8 insights on media mix modeling for mobile user acquisition from Meta, Tinuiti, Rocketship HQ, and Singular

By John Koetsier October 8, 2022
  • Where MMM fits and works best
  • First-party data and media mix modeling
  • How to know your MMM is actually working
  • Media mix modeling as part of a hybrid measurement strategy
  • How do to incrementality without pausing campaigns

What can you learn about media mix modeling from an hour with the absolute experts in the trade? Far too much to fit into one blog post, unfortunately.

We recently asked experts from Meta, Tinuiti, and RocketShip HQ to join Singular CTO Eran Friedman in a discussion on how MMM can reasonably be expected to impact the high-pace, granular, data-driven world of mobile user acquisition. To catch that incredible show, check it out now, right here.

Those experts include:

  • Liz Emery, VP, Mobile + Ad Tech, Tinuiti
  • Shamanth Rao, Founder & CEO, RocketShip HQ
  • Igor Skokan, Director of Marketing Science, Meta  
  • Eran Friedman, CTO & Co-founder, Singular

Here’s a sampling of their insights …

1. Media mix modeling is also marketing mix modeling

The first thing to understand about media mix modeling for mobile user acquisition is that it’s also marketing mix modeling. Sure, it’s about channels and partners and advertising effectiveness. And as such, Meta director of marketing science Igor Skokan calls it the “gold standard” for ensuring brands can measure campaign and marketing performance.

But it’s also more than that.

“We also sometimes refer to this as a marketing mix modeling, not just media mix, because MMM itself actually is much more powerful … it’s more like holistic business modeling, so it goes way beyond media, even way beyond marketing into business drivers, price, distribution, other things, even whether all kinds of things that impact business … should be included in the modeling.”

Igor Skokan, Director of Marketing Science, Meta

That makes sense, and it’s worth remembering. 

The bonus: additional insight beyond simply marketing.

2. Yes, MMM can achieve high levels of accuracy

The second thing to understand is that media mix modeling can actually be quite effective. 

The knock on MMM has always been that it provides high-level general feedback only. There’s some truth to that, for sure. But in the right hands and used correctly, media mix modeling can provide a fairly high level of accuracy. 

“It’s very much a function of the volume of data but I would say even for smaller clients we work with, we see R-squared of like 90%-93%. But that’s the minimum we try to look at. So, you know, definitely, I would say that we do see very strong accuracy that informs a lot of our decisions.”

Shamanth Rao, Founder and CEO, RocketShip HQ

R-squared is data science for, essentially, real

As in, the percentage likelihood that the good things you want are being created by the expensive things you’re doing and not by competitive forces, organic factors, or the pink pixie dust spread in the air by magic marketing elves. 

Simpler yet: your advertising is working.

“Using MMM? The best fit is that you can use your past marketing performance to influence your future ROI. Not only that, but you can do that by optimizing your allocation by channel, by tactic. And that’s not just digital landscape, that’s traditional as well. So, that’s that full spectrum influence in terms of optimizing, promotions, pricing, competitor spend, you just have so much insight at your disposal when you do your MMM.”

Liz Emery, VP Mobile + Ad Tech, Tinuiti

As we see mobile app advertising ad spend proliferate beyond just in-app to in-app, that’s not only helpful, it’s essential.

3. But, media mix modeling is strategic, not tactical

The third thing to understand about media mix modeling is that it is strategic, not tactical. 

It’s crucial to understand the value and use cases for every measurement methodology in a hybrid measurement world, and that’s doubly true for MMM.

“Part of the challenge with MMM is with very, very granular and very tactical kind of decision-making. It’s difficult to get insights, for example, at the creative level ROI … of course, if you have a lot of data, if you’re running creatives at scale … then you can optimize things. But generally speaking, usually, it would be easier to seek a solution from the other methods such as deterministic attribution versus MMM for the operational or daily stuff.”

Eran Friedman, CTO & Co-founder, Singular

In other words, you’re not using media mix modeling to optimize adsets or creative or fine-grained targeting. You are using MMM to optimize channel and partner mix, or media intensity.

But when doing so, you get things you couldn’t measure before.

“I think it starts to become really interesting once you start digging into the other things that MMM can give us. Competitive intel, insights on game launches, cannibalization from within portfolio, things that we couldn’t probably measure before …”

Igor Skokan, Director of Marketing Science, Meta

Interesting indeed!

4. MMM is part of an overall hybrid measurement strategy

MMM might be a shiny new thing right now thanks to privacy and the drop in granular attribution data  (even though it’s 60 years old). But it works best paired with multiple methodologies, each being used where it makes the most sense.

“MMM is not a silver bullet. There is no silver bullet. We believe that advertisers need a mix of complementary solutions and that is attribution, experimentation, and modeling mix of approaches and every company needs to figure out what works for them. There is not even a single template or a right balance between these methodologies.”

Igor Skokan, Director of Marketing Science, Meta

I love silver bullets.

Everyone loves silver bullets.

In fact, everyone wants silver bullets.

