The loss of third-party cookies and mobile ad IDs, as well as ongoing data regulation, is driving huge investment in first-party identity and measurement solutions to help brands effectively plan, execute, and measure marketing campaigns. These solutions are deterministic in nature. And we’re seeing some seriously impressive solutions emerging in this space. But with so much emphasis on deterministic solutions, such as first-party identity graphs, I believe marketers may be overlooking the power of probabilistic modeling—in particular, a form of privacy-safe omnichannel measurement that’s making a comeback.
Marketing mix modeling (MMM) is back. (Truth is, it never really left.) But it’s more granular and faster than before. And used alongside deterministic measurement, it can help marketers make more confident, informed decisions.
Here’s what you need to know to get started with the new, rebooted MMM.
The missing piece of the puzzle
You might already be familiar with MMM. Hardly a new entrant to the marketing measurement landscape, it’s been tried and trusted by CPG brands since the 1980s and—at its simplest—it helps marketers understand the ROI impact of channel-level performance.
Fast-forward several decades, and MMM has been widely cast aside by performance marketers in search of the deterministic truth. They didn’t see how a probabilistic approach could possibly deliver the quality and accuracy of granular-up measurement: a fair argument to level at the MMM of the ‘80s.
What many marketers don’t yet know is MMM has evolved. The new iteration drills down deeper than its predecessor to optimize at audience, message, and channel levels. It also brings valuable insights across the entire marketing ecosystem—even within elusive walled gardens that typically present frustrating data access roadblocks.
Indeed, the companies that are currently most vocal about MMM are Google and Meta. They’re talking to performance marketers from sectors like financial services that have never used MMM before and explaining how it is evolving to enable quicker data analysis and accurate evaluation of incremental lift.
And many brands are also using the evolved version of MMM, even if they’re not aware of it. For example, if you’re doing cross-screen sync work to understand reach across TV and other devices, there’s a good chance you’re already using some MMM techniques.
MMM is also expanding to create other models such as full funnel modeling (FFM). This advanced iteration analyzes interactions at each stage of the customer journey, showing how activity in different parts of the sales funnel impacts other efforts up and downstream.
What’s more, MMM incorporates privacy by design. That’s to say, it delivers non-PII (Personally Identifiable Information) data that brands can move across borders to extract maximum value.
Our post-cookie, privacy-first landscape has created the perfect storm for a renaissance of MMM—and probabilistic modeling in general—as a valuable complement to deterministic measurement.
It’s all about the results
Some marketers will still see MMM as statistical mumbo jumbo compared with the deterministic truth they’re used to. And, with MMM involving everything from multivariate regression statistics to Markov chain Monte Carlo simulations, marketers are unlikely to understand exactly how it works.
But what they will understand is the results it can achieve. Numbers talk louder than words, and we’re seeing brands that combine probabilistic MMM with other forms of measurement to do channel mix budget optimization typically achieve 10-15% improvement on outcomes year-on-year.
Marketers will also get to grips with the predictive power they’ll enjoy as a direct result of using MMM. When you can forecast accurate outputs for next month and next quarter sales, on repeat, statistical mechanics aren’t likely to keep anyone up at night.
The point is, with the right engine in place brands can unify and connect all this data in real time with exciting possibilities. Think competitive pricing, site optimization, direct marketing activity, and paid promotions that combine to improve future outcomes.
With probabilistic as well as deterministic techniques, marketers can connect metrics that have gained prominence in our digital world (the clicks, views and so on) with the long-term objectives brands really care about such as sales outcomes.
The best of both worlds
Considering all the exciting advances in MMM, marketers might wonder whether they can do without deterministic measurement—but that’s really not an option. Looking to the future, they will need to do both together. Probabilistic and deterministic methods aren’t mutually exclusive. In fact, they’re perfect partners.
An effective way forward is to use deterministic measurement as far as you can as a Plan A, and then combine it with MMM wherever it breaks to get the fullest picture across audience, message, and channel.
Time for a renaissance
Ultimately, the two questions on performance marketers’ minds are: 1) ‘Is it working?’ and 2) ‘How can I do it better?’. But stricter regulations over consumer privacy and data collection, as well as the loss of third-party IDs, means they’re often struggling to answer these.
That’s why the timing is just right for a resurgence of probabilistic modeling that is helping marketers step outside the log file and widen their source of truth.
If you could use MMM to complement your existing measurement strategy and hit higher levels of confidence, why wouldn’t you?
