What you need to know:

• Designer retail brand, Michael Kors, is one of five businesses to have participated in pilot for Adobe’s Mix Modeller solution, a combined marketing mix modelling (MMM) and multitouch attribution (MTA) tool.
• Having cobbled together a suite of solid but limited manual marketing mix modelling measures and a reliance on last touch attribution, Michael Kors global VP of analytics, Manuel Neto, was keen to try out a tech that could unify and unlock insights and importantly, empower teams to drive better business results. And he believes he’s found it in the combined marketing mix modelling and multi-touch attribution approach and “atomic-level insights” promised through the Adobe platform.
• As well as insights to inform channel mix, Michael Kors has been able to validate existing as well as ‘what if’ scenarios in minutes, opening up out-of-the-box thinking and stronger marketing business cases.
• Yet even with the best tech in place, marketers are going to need to be culturally brave if they want to truly embrace marketing modelling and enact a data culture, Neto says – this means big changes to the way they work, plan and demonstrate business impact.
• According to Adobe director product marketing, Experience Intelligence, Lily Chiu-Watson, one of the big reasons why Adobe is pursuing a joint MMM and MTA approach is because of fragmentation in the kinds of data customers can even access anymore and its granularity. Cookie deprecation, walled gardens and ever-tighter privacy regulations are all making this even more pronounced, she says – and marketers are going to need to find other ways to demonstrate the effectiveness of their marketing programs across channels to truly demonstrate impact.

Slashing the time it takes to action insights by half, creating models in minutes, plus unlocking – and validating – marketing scenarios that didn’t even exist to innovate programming, are just three ways a marketing mix modelling (MMM) pilot is helping retailer, Michael Kors, rethink its marketing approach.

The designer retailer, part of the Capri Holdings Group, has been one of five brands participating in a pilot program for Adobe Mix Modeller, trialling the solution as well as testing new features in beta. First announced in 2022, Mix Modeller is Adobe’s answer to two measurement problems marketers continue to grapple with: Multi-touch attribution; and better gauging the impact of their marketing mix across paid, owned and earned channels through AI-driven mix modelling.

Michael Kors already used Adobe Analytics as its main source of analytics truth. Historically, it’s operated a combination of custom models and last-touch attribution to plan and understand impact of its marketing spend. It’s an attribution approach VP of global analytics, Manuel Neto, readily admits is “outdated for what we want to do and where we want to go”.

“It was super exciting: To come along, look at where we are, then start applying some of the magic MMM is bringing to life to enable us to get to the next level,” Neto told Mi3 during an interview at the recent Adobe Summit in the US.

“We did have a lot of validation elements in place internally, because what we had done in the absence of existing technology or tools to give ourselves insights was create stop-gap solutions.”

Yet even with marketing and insights working in concert, things were fragmented. Enter Adobe Mix Modeller, which initially ran as a four-month pilot across the Michael Kors brand in the US. The retailer, which has three brands and operates across 56 countries, is now running an extended trial to the end of April, with discussions underway to potentially roll out Mix Modeller solution long term.  

Neto said “all the right stakeholders” from other Capri brands have been part of understanding its parameters as well as readouts and insights it’s showcasing. He’s confident the solution can scale and be replicated across the portfolio.

“It’s been one of the most transformational, reassuring journeys we have been on. It’s transformational from the perspective of unlocking insights,” Neto said. “We knew by moving methodologies and evolving measurement internally that we would receive new discoveries and insights. But the atomic level MMM has unlocked for us was a whole new game. It’s beating expectations and now we’re in the process of how we apply this for a longer journey and change from pilot to a longer-term partnership.”

Why Mix Modeller

Neto pointed to distinct localised marketing approaches for Capri’s three brands across all markets as one of the reasons why the group needed to find a better way of planning and measuring marketing programs.  

“The way marketing is often structured is touching all of the funnel – from an upper, middle and lower perspective – and with the aim of creating omnichannel experiences. So in addition to having a phone, app or desktop experience, it’s also translating into experiences cascading as a common thread to the store and offline level. In our group, there’s a big focus on omni going across every channel you can imagine,” he said.  

