< Back to articles

Podcast: How MMM Can Create a Holistic View of Your Marketing

Listen to the podcast to find out how to take the first steps towards an automated MMM strategy that can transform your marketing campaigns.

In this episode of the Undiscovered Metric podcast by Adverity, host JJ Haigh welcomes Arno Witte, Senior Vice President of Data Science, and Sven Meijer, CEO of Objective Platform, to delve into the critical importance of a holistic approach to data in marketing.  

Both guests bring a wealth of experience, with Arno's background in econometrics and model development, and Sven's extensive history in big data and commercial strategy, setting the stage for an insightful discussion on leveraging comprehensive data analysis to drive marketing success. As leaders at Objective Platform, they share their perspectives on the evolving landscape of marketing measurement, the integration of brand and performance KPIs, and the power of predictive analytics to shape future strategies.  

Join us for a deep dive into how Objective Platform is revolutionising client outcomes through data-driven insights, the challenges and solutions in naming conventions across platforms, and the future trends in marketing that professionals should prepare for. Plus, Arno and Sven offer valuable advice for those embarking on a career in data analytics, emphasising the intersection of technical skill and business acumen in achieving industry-leading results.

Topics covered in this podcast are:

·       Why it's so essential to have the data immediately at your fingertips?

·       Why does it seem that people are leaning towards automated MMM more so than MTA now?

·       What trends should marketers and advertisers really be focusing on over the next 12 to 18 months?

Listen to the podcast to find out how to take the first steps towards an automated MMM strategy that can transform your marketing campaigns. The transcribed version is available below.

JJ Haigh: Welcome to the latest episode of the Undiscovered Metric with me, JJ Haigh. Today, I'm joined by Arno Witte, Senior Vice President of Data Science at Objective Platform, and Sven Meijer, CEO at Objective Platform, to discuss why having a holistic view of data is so important. Firstly, thank you both for joining me. Just before we get started, could you please provide a short background on your role and career history?

Arno Witte: My name is Arno Witte; I have a background in Econometrics. After that, I started working as a consultant. I had a lot of contact with clients, and at Objective Platform, I was responsible for creating the models we have, developing them, and developing the platform to make sure that these models are usable in practice.

JJ Haigh: Brilliant, it sounds like you're going to have a lot of great insights for us on the data side later in the conversation. And Sven, if I could just ask the same of you.

Sven Meijer: Yes, thank you. I'm Sven, and I'm the CEO of Objective Platform, having been with O|P for many years. I have extensive experience in Big Data and have been working in this industry for a long time. I've been leading O|P as CEO for almost two years now. My background has always been in more commercial roles, assisting clients in getting the most out of the platform by really utilising it. That's also my main focus as CEO: to ensure our clients are very satisfied with the use of analytics on a daily basis.

JJ Haigh: Fantastic, thank you both so much for joining us today. Sven, I'm going to start with you. Would you be happy to provide a bit of an introduction to O|P? What are the goals, how are you working with clients, and what are the objectives that help you achieve those goals?  

Sven Meijer: Yes, we are a marketing measurement company. We utilise all the client's data to create models and provide them with insights into those models. We believe in full-funnel measurement, not just modelling the performance KPIs but also incorporating the brand KPIs to give insights into how the brand impacts performance KPIs. For us, that's the essence of the full-funnel approach. What we do is automate this process; instead of doing this once, we update models for our clients every month. This helps them to not only learn once about what the media or marketing achieved in the last campaign but also to constantly update those models and learn about changes, new things, and to test and experiment with new strategies. A major part of O|P is also the tooling we provide, making historic data predictive for the future. This is crucial for predicting future outcomes and optimizing strategies. That's what we do for our clients, and it's actually the main thing that Objective Platform is doing.

JJ Haigh: Great, so considering the comprehensive analysis of historical data, predictive data, and those models, why would you say that having a holistic view of data is so important? Furthermore, how do you translate that into a tangible impact for clients?

