Top-Down vs. Bottom-Up Attribution

Entering 2014, most would agree cross-platform marketing has become a must for any company hoping to reach consumers on their terms. They would also likely concur that such a mandate carries with it an expectation that as marketers tack on additional channels, formats and platforms, they would be met with a proper method to validate those decisions. Cross-platform attribution is that solution, according to a new eMarketer report, “Cross-Platform Attribution: A Status Report on Overcoming Select Attribution Challenges.”

There is undoubtedly great demand for cross-platform attribution—a method of assigning credit to a particular marketing-driven interaction or other brand-impressed touchpoint—but also great confusion as to how to best approach it and access the data necessary to prove bottom-line effects. As such, few marketers are attempting to undertake cross-platform attribution modeling.

Traditional marketers often rely on marketing mix or econometrics methodologies to understand how to allocate their ad spending for maximum effect. This requires inputting media spend for all channels and formats, as well as gross rating points, print exposures and other marketing impressions, into a model. Marketers can then add as many additional datapoints as they want, including sales revenues or external influences that could affect performance.

This model aims to provide a general understanding of the incremental revenues generated by introducing one or more channels, which marketers can then use to justify spending and inform future media plans. Marketing mix modeling is often referred to as a top-down approach for its more broad-level view of channel influences and interactions.

The marketing mix approach’s lack of granularity is a far cry from digital-focused marketers’ use of path analysis attribution—a bottom-up approach that relies on cookies to provide user-level insight into each channel, format and ad creative prior to the point of conversion. But unlike the majority of econometrics models that ingest data and then use statistical analysis to assign a credit value or weighting to each channel, the majority of path analysis models use simplistic crediting methods such as last-click attribution.

Such an approach is perilous not just because it skews attribution heavily toward a particular channel or format, but also because it pushes marketers to optimize those channels and formats, devaluing the effects of more upper-funnel digital formats and completely negating the effects of offline channels.

Courtesy of eMarketer