Businesses have long used models to determine the best allocation of marketing budgets across all available media formats. The Council for Research Excellence recently studied this topic and issued its findings in the Current State of Marketing Mix Model Report. As digital formats grow more popular, marketers are taking a closer look at what their models should be accomplishing and how to change them to incorporate new media.
While some advertisers are staffed with analysts who understand how to design and apply models, others rely on their agencies to do this job for them. The authors of this study say marketers should understand what they're hiring agencies to do, to be sure they are applying the right models to achieve the desired outcomes. For example, popular models in use today include regression analysis, which is highly efficient, and cross-channel attribution, which applies only to digital formats.
Some of the details marketers are considering as today’s marketing mix models are developed include:
- Return on operating objectives – as distinct from ROI
- Upper funnel effects – such as awareness
- Repeat purchase patterns
Marketers are also struggling with how to get modelers to work with media-research experts. These two fields are staffed by distinctly different professionals, yet they should be working together to design good models, input the proper data, and then analyze the results of campaigns.
Another challenge identified by the CRE study is one many industry experts have heard before – how to deal efficiently with the flood of data in the digital channel. Without finding a good way to analyze the data, the information is meaningless. A similar problem exists when measuring the impact of social media investments. And, with the mobile channel growing exponentially, marketers are desperate to find ways to accurately measure what’s going on. Analysts point out that it’s easy to track mobile web traffic but it’s far more difficult to track app activity.
Have you been reviewing your marketing mix models? Have you found a good way to incorporate app activity into the models?