Sven Meijer: Yes, "holistic" can have different meanings. For us, holistic first means including all media channels, not just focusing on online channels but also incorporating TV, radio, and out-of-home advertising, so there are no separate silos. Additionally, for us, being holistic involves integrating brand KPIs alongside performance metrics. This approach is strategic as well. It means not just looking at online conversions but understanding the entire picture—what online efforts are achieving, how physical stores are performing, and the overall health of your brand. This holistic view provides insights into significant strategic questions, benefiting senior marketing personnel and those responsible for individual channels. At Objective Platform, we offer multiple insights dashboards where clients can access their specific data. For example, they can see how their Facebook channel is performing and compare it with TV advertising.

JJ Haigh: Great, and I'd like to bring Arno into the conversation a bit here as well. From a data scientist's perspective, what does a typical day look like for you when you have this holistic overview, with access to various data channels and insights? How do you then translate those for clients and really help them achieve their brand KPIs and goals?

Arno Witte: Yes, I believe the most critical aspect is our extensive experience in modelling and our deep knowledge of best practices in this field. However, we recognise that our clients possess more in-depth knowledge of their own businesses, so collaboration is crucial. This collaboration begins when we start working with a client, where we first develop a blueprint to identify the most critical KPIs, understand the media strategies they have previously implemented, and learn about their plans for the upcoming year. This serves as our starting point. We then proceed to model the KPIs of their interest, taking into account all media channels and possibly other influencing factors such as pricing. Once this foundation is set, two main actions occur. We update the models very frequently, as soon as new data becomes available to our system. This enables us to assist the client effectively in applying these models in practice. I have observed numerous projects where clients had sophisticated media mix models but failed to make improved decisions based on those models. Thus, our initial focus is heavily on getting the models right, earning the client's trust, and then actively using those models in practice to inform decision-making. If decisions are not made based on these insights, we believe we are not delivering the value we are capable of.

JJ Haigh: And with all these models that you're creating, is that why it's so essential to have the data immediately available at your fingertips?

Arno Witte: Yes, exactly. For our clients, it's crucial that the model is trusted first and foremost, so there's a significant emphasis on ensuring trustworthiness. After establishing trust, it's vital that the model remains up to date. Often, the most recent insights are the most critical for planning the next campaign. It's important for our clients that the results are delivered quickly, ensuring that decisions for upcoming campaigns are not based on outdated models from the previous year.

JJ Haigh: And Sven briefly mentioned Facebook and Meta, and I know you work closely with your clients. How would you say a platform like Meta contributes to a client's marketing mix?

Arno Witte: That's an intriguing question because the impact of Meta greatly depends on the goals a client aims to achieve with it. We've observed various objectives that clients have with Meta. For example, Meta can play a significant role in brand building, as some of our clients have successfully used it for this purpose, but it can also be a driver for performance. Depending on the objectives and how the channel is utilised, the impacts can vary. However, a crucial aspect for us is our ability to measure the impact of a campaign at all stages. This means we can determine whether a recent campaign increased traffic to a website or led to conversions for the advertised products.

JJ Haigh: Are there any specific metrics or insights that you believe clients should focus on, or that your teams particularly emphasise when integrating this information into the holistic view of data for your modelling or any other purposes?

Arno Witte: I believe for a holistic view, it's essential that the channels are comparable. You can, of course, examine metrics like impressions, reach, or clicks from your ads. However, more importantly, it's about having the correct mappings and conventions between different channels. This involves identifying the types of campaigns conducted, whether they were performance-based, focused on brand building, or product communication campaigns. By consistently identifying these aspects across all channels, including Meta, you make it significantly more manageable to compare them when making holistic media decisions. I always recommend starting with this approach and then, if necessary, delving into more detailed information such as targeted devices, specific target audiences, or particular impressions.

JJ Haigh: It seems naming conventions are a recurring topic throughout this series and with many people we speak to. With all these different platforms having slightly different naming practices, for example, reach may have different meanings in Google compared to Meta. Is this a daily challenge, and how do you overcome it when creating these models?

Arno Witte: Yes, this is a challenge our clients face, and it was a challenge for us too, especially when we began modelling and needed to update these models quickly; it quickly became a bottleneck. There are two approaches to address this. One is to establish conventions, so everyone internally follows them, but we know it's nearly impossible to ensure everyone adheres to these recommendations. Some clients manage this better than others. What we can do is facilitate the mapping of data, making it straightforward. We allow all input data to maintain its original format and conventions, and we map them to a unified convention used in our models. We provide our clients with mapping tools that enable them to reassign data to the conventions we use in the models. Streamlining this process, I believe, is perhaps the biggest time saver in such procedures.

