A New Era Calls for New Marketing Measurement 

A faster Marketing Mix Modeling project turnaround emerged as a necessity following the way COVID-19 pandemic has shaken the measurement industry. Accelerated changes in consumer behavior and marketing spending became the new norm to be measured and optimized. According to Alex Owen, Global Head of Data & Analytics at Unilever, “COVID-19 has accelerated what could be a 10 years’ worth of change”. 

This unexpected change has enormously affected the MMM industry. The old standard of 12 to 16 weeks MMM project duration is no longer able to keep up with the accelerated changes. As such, a new measurement framework is needed, one that is based on Agile practices, project efficiency, and the full leverage of modern technologies.  

At MASS Analytics, we have identified 5 major areas in the MMM workflow that can be optimized to achieve a significantly shorter project turnaround. How short? These tried and validated techniques have been proven to decrease MMM project duration by as much as 50%, without a significant cost increase. Advertisers aiming to get more frequent insights from their Marketing Mix Modeling projects would find these five methods efficient and intuitive. 

The Five Keys to Shortening Marketing Mix Modeling Project Turnaround 

1. Streamline your Data Preparation for Better Efficiency 

Improving the efficiency of the data preparation phase is a must, especially that nowadays more and more frequent project updates or refreshes are needed. At the start of each project, the modeler has to juggle multiple data sources and hence multiple streams of data coming in different shapes and forms. That data then must be manually put in a format that is acceptable by the modeling software used. This pain-staking data preparation phase ends up wasting 30 to 40% of the project timeline. That percentage can be saved by having a process built to automate the data preparation phase. Using automation technology means that the next time a new stream of data is available, the predefined data-prep sequence is applied to seamlessly generate data ready for modeling. This is proven to be most valuable to companies seeking to create multiple Marketing Mix Models with minimal investment. Our Data preparation solution MassFeeds is a prime example of technology that would enable automated and scalable Data Preparation.  

2. Automatically Transform your Data to Proxy Real Consumer Behavior 

While Data Preparation involves getting through data and harmonizing it in the right format that will be ingested by the modeling software, Data transformation is something else. It is the replacement of a certain variable by a function of that variable that would make it more aligned with what is being modeled, i.e., real consumer behavior. It involves thinking about the variables collected and the way consumers have responded to the different stimuli to make sure that variables are transformed in a way that mimics how customers and consumers have responded to marketing activities.  

Data transformation is one of the key elements, organizations should seek to automate in order to achieve that ambition of shortening the project turnaround. If full automation is difficult to achieve, modelers should have the tool, or the technology needed to create thousands of transformed variables quickly and efficiently. 

3. Leverage Artificial Intelligence to Accelerate Modeling 

Manual selection of variables into Marketing Mix models is not by any means a process that can be scaled and performed indefinitely. The changing dynamics of Media and Marketing in general means we have to test tens of thousands of different variables to build a model as robust and accurate as possible. Such a level of sophistication makes it very hard to run MMM projects manually without sacrificing efficiency and time.  

Even though manual modeling allows the analyst creating the model to have full control over the inclusion or exclusion of specific variables, this approach remains time-consuming and requires high skilled modelers to manipulate it efficiently. Therefore, auto-modeling is the perfect alternative in terms of efficiency and time saved. 

Automatic modeling is generally based on genetic algorithms powered by AI that would test various combinations to provide the best combination of variables to build a model. The genetic algorithm reflects the process of natural selection in its workings. It selects the finest variables that would best fit the equation by optimizing a predefined objective function (e.g., a combination of standard error, R square, VIF…). 

4. Perform Modeling and Optimization Iteratively 

The juiciest part of any Marketing Mix Modeling project is the insight and recommendations delivery. Automated Modeling can be leveraged to create models that are sound statistically and commercially. However, these models need to be deployed to produce actionable insight on how to optimize media and marketing going forward.  

To achieve project efficiency and save weeks’ worth of work, it is essential that there is a direct link between modeling and optimization and that we do not consider modeling and optimization as a sequence, i.e., when modeling finishes optimization starts! The optimization process must be agile and iteratively performed over time as opposed to waiting until the modeling phase is over in order to begin optimization. This will ensure the best quality possible while slashing modeling phase duration for improved efficiency.  

5. Link Modeling to Reporting for Efficient Results Exploitation 

It is not uncommon for the data extraction and model reporting to take between one to two weeks from the time model is finalized to the time insights are ready to be presented. This bottleneck can be eliminated by leveraging technology. In fact, by streamlining MMM software results extraction with reporting tools such as Excel or Tableau, results are always linked and reflect changes in the model in real-time. This means that as soon as any changes are made to the model, reports and results will get updated automatically. That will save substantial resources and time while protecting against human error. 

Technology is the Future of Marketing Mix Modeling 

Marketing Mix Modeling is science and art. The ability to turn raw data into sound variables that actually capture how customers have responded to your different marketing activities, and then use these variables to create a model depicting the impact is nothing short of an elegant display of human ingenuity and imagination working hand in hand. By adding a layer of symbiotic reinforcement through technology, MMM can transition smoothly into a post-covid era bound by accelerated changes and a higher focus on Marketing ROI.