The fact is, however, that we live in the real world, not the fantasy world. And in the real world, problems are multi-dimensional, circumstances are complex, outcomes are overdetermined, and solutions need to take into account multiple situational intricacies. (This is one reason data scientists are sometimes called data janitors.)

Ultimately, the only person selling one-size-fits-all solutions is the snake oil vendor. Simple answers are great, but they’re incomplete.

“There’s SKAdNetwork and GAID [and Privacy Sandbox on Android] and device-level data and incrementality testing and MMM … instead of relying on a single method to base all your work on, in our view, the future is going to have to require multiple methodologies and it’s really the combination of these solutions that’s going to provide the most actionable and insightful measurements.”

Eran Friedman, CTO & Co-founder, Singular

That’s a more challenging world than one in which you could just rely on freely-available IDFAs and GAIDs. 

But ultimately, it’s likely to be a better, more effective marketing measurement world, too.

5. MMM can save you money

It’s possible that everything that makes sense individually under the microscope doesn’t make sense in aggregate, from the 30,000 foot level. 

And that can include mobile user acquisition.

“You’re just basically trusting Google-reported numbers on iOS [and] the numbers looked wildly optimistic, the numbers looked crazy, because Firebase … [is] the referee and the player … when we did cut the budget, we just saw no change in the overall trials baseline. We’re like, ‘Guys, we just cut your budget in Google 66% because analysis said Google wasn’t driving anything incremental. Your trials just haven’t changed, nothing more.’ And that was, I think, just a huge, huge win.”

Shamanth Rao, Founder and CEO, RocketShip HQ

Saving money on ads that don’t provide impact?

Absolutely a huge win.

And it’s not just about saving money. It’s about optimizing allocation.

“We were able to leverage a combination of an MMM that we have built with them out of Tinuity, some incrementality testing, and projected models … and as we implemented these models, we were able to …  get a much higher return, we were able to actually prove out which channels worked, and we were also able to take into account environmental factors.”

Liz Emery, VP Mobile + Ad Tech, Tinuiti

Thinking you know what’s working is dangerous. Knowing you know: that’s power.

6. First-party data is critical to media mix modeling success

We’ve already seen the power of first-party data in the ongoing privacy revolution. It’s clear that first-party data is also critical for media mix modeling success.

“First-party data is critical to ensure the accuracy of your model. You can use that for so many things, including for tests for ground control, and making sure that your MMM is as accurate as possible … as always, first-party data is gold, right? The more we can have it, the better it is for everything basically.”

Eran Friedman, CTO & Co-founder, Singular

What that means is when you see actual use in an app like conversions, purchases, activity, and engagement, you can correlate that to marketing activity. 

Connecting this data to your MMM model ties the probabilistic projections of the modeling to the deterministic reality of what you’re actually seeing in-app.

7. How to know for sure that your media mix modeling is working

Marketing results aren’t just an input into your modeling. They’re also a way to ensure that your models are correct, accurate, and — the critical part — predictive.

“Whether you do [media mix modeling] in-house or through a partner, it’s really important to put actionability at the core. So request forecasting and simulations to be done to verify the accuracy of models over time because only that’s how you’re going to find that it actually works.”

Igor Skokan, Director of Marketing Science, Meta

It’s simple.

If your marketing measurement methodology can make predictions about the future that have truthfulness and accuracy to them, and if you can test those predictions and see if they actually pan out … then you know you have something that works.

Make predictions with your MMM models. Then test them.

8. Incrementality in an MMM world doesn’t mean stopping everything else

One of the challenges in my mind around measuring for incrementality has been that doing so in a purely clinical and completely scientific gold-standard way is sometimes interpreted as “pause everything else, only do X, and we’ll see what results X generates.”

There’s some sense to that of course, but there’s also a huge disconnect with marketing and business reality, where it’s extremely difficult to just push a big pause button and potentially negatively impact your next quarter results. Doing so has potentially huge impacts for CEOs and CMOs and growth leaders and, in fact, entire companies and their investors.

That, however, is not necessary:

“You need variation in the data to measure it. MMM is a backward-looking analysis, so as long as there is enough variation in the channels, then you are able to get a read on them … you could, for example, switch off a channel for a couple of days, or increase intensity somewhere and decrease intensity somewhere else, and then introduce variability into your factors so you can actually measure them … you can do regions … or different gains within the portfolio.”

Igor Skokan, Director of Marketing Science, Meta

That’s reassuring.

And it’s a reminder to not keep campaigns on autopilot. Adjust intensity, wind up and wind down, change targeting a bit, assign certain creative and certain offers to a specific channel or partner, and then switch things up a few weeks later.

Summing up: you really need to watch the entire show

There’s a lot of gold in this chat with some of the best minds available today on media mix modeling for mobile user acquisition. I get it: it will consume an hour of your time. Trust me, it’s worth it.

Check out the entire chat here.

Stay up to date on the latest happenings in digital marketing

Simply send us your email and you’re in! We promise not to spam you.