“Atomic-level insights are enabling us to understand the best mix of those channels and strategies that will enable us to reach a better outcome. And if we have an existing plan in place, help validate that plan and how we would go against baseline.”

So far, Mix Modeller has backed up several elements that, in the absence of a unified tool generating insights across paid, owned and earned channels, were generated through the team’s more manual measurement and test-and-learn processes.

“But what it’s also done is unlock something we’re not capable of doing,” Neto said. The big one for Michael Kors is scenario planning.

“The ability to validate scenarios we have already done was very good, but it’s humans doing this rather than technology amplifying that.” From being able to look at two scenarios over a 40-hour period, the team found itself able to explore 100 scenarios in 20 minutes.

It’s also created capacity for testing scenarios that didn’t exist. “For example, if we received X amount of additional money today, the scenario planning showed us what the universe could look like with that. That insight is incredible because it could enable us to negotiate further funding – it makes for a very strong business case,” Neto said. “It allows you to say: Here’s what else we could do if marketing were to be given this extra money. We’re sharing enough to make that impact understandable.”

Then there’s what Neto described as exploring “breaking the model” approaches to marketing planning.

“Say marketers have already been focused on getting customers to buy a product in blue for the last 10 years, but I want to try out yellow. This [modelling] allowed us to show the actual results from that yellow product you tried out through A/B testing from a brand perspective,” he said.  

“In addition, it helps validate the channels or things that often don’t receive the credit they deserve and show the credit given. It’s a powerful tool to say hang on, the thing we did as an isolated test and a ‘thinking out of the box scenario’ should be part of either our seasonal routine or ongoing routine, because it’s yielded results. Here are the numbers to show that. It means an innovation can become part of our strategy going forward, not just a ‘should we do this’ scenario.”

Automation driving empowerment of actionability – and change

In the absence of global governance of an MTA or MMM, many marketing and insights teams are partnering with agencies and groups to find solutions. That inevitably leads to fragmented information.

“You have group A running a particular approach with a group which makes sense for them; then you have group B who is saying ‘well I did it with my partner in a way that make sense for me’. There are great outputs but they’re in isolation. What happens frequently is when you try to combine those, they are conflicting. By nature, they were set up that way,” Neto said.  

“Something else we’re looking forward to doing with this technology is creating a ‘single source of truth’. For me, that has to be part of any data culture. Because if you want to drive actionability, and you’re telling people I want everyone to have access to information to drive empowerment, you also have to be careful data isn’t becoming a weapon. You don’t want to accidentally give out data that’s wrong or conflicting. It becomes a whole other universe of problems.”

Yet even with the best tools in the world, Neto said nothing can really change unless teams are brave enough to have challenging conversations to ensure insights are enacted for business benefit.

“I’m a huge believer in driving a data culture and empowerment of actionability. In essence what that means is you have this big discovery and new ‘insight’ story, but it’s only as powerful and valuable as your ability to put it forward and drive change and actionability,” Neto said. “Often, that actionability and change faces resistance. If the discovery is that powerful, it will be followed by change."

But people don't always welcome change, he acknowledged. 

“Because depending on the level of change, you may need to restructure an entire process. The example I gave is merchandisers are buying a blue bag, yet my data and mix modelling is saying consumers actually want to choose the yellow. I know internally it could take six months to change the entire process of buying a bag – I need to talk to planning about buying new products, talk to product development, update marketing materials.

“Imagine adding in the extra layer of so many months to get that insight itself – that’s bananas…  This fixes the first lag, which is six months getting from tech insight to action. But it also enables us to move faster when we have the insight. That’s the biggest challenge today – people and processes. This tech enables us to be faster at least 50 per cent or more of the time, which is huge.”

Given Michael Kors didn’t have any monetary investment in the Adobe Mix Modeller pilot, Neto couldn’t give an ROI in dollar terms. But there is no doubt in his mind it’s generated “immeasurable amounts of efficiency”. From creating models that can take up to five days, and up to three weeks of lag time, Mix Modeller is cutting modelling processes down to minutes or days.

“The smarter we get, the quicker we can output those models. It’s cumulative and grows as we grow,” he said. “It's already undeniable from an efficiency perspective. And for me, efficiency is money. If you’re able to act faster you’re able to get competitive advantage faster. As a consequence, KPIs will improve. And you’re going to fail faster and learn faster too.”

