News

Performance Power Players: Gary Danks on the importance of Media Mix Models

Dec 13 2023 | Data Analytics Tech, Market Insights, Mobile

Gary Danks, General Manager, AIM by Kochava

Key Takeaways:

Media Mix Models provide a holistic view of media spend and enable marketers to understand the impact of various marketing channels on KPIs. MMM answers critical questions such as how to plan budgets depending on seasonality and where to focus budgets for maximum ROI.

Having a Media Mix Model in place in 2024 will be highly beneficial as user-level signals will be further reduced due to ongoing changes in privacy brought in by both Apple and Google.

Advertisers can start building Media Mix Models with campaign data via an MMP, plus internal sales and media spend numbers.

Why do Media Mix Models matter?

Media Mix Models (MMM) are a statistical method of measuring marketing mix impact and, more importantly, the contribution of each marketing channel to achieving KPIs.

Fundamentally, these models consider a range of marketing channels, from ad networks such as Google, Apple, Facebook, etc, to Digital Out of Home (DOOH), Connected TV, and offline sources such as Radio, TV, Press, etc. They then isolate the incremental benefit of each media channel or campaign, assisting marketers in understanding a range of questions such as;

What is the most effective channel mix for my brand?
How should external factors such as seasonality be considered when media planning?
What are the critical points in the year to invest media budgets?
Which channels offer the most robust ROI?
What media sources are most incremental for my brand?

Attribution has been the industry staple and accepted measurement method for many years. However, attribution only shows part of the story; last-touch attribution recognizes the last media channel the user saw before converting.

“I often use an analogy of someone who goes to the cinema and decides to see a film based on a film poster they see as they walk in through the door. The reality is that they are likely to have seen advertising for several films in many places over some time, for example, while watching TV, on billboards, on social media, etc.”
Gary Danks

He continues: ‘If a marketer measures the impact of media using last-touch attribution, the film poster is rewarded. All the other channels that have influenced the decision to go to the cinema are not considered. However, a Media Mix Model considers the impact of all the channels in the campaign and enables the marketer to understand the effect of all the channels involved when someone goes to see the film.”

Can MMM and Attribution work together?

Yes, both measurement techniques work well together. MMM gives a top-down view, providing aggregated strategic insights as an output that enables more informed decisions about the media investment. Last touch attribution gives a bottom-up view at a user level. Therefore, with both systems in place, marketers will be able to see different insights and have the opportunity to use other levers to optimize. For example, understanding the impact of seasonality.

Why are Media Mix Models an essential tool for mobile marketers?

“As a marketer, you are there to improve results and be a valuable asset to your business. In addition, championing Media Mix Models provides a skillset marketers should understand for personal development. Being the champion increases the likelihood of being the ‘go-to’ person, the expert in the company, which means you can build trust.” says Gary.

One of the advantages of MMM is that the model can measure offline media as well as online media, which is often a frustration for marketers. Marketers often know their offline media does something to ROAS but need help to measure it effectively.

Media Mix Models measure the incremental impact of a media channel, considering a base level of sales completed by customers thanks to factors such as brand loyalty, immediate need, convenience, etc.

Why the interest in MMM now?

There has been a noticeable shift in interest during 2023. Much of this is due to the changes in Privacy, with Apple’s ATT and Google’s Privacy Sandbox changing the measurement picture. Privacy changes were becoming apparent three or four years ago, and it was clear there would be a need for non-user-level measurement. At that time, MMM models were expensive and were too slow.

“Everyone has to take notice of these changes, and they are coming down the track fast,” says Gary. For example, fingerprinting will be banned in 2024, and brands not planning for these changes will be impacted.

The industry has been preparing for this for several years, and the models are available, so brands should start getting ready for measurement change, too.

“MMM not only future-proofs against these privacy changes as the models are not based on user-level data but also enriches available insights and provides more information for marketers to make better strategic decisions,” Gary says. He continues, “Previously, MMM was a tool for brand marketers; now it is a tool for performance marketers too.”

Can Media Mix Models be applied to any vertical?