JJ Haigh: And do you think that as the number of platforms increases—our research last year showed that CMOs are now operating with around 10 different data sources, some up to 14, compared to about 6 pre-COVID-19—is embarking on this journey of establishing naming conventions and mapping from the start essential, or can they retrofit these conventions and mappings later on?

Arno Witte: Of course, it's always possible to implement these practices later on, but it's generally better to do so from the beginning. Ensuring that everyone is aware of and adheres to the conventions from the start is ideal. However, what we've found in practice is that having mapping tools in place not only makes it easier to align data but also helps in identifying outliers that don't fit the mapping rules. This allows us to promptly provide feedback to those responsible for tagging or establishing conventions. Incorporating this feedback loop can save a considerable amount of work in the future. While it's always feasible to map data retroactively, doing so from the outset is preferable to maintain timely and accurate models and data.

JJ Haigh: Great, and something I wanted to discuss with you, Sven, is the ongoing industry debate between marketing mix modelling (MMM) and multi-touch attribution (MTA). Where do you stand on this issue, and why does it seem that there's a trend towards favouring automated MMM over MTA now?

Sven Meijer: First of all, we do both. We have a strong background in MTA and still maintain a robust MTA model, and I firmly believe in the power of MTA for organisations. However, what we've seen with automated MMM is its ability to provide more strategic insights, which is crucial. For instance, it can help answer significant questions like whether to invest more in brand or performance, which is not only strategic but also vital. Automated MMM allows for testing campaigns and comparing different channels, such as Facebook and TV, both for sales KPIs and brand KPIs, which are key concerns for major advertisers. The choice between MTA and MMM depends on the specific business questions and the organisation's data maturity. With the decline of third-party cookie data, tracking display views is becoming more challenging, making automated MMM a more suitable option for measuring effectiveness. However, for organisations with high data maturity, MTA can still offer considerable value. We're proponents of automated MMM because it tends to be more accessible for marketers to work with, enabling them to achieve and demonstrate results more effectively.

JJ Haigh: And for a client or marketer looking to embark on creating an automated MMM model and utilising those insights, what would be the top three pieces of advice you'd offer them before even reaching the stage of delivery?

Sven Meijer: The first requirement is data, unsurprisingly. We need to ascertain if the necessary historical data is available. Secondly, it's crucial for them to define the business questions they aim to address clearly, as this clarity will enable them to measure the success of the MMM effectively. Lastly, it's important to have individuals within the organisation who can manage the work and utilise the insights. Having at least one ambassador who understands modelling and its application can make a significant difference. Such individuals are essential for leveraging the model's full potential and supporting its adoption within the organisation.

JJ Haigh: Regarding having a data and modelling champion within a company, would you say this indicates a high level of data maturity, or how can businesses develop this maturity if they recognise the need but feel they are not quite there yet?

Sven Meijer: For me, it's perhaps not always about data maturity, although Arno might have a different view on this. Rather, it's about how, once we have the model outcomes, we can apply insights across different CPOs, making them comparable, for instance. It also involves observing how CPOs develop over time and then diving deeper into various campaign types or campaigns. This is essentially about translating data into actionable knowledge for internal use. We run customer success programmes around this concept. The necessary knowledge isn't solely about understanding the data or someone who delves deeply into it, but rather about individuals who grasp how the model works and how to utilise it effectively. That's why we're big fans of automated MMM. With a one-off model, it's crucial to recognise that interpreting the results is just the beginning of the journey. It's about understanding what the results mean, how to test them, and how to implement them, requiring people who grasp these concepts. That's the type of knowledge, from my perspective, that's needed.

Arno Witte: Sven's point is crucial. Naturally, it's vital to have people who understand the data and ensure its accuracy, but more importantly, there's a need for readiness to act on the model's outcomes. As Sven mentioned, that's when the journey begins—translating model outcomes into actionable steps. For instance, when planning the next media scenario, it's about assessing its impact on the KPIs modelled—how it might enhance brand value, increase traffic, or boost sales. If these results aren't available or not presented in a format that facilitates decision-making, then adjustments are necessary. The tooling must align with the existing business processes. Changing how people work based on model results alone is challenging; the model must adapt to them. Over time, if a change is needed, it's feasible. However, the insights must be compatible with decision-making processes. If they're not in the right format, their impact will be less significant than it could be. That, I believe, is paramount.

