Now that the industry is moving toward MMM—largely because we all know attribution alone doesn’t cut it—there’s a new risk emerging: overconfidence in static models. Most MMM models assume that marketing effectiveness is constant, but in reality, it’s under constant change. If your model treats performance as a static, without explicitly modeling time, it’s fueling poor decisions and wasted ad spend.
Why Marketers are Losing Faith in Traditional MMM
The most expensive thing you buy is either the thing you don’t use, or the thing that’s generating losses. Sadly, most MMMs fall into either one of these categories. Still today, most MMM models fail to capture the complex reality that marketers navigate on a daily basis. It’s understandable that marketers feel skeptical that a model could understand their data better than they could do themselves.
With the rapid growth in interest of MMM, how can this still be the case? In short, it’s incredibly difficult to capture the complex world of marketing in a single equation. However, many vendors are unconcerned with this. They see the MMM as a silver bullet to their clients’ needs, since the general discourse is that it’s so much better than traditional ea
This is putting the entire industry at risk. Unserious vendors are fueling a distrust in the methodology itself for no other reason other than complacency. The main culprit? Very few MMM models are concerned with time.
Traditional MMM Limitations
There is a growing discrepancy between the requirements marketers place on the models, and what is feasible from an econometric perspective. Historically, brands have used MMM once or twice per year to assist with overall budget planning. But today, it’s becoming increasingly popular to work with MMM continuously, as a way to shift away from tracking-based attribution tools. This means that the model needs to be continuously updated with new data. It’s no longer enough to mask short-term trends with long-term averages.
Traditional MMMs assume that marketing effectiveness is constant. But every marketer knows that performance shifts week to week based on platform changes, competitor activity, the creatives you’re running. The list goes on. A marketer will have little use of a MMM model that tells you that your Facebook ads deliver a 3x ROAS over the last 12 months, but fails to show that ROAS was 5x in Q4 and 2x in Q2. This is amplifying misinformed budget decisions.
Last but not least, incrementality is always a top subject within the field. A significant drawback on traditional MMM approaches is the assumption that the baseline level of sales (what you would have sold if you didn’t do any marketing) is static over time. This is by no means a realistic assumption. The underlying demand of your products shifts over time, as does your brand equity. A model that fails to capture this won’t have a chance to accurately capture your ROAS.
The Rise of Modern MMM
Instead of assuming that marketing effectiveness is static, MMM needs to be dynamic and time-aware. At the start of 2025, Google released their long-awaited open source package Meridian which partly addresses this. A clear signal to the rest of the industry that MMM needs to evolve in order to remain relevant.
So what sets time-based MMM models apart from their static counterparts? First of all, baseline sales must evolve. For many larger brands, understanding the evolution of the baseline itself is a goal as it allows them to understand how their marketing is driving long-term effects in terms of a growth in brand equity. Second, media effectiveness must change over time. Instead of a static effect per channel, modern MMM should track how this fluctuates weekly or even daily. Last, diminishing returns need to be adaptable over time. This is key in order to derive optimal budget allocations. It’s no longer enough to assume that these relationships are the same across the year, especially in industries heavily affected by seasonality.
From the perspective of an MMM vendor, it’s also worth mentioning that as brands work with this on a continuous basis, the flexibility that time-based models grants help to establish trust in the methodology. Static models that are trained on new data will likely shift around significantly, since they’re incapable of handling shifts in the data. This leads to mistrust by marketers, often asking themselves why the ROAS that model reported for last month all of a sudden looks different. The alternative, to not retrain models is worse still. Models become outdated quickly, and a stale model fed with new data will likely create a false sense of comfort.
How Can You Shift to Time-Based MMM
It’s clear that you need to address the limitations of traditional MMM in order to stay competitive in today’s fast-changing marketing landscape. It’s also clear that a big step in the right direction is to incorporate time into your model.
- Move away from static MMM approaches. If your model assumes that marketing impact stays the same throughout the year, it’s time for an upgrade.
- Use advanced regression or state-space models. These both allow marketing effectiveness to evolve dynamically over time, rather than reporting on long-term averages.
- Ensure you update your MMM continuously. Instead of running an MMM study once per year, the modern approach to MMM should be always on, providing updated insights weekly or even daily.
Key Takeaways
Most MMM models assume marketing effectiveness is static, but in reality, it shifts constantly. To remain relevant in today’s fast-moving landscape, MMM needs to be always-on, dynamic, and explicitly time-aware; tracking changes in marketing impact weekly or even daily. Brands that fail to move away from static MMM risk making decisions based on old data, eroding trust in the methodology, and missing critical market shifts. The future of MMM is flexible, real-time, and built for continuous optimization because in marketing, time is everything.