Fundamentally, Media Mix Models can be a helpful measurement tool for all types of brands across many verticals. MMM works well for clients with ongoing media budgets, always trying to drive new sales or acquire new users. However, the models also benefit from times when there is variation in your media budget, as this helps the models isolate the incremental impact of the measured channels. Overall, Media Mix Models are well suited to many mobile-first companies.

What are the biggest challenges in setting up MMM?

MMM is a complex concept; until now, many marketers and businesses have relied upon attribution methods, so there is a shift in thinking and understanding when considering setting up MMM. Internal education of stakeholders may be needed to ensure the technique is understood and trusted.

Bear in mind that the numbers an MMM model shows can be different from last touch measurement, and many businesses have built Lifetime Value (LTV) models based on the last touch, so changing to an MMM model is something that needs to be thoroughly understood as there are implications across the business to take into account.

One of the challenges is the perception that MMM can be slow, expensive to implement, and complicated to use because data is static and quickly out of date. The technology is improving faster than it did in the past. However, it is still a complex method requiring stakeholder education, time to set up, and specific budget requirements to succeed.

How should brands get started?

Here is a good starter guide to understanding MMM; getting started is more accessible now that the tech is improving, especially for digital-first advertisers. Sales happen mainly online and with digital audiences, making them ideal for testing an MMM approach.

“No technical resource is required, no SDKs are involved, no need to move MMP, no need to swap tags, and no campaign disruption. So the level of involvement for the client is minimal,” says Gary.

Get your data ready. MMM works off a foundation of historical data, and the more, the better. While a minimum of 12-18 months of sales data will suffice, more robust models can be built if there are several years of data, as it allows you to understand the underlying seasonality. This data should be at a weekly or daily granularity. MMMs can measure the impact of media on any KPI, such as;

App-based installs
First purchase
7-day revenue events
Subscription event
Repeat purchase events

Alongside internal media and sales data, external historical data, such as weather, public holidays, sporting events, etc., is also ingested, which can impact sales on many brands.

“The results can be fascinating; for example, with one food delivery service client, the MMM showed that food deliveries built up through the week to Fridays, over the weekend, and paydays when people ordered significantly more takeaways. This enabled the brand to understand that achieving a sale is a lot harder on a Monday compared to payday. They could then decide whether to invest media budgets away from paydays or weekends and support other days because the service was top of mind anyway,” continued Gary.

What advice do you have for brands interested in MMM?

“Don’t run before you can walk. Start with digital sales and digital media data initially. This first step will build the concept of Media Mix Models as a measurement solution and ensure the business is comfortable with the results.”

Once the value of Media Mix Models has been established, other signals, such as PR, can be factored in. CTV is considered a digital channel and can easily be ingested into a Media Mix Model.

The results can be divisive and surprising, and it’s wise to manage expectations internally. An MMM will only improve results after some time; instead, think of it like a GPS, always seeking the best route.

First, you can build confidence in the model. The model will constantly predict what will happen in the next 2 weeks. For example, what installs, sales, repeat revenue, etc, can be expected? These predictions can be validated to build trust in the model. The next step is to make small changes, such as decreasing spending in one channel and increasing elsewhere rather than drastically altering the budget mix. Minor adjustments are suitable for the model and managing expectations internally.

About Gary Danks

In the dynamic digital marketing realm, Gary Danks has carved out a niche for himself, boasting over two decades of expertise in mobile marketing. His professional ascent commenced with pivotal leadership roles at ITV and YOC.

In 2015, Danks established the award-winning Machine Advertising, quickly gaining recognition for its unparalleled product excellence in ad fraud detection and Marketing Mix Modelling. Fast forward to 2023, and Machine Advertising caught the attention of the mobile measurement titan, Kochava, leading to its acquisition.

Gary holds the position of General Manager for AIM – The Next Generation MMM tool. In this capacity, he steers this groundbreaking initiative, ensuring its position at the forefront of the industry.

Read more about MMM from Dane Buchanan, Global Head of Data, Analytics, and Tech at M&C Saatchi Performance, here:

Media mix modeling: the future-proof, silo-busting future of measurement