Working Paper

Attribution Modeling: Understanding the Influence of Channels in the Online Purchase Funnel

Report Summary

Online marketers invest significant resources in driving traffic to their websites through multiple marketing interventions and channels. Authors Li and Kannan propose a conceptual framework to examine the nature of carryover and spillover effects of prior visits through different channels to a firm’s website across a number of commonly used online channels, both at the visit and purchase stages.

They develop a three-level measurement model of customers’ consideration of online channels/ sources, their visits through these channels over time, and subsequent purchase at the website, which accounts for carryover and spillover of visits. Based on customers’ path data of visits and purchases at a travel and hospitality firm’s website, they find significant carryover and spillover effects—for example, e-mails and display ads trigger visits through other channels, while e-mail leads to significant purchases through search channels. Attributing credit for the different channels for the purchase conversions using the model estimates, they find that the relative contributions of these channels are significantly different as compared to estimates of the widely-used “last-click” attribution model. A field study conducted at the firm’s website by turning off paid search for a week validates the ability of the model in estimating the incremental effect of a channel on conversions.

This study highlights that the commonly used last-click attribution or the averaging attribution models are not good metrics for understanding the real impact of firm-initiated channels as well as customer-initiated channels on conversions.

Some of these metrics consider only those visits that result in conversion immediately to the exclusion of the other visits, and all of them consider only the visits that result in conversion while excluding the information that can be derived from non-conversions. While they may provide passable results in product categories with very short purchase funnels (with one or two touch points) and with fewer channels, they will invariably be misleading in product/service categories with longer purchase funnels as in high involvement categories—such as consumer durables and travel services—and for firms with multiple channels, both customer-initiated and firm-initiated. In the latter case, the effectiveness of firm-initiated efforts is generally underestimated using the last-click model.

In addition, there are limits to the effectiveness of firm-initiated efforts such as e-mail and display retargeting campaigns. Repeated e-mail targeting may actually hurt conversions in certain cases and retailers should be well-advised to carefully calibrate the targeting of customers with these instruments.

Hongshuang (Alice) Li is a doctoral candidate in the marketing program and P. K. Kannan is Ralph J. Tyser Professor of Marketing Science at the Robert H. Smith School of Business, University of Maryland.

Acknowledgment
The authors acknowledge and thank the support of Marketing Science Institute through their grant #4-1680 for this research.

Featured in Insights from MSI.

Comments from members