

Self-serve and software as a service (SaaS) MMM models

Example: Recast, Keen.

SaaS marries the flexibility of the homegrown models with the advanced software offered by full-service partners, but at a fraction of the cost. This approach does require more involvement from the brand, particularly in the initial data collection phase where having more data sources or inconsistent data infrastructure may add time-consuming complexity.

Self-serve models might be best for businesses with access to data and analytics resources who can lead the implementation of econometrics but may not have the bandwidth or technical skills to build and iterate models from scratch.

Full-service MMM

Example: Neustar, Analytic Partners.

These spring to mind first for MMM: big players that have been in the business for decades. Their platforms and econometric modeling techniques are highly refined and often include a ‘combo’ offer with measurement tools like brand health tracking, ID resolution, or multi-touch attribution (MTA).

Full-service providers offer hands-on support from collating data to building models, generating insights, and recommending media optimizations. With the support of account teams, marketers can focus all their attention on actioning relevant insights.

For marketers with multiple KPIs, differentiated brands, or highly complex media programs, full-service MMM might be well worth the investment.

Beyond modeling, experimentation is critical

As strong as econometrics is for measuring incrementality, the results are still based on correlation.

Testing, particularly a robust and future-proofed methodology like geo-lift (matched market), is the best way to prove causation. Every marketing team should be testing rigorously to measure incrementality and calibrate their MMM model based on the results.

The demise of the cookie will continue to disrupt the way that brands measure marketing performance, but marketers can still measure with confidence when they dust off the science and stats-based tools of testing and econometrics.