JJ Haigh: Is it a common pitfall that organisations are not prepared to act on the insights they receive, or they don't integrate well into existing business processes? Or are people generally open once they have these insights, to actively utilise them and drive change?

Sven Meijer: I think sometimes people are apprehensive, worrying that introducing modelling will critique their past media investments. There's a tendency to view this negatively. However, that's not our approach at Objective Platform. We collaborate closely with clients to maximise outcomes. From my perspective, I don't regard these concerns as pitfalls for our clients. Those who begin using Objective Platform are typically well-prepared and understand what to expect. We support them from the start of the process. Yet, it's true that some advertisers may feel somewhat intimidated by the prospect, which is unwarranted because historical analyses always yield valuable learnings. Especially, I advocate for a test-and-learn approach with MMM. The pandemic presented a challenging test case but offered a fascinating perspective from a modelling standpoint. It led to significant changes, providing us with many insights that advertisers can now use to employ media more effectively.

JJ Haigh: Brilliant, and speaking of COVID-19, it seems we are transitioning out of that era of advertising, moving into the next phase with those learnings in mind. Over the next 12 to 18 months, are there any particular trends that you both think marketers and advertisers should focus on, any emerging trends that need immediate attention rather than a wait-and-see approach?

Arno Witte: Generally, we're observing platforms like Meta and Google increasingly focusing on MMM, which suggests that marketers will soon realise the necessity of MMM for making holistic channel decisions. This shift towards MMM is expected to accelerate. Additionally, as these techniques become more prevalent, it will make sense to also focus on upper-funnel KPIs, given the utility of these techniques in incorporating such metrics into actionable strategies. I foresee a promising future for automated MMM, which is why we have dedicated significant effort towards this area recently. The transition towards MMM is already underway, but I anticipate it will gain even more momentum shortly.

Sven Meijer: I believe branding is becoming increasingly important, even for smaller companies. It's not just about focusing on the low funnel but also steering the mid funnel. For large advertisers, brand development is crucial, but in some countries, TV inventory is saturated, making it challenging to invest further in TV for brand promotion. We've observed that some clients achieve significant success on platforms like Facebook and Instagram for brand driving. However, the question remains whether these platforms represent the future of brand building or if new platforms will emerge. In the context of MMM, there's a common perception that it simply provides a historical overview through a PowerPoint full of numbers. Yet, the predictive aspect is far more intriguing to advertisers and is a growing focus. Predictive analytics and AI, which we now employ to forecast and optimise media across channels for various KPIs, represent the future in our field, and it's an area we're investing in heavily.

JJ Haigh: Thank you for joining us. One final question for both of you: What advice would you pass on to someone starting their career in data analytics, or what advice would you have liked to give to your younger self?

Arno Witte: Should I begin? My advice is to start with the business question. Coming from an econometrics background, I've learned that starting from the business perspective, rather than the econometrics alone, leads to better models that are more aligned with business challenges. Understanding what the business anticipates for the future allows us as econometricians to model that reality more accurately. For those more technically inclined, I believe understanding Bayesian statistics is crucial. It's vital for modelling with varying degrees of certainty and integrating outcomes from experiments, such as blending MTA and MMM into a unified model. I wish we had embraced Bayesian statistics from the start.

Sven Meijer: From a less technical perspective, and considering myself still quite young, I'd say we're in a fantastic era. I'm passionate about this industry, combining data with marketing. Marketing itself is exciting, but using data to enhance marketing practices daily is truly rewarding. Having worked in this industry for over eight years, my advice to my younger self would be to enter this field directly, bypassing my prior involvement in the less stimulating financial sector. This industry is dynamic and continually evolving, making it an exciting career choice. For anyone looking for solid advice, especially on technical growth, listening to Arno would be beneficial. Learning Python, for instance, is invaluable for anyone in this field.

JJ Haigh: Brilliant, it's clear you're both very optimistic about the future of marketing, which is incredibly encouraging for everyone involved. Thank you both for your time and insights. Have a wonderful day!

Our Case Studies