Prior research on advertising media mixes has mostly focused on single channels (e.g., television), pairwise cross-elasticities, or budget optimization within single campaigns. This is starkly detached from advertising practice where (i) there is an increasingly large number of media channels available to marketers, (ii) media plans employ complex combinations of channels, and (iii) marketers manage complementarities among many (i.e., more than pairs) channels. This research empirically learns complex channel complementaries using Latent Class analysis. Latent classes have three useful properties: (i) they account for non-random selection of channels into campaigns, (ii) they capture pairwise and higher-order interactions between channels, and (iii) they allow for meaningful interpretation. We empirically describe the most common media channel archetypes and estimate their relationship to the effectiveness of a set advertising campaigns on a set of common brand-related performance metrics. We use a dataset of 1,083 advertising campaigns from around the world run between 2008 and 2019. We find that there is not a systematically “best” media mix that correlates to dominant performance across all metrics, but clear patterns emerge given specific metrics. We find that traditional channels (TV, outdoor) are commonly paired with digital channels (Facebook, YouTube) in high-performing campaigns. We also find that current marketing practice appears far from optimal, and simple strategies have the potential to increase brand mindset metric lifts by 50% or more.
Read full abstract