Tackling cookie deprecation with Marketing Mix Modelling and MTA

The origins of Adobe Mix Modelling are within the Adobe marketing team itself, with underlying models used by the internal team for the last five years. According to Adobe director product marketing, Experience Intelligence, Lily Chiu-Watson, improvements came not only through seeing the ROI of spend being investing and adjusting allocation based on the models; but also the ability to bring together Marketing Mix Modelling (MMM) and multi-touch attribution (MTA).

As Chiu-Watson explained it, the platform’s MMM and MTA capabilities have bi-directional learning, ensuring results are consistent. Mix Modeller has also been built on the foundations of the Adobe Experience Platform and the vendor’s AI underlying engine, Sensei.

“We have talked to so many companies where they’re using 2-5 measurement solutions. The MTA tells them one thing, MMM tells them that, and they sometimes don’t agree, so they end up tweaking them to get them closer. It’s funny – it’s like you’re trying to be so data-driven but in the last mile, you’re facing a gap,” Chiu-Watson told Mi3.  

The other reason why bringing MMM and MTA together is important – and what’s driving all this interest – is the perfect storm of cookie deprecation and stronger privacy measures across different regions.

“We’re seeing a lot of fragmentation in the kinds of data customers can even access anymore and its granularity,” Chiu-Watson said. “If you’re in your own channels, you still know who is doing what and when; but if you’re talking about the walled gardens, you don’t have that, and you’ll never get it again. You can spend a lot of time and money trying to guess where people are, then run MTA on top of that once you have gone through that time-consuming and expensive exercise. And you might be right, or you might not.

“The process we are taking is to say, give us the data to the best granularity you have – and with walled gardens, that’s going to be summary level data; tell us how much you spent and what campaigns you spent it on; tell us the impressions and clicks the publishers are reporting. But importantly, we’re not just taking their conversion and revenue numbers, we’re taking it from the source of truth, which is often Adobe Analytics or Customer Journey Analytics, which conveniently, we already have.”

This means the solution can combine all channels, offline and online, Chiu-Watson said, as well as data along the time series. While it’s being sold as a standalone solution (models come in packs of 10), customers already using other AEP apps, such as Adobe Analytics or Journey Optimiser, will have interconnected data capability and sharing across products.

“Having Mix Modeller on AEP alongside Analytics, Journey Optimiser and Real-Time CDP means there’s lots of compounding value,” Chiu-Watson claimed. “At Adobe, we definitely get questions like: You’re helping me understand incremental value of the channels I invested in – great, I really need that to talk to my CFO. But how do I now tactically understand which campaigns were most effective?

“Mix Modeller and the data this ran on is the data you have in Customer Journey Analytics; that’s the perfect exploratory tool to really get into the weeds of whether a campaign was successful or not, and where the drop-off happened in this part of the journey. Then, you go into Optimiser and create a new journey to fix this problem. Then in Mix Modeller, you can see the incremental value of that channel go up.”

For Chiu-Watson, the key imperative is getting marketers onto a better, more regular cadence of marketing planning and reporting than once-a year or biannual marketing schedules. And there’s no doubt being able to quantify the value of marketing is a question being raised more often. For Chiu-Watson, Adobe’s Mix Modelling solution is “a pragmatic approach” to the task at hand – and one made more affordable and accessible thanks to AI and the tech and data sets in play today.

“In Mix Modeller, a marketer or data scientist or someone in between can through a user interface, create a model, incorporate priors, assign a numeric of confidence they want to apply to the model, they can bring in internal and external business factors, save the model, then in a few hours, get the insights. Then on top of that you start scenario planning and choose the model you created,” she said.  

But even with the modelling available today, marketing remains an art and a science in Chiu-Watson’s book.

“Even if we tell you this channel has really underperformed for you, it doesn’t mean the answer is to just stop investing in it. Maybe, you need to examine what you’re doing there and uncover the areas where you can increase the value of that channel – perhaps your approach is wrong,” she said.  

“With Mix Modeller, we’re trying to provide the best science. But there’s this whole layer of art too.”