It’s been a long time where advertisers have long used marketing mix modeling (MMM) to understand how marketing strategies affect revenues, and it has proven to be productive in delivering insightful insights into conventional media. However, for today's dynamic and ever-changing digital platforms, the tried-and-true method to modeling is not well-suited based on years of stable data.
Modern platforms such as Facebook's family of software and services (Facebook, Instagram, WhatsApp, etc.) have vivid and complex advertisement possibilities. Also, the applications are continually changing to keep up with customer behavioral change. Compared to existing traditional media outlets such as print and television, it’s not as easy as it seems to determine their true impact.
Although this uncertainty makes it more difficult to model the marketing mix, it is not impossible to make it work. In collaborating with partners, developers have noticed that if marketers improve their strategies and take these main approaches outlined below, the effect of digital platforms can be accurately captured.
#1 Adjusting Time Frames
Generally, marketing mix modeling has required constant observation over the course of many years. Given the rapid pace of creativity and the evolving nature of interactions, it doesn't seem to fit well for most of the digital channels if it takes so long time to wait for the result. Today, the modern MMM helps to shorten the length of time needed to analyze the required information through data.
A crucial thing to note is to change the model approach of your marketing mix to account for the statistical power that is lost by using a shorter time frame. This can be solved in many ways, including by developing dynamic models with different time parameters, by running geography-segmented models and by increasing the data granularity.
#2 Utilizing the Right Insights
While digital channels bring new challenges to MMM, they also offer new possibilities. Advertisers now have access to much more granular insights than in the past, which can generate richer, quicker analysis. The key is to use the correct inputs in models: specifically, use those that enable successful cross-channel comparisons, such as paid impressions, rather than those such as engagement metrics or clicks. It’s always recommended to use paid impressions for digital channels in models because they are actionable, purchasable and predictable.
#3 Contextualizing Past Performance
It is important to note that not all impressions are generated equally when integrating digital platforms in marketing mix models. Performance differs from campaign to campaign because there are differences in media and creative execution. In order to obtain a full understanding of the effect of different approaches, it is important to account for these variations when looking at results.
Advertisers should define key best practices in order to contextualize past success and evaluate media and creative efficiency, in order to test which approaches are best for their specific situations, and then integrate structures such as quality scores into marketing mix models.
#4 Calibrating With Experiments
Modern MMM will vary significantly from their predecessors: they often use fresh approaches, integrate new data and have been updated over time. Given this, in order to verify your conclusions, it’s important to select between models, test against established results and tweak models to make them more reliable. By doing all these, it can be very beneficial to run experiments.
Ultimately, to create a successful modern marketing model, marketers should look at shorter time frames when accounting for channels such as Facebook, utilize inputs such as paid impressions that enable effective comparisons, account for differences in media and creative quality in your models and run experiments to check assumptions and optimize your models.