Three trends paving the way for the future of marketing mix modeling

For the last three years, we have been building models to help companies and our clients understand the ROI of their media, advertising, and other types of things that grow new types of investments.

We’ve developed several models, trying things as simple as regression and as sophisticated as Bayesian networks. It’s a very exciting time for us.

We see three main trends with our clients on a fairly global level.

At Artefact, we address this through our internalized approach, which may seem counterintuitive for a consulting company, but it aligns with how we assist clients in building this capability within their organizations.

A lot of our clients want to internalize MMM, but they need help to take this first step. Our solution is to work with business teams to figure out what the real learning agenda is: what do clients want to learn about their marketing or sales activities? That would be a first step. And as a second step, we’ve developed our own model-as-a-service that we’re going to put on different platforms and make available to our clients.

Our tech teams work hand in hand with our clients to implement our IP and bring their teams up to speed. Normally, we then leave our clients to continue their MM journey on their own, but if they want extra support, we of course provide it, that’s our philosophy.

There are two main types of capabilities clients want; the first is pragmatic: to get fast results and rapidly iterate with available data. The second is an extension of the first and involves innovative AI solutions that we have with the Bayesian network or that we can incorporate with GenAI to increase team efficiency.

To quickly get to this first ROI, we’ve developed a way of working with clients where we start by implementing the first pragmatic approach in the beginning so that we already enable the teams with these fast insights and then we continue iterating and here the in-housing really helps because this way we have really quick access to all of the tech stack; we have quick access to to the data and we can do this link between model updating, between updating the results, and between implementing them in the business really fast.

This way our clients get fast results, they still have their in-house capability and we have the time capacity, staffing and budget to implement great innovative things that are available now to us with GenAI.

We try to save time figuring out what’s the best model to use because we’ve tried it so many times, we know more or less what approaches work, so within your regular project timeline to connect strategy and operations, we need time to talk to people, we need to align the global learning agenda with the local learning agenda and one way or another. And then the rest of the time we spend with the business to make sure that the results that our MMMs will deliver align with other studies that the client has done, either with AB tests or with other MMMs, and we actually make that link and bring all the insights together, it’s more of an analytical and consultative role rather than bringing in a bunch of data scientists and building the AI from scratch.

I’ll give you an example from connecting the operational and the strategic department and decision making in the company.

We had a client – no names, sorry! – and this client was very far along in their measurement journey. They’d done a lot of studies, AB tests, marketing mix models on global and local levels. And another person within the organization invited Artefact to do a study on a slightly different scope. When we arrived, we discovered that there were a lot of studies that had been done in the past within almost the same or adjacent scopes, and we found a lot of value in reusing all of these insights because some of them were very strategic, charting the company’s path, and one of them was very operational from AB testing, trying a lot of different innovations. We really wanted to tie these insights together to present a holistic picture to the business that hadn’t been done before.

And because our model and our MMM approach is based on Bayesian methods that learn from the customer’s business context, we were able to succeed, and the first value-add of our model was “here’s the ROI of this, here’s the ROI of that,” but the second value-add was really unifying the entire business value chain, from operations to strategy, using AI.

True, but sometimes there can be challenges because there are different teams doing the measurement and AB testing and bringing it into the business. We were able to create a story that was shared by everyone, and that really helped speed up the business without having to deal with little inconsistencies here and there. When you have a holistic view of all the insights you’ve gathered over the last few years, it really solves the problem of operational communication.

I can tell you one thing: marketing mix modeling is here to stay. Companies want to see the results of their actions. Companies are doing a lot of initiatives and reorganizing their data lakes and the way data is consumed inside. Organizations are paying a lot of attention to data governance. I feel the future is in efficiency: marketing mix modeling will have to become faster, more rigorous, simpler to explain to business stakeholders, and more easily accessible to business analysts. And GenAI is important because it can enable clients to query MMM results and build reports by themselves without necessarily having a lot of technical skills, so marketing mix modeling is definitely not going anywhere: it’s going to stay and evolve to become more efficient and more easily accessible.