Agile marketing mix modeling extracts two signals: how spend moves price, and how fast an audience saturates. This matters because flat historical spend, a small bump held steady for months, carries no real signal to learn from. Without deliberate variation through match market tests, comparing baskets of markets, marketers are stuck guessing. This video explains how agile mix modeling and match market testing turn weak data into reliable platform insight.
Why Agile Marketing Mix Modeling Needs Match Market Testing to Work
Agile marketing mix modeling needs match market testing to work, because historical spend data is often too flat to reveal how price and saturation actually respond to spend, no matter how sophisticated the model is. In this video, John Wallace, CEO of LiftLab, names the two signals every ad platform relationship boils down to: what happens to media price as spend increases, and what happens to audience quality, saturation, plateau, fatigue, as a platform runs out of its best prospects. LiftLab’s answer is what it calls an Agile Mix Model, a two-stage approach that quantifies both the price effect (the same dynamic behind rising CPCs as you bid deeper into a search auction) and the saturation effect, going beyond directional intuition to a precise degree.
The catch is that the model is only as good as the data feeding it. Wallace shares a real example: a customer who ran $300 a day on Snapchat, bumped it to $375, and held that flat for two months. That’s simply not enough variation for any model to learn from. The fix is to manufacture better data through geo-based match market tests: not a single market compared to another, but a whole basket of markets tested against a comparable basket, with spend deliberately fluctuated to build a real body of evidence about price and saturation. What those experiments reveal then gets fed straight back into the model.
In this video, you’ll learn:
The critical importance of identifying price and saturation signals in ad platforms
Why traditional marketing mix models struggle with low-signal historical data
The methodology behind the “Agile Mix Model” approach to data analysis
How randomized, geo-based experimentation creates higher-quality data for decision-making
Why matching baskets of markets is superior to simple market-to-market comparisons
TIME STAMPFull Video Transcript: Agile Marketing Mix Modeling and Match Market Testing Explained
+–
John Wallace explains how agile marketing mix modeling and match market testing work together to reveal ad platform price and saturation signals.
[0:00 – 0:22] Defining Necessary Ad Platform Signals
If you agree with me that the signals that you need to get about all of your ad platforms are, when I spend, what’s the impact on the price of the media, and what’s the saturation, what does the quality of the media, how does that taper off over time, or we sometimes think of that as a saturation or plateau or fatigue, we have two methods that we’ve developed in order to get to that.
[0:23 – 1:24] The Agile Mix Model
One is looking at all of your historical data simultaneously, and what we refer to as an Agile Mix Model. Some people think of it as a next generation marketing mix model, some people think of it as a high frequency marketing mix model. It’s becoming a best practice in the industry, and we’ve helped revolutionize and pioneer this form of mixed model. It’s a two-stage model. It looks at what’s the impact of spend on price. A great example is search: the deeper you spend into an auction, the higher the CPCs will be, and everyone knows that, but we need to know to what degree, it can’t be directional anymore. We also need to know the second element, which is, as I spend in any of these ad platforms, what happens in terms of saturation. They are ultimately going to run out of their best prospects; they don’t have an infinite list of people that want to buy from you, and that’s why you’re buying media from them, is their ability to find those people, don’t serve ads to someone who’s never going to buy my product, serve ads to people that are likely to buy my product, and they do such a great job of that. The mixed models are designed to tease those two signals out of your data.
[1:24 – 1:56] The Limitations of Historical Data
But ultimately, what marketers are faced with is, sometimes the historical data that’s going into a model like that can’t get you there, it’s low signal data. A great example of that: we saw a customer the other day that had spent $300 a day on Snapchat, and after a few months they bumped that to $375 a day, and they did that for the next two months. That’s a very flat line; there’s no signal in that data for anybody’s model to work with. I challenge anyone building a model to say they can get an answer from that.
[1:56 – 2:50] Experimentation and Match Market Tests
And what are you going to do? You can kind of shrug your shoulders, or you can resolve to make better data. Our approach to making better data is to run experiments, sometimes people call them randomized control trials, in our case they’re geo-based experiments, and more specifically, they’re match market tests. And that does not imply you pick one market and another market; you have to pick a basket of markets on one side and a basket of markets on the other, and we have algorithms to do that. Now you’re making data, you’re going to control the environment, we’re going to fluctuate the spend, we’re going to take spend away from some markets, and we really build a pile of evidence, a mountain of evidence, to figure out what is happening to price when I spend on this media, and what is happening to saturation when I spend on this media. And what we learn from that experiment, we route back into our model.
Key Lessons: Combining Agile Marketing Mix Modeling With Match Market Testing
Data quality: Historical data can often be too flat or static to provide valid, actionable modeling insights.
Strategic experimentation: Running match market tests allows marketers to actively control environments and generate the evidence needed to understand platform performance.
Performance signals: Successful media planning requires understanding the specific degree to which spending impacts CPCs and audience saturation levels.
Integrated modeling: Learning derived from controlled experiments should be fed back into agile mix models to refine future media effectiveness strategies.
