Learning Multi-touch Conversion Attribution with Dual-attention Mechanisms for Online Advertising

Abstract

In online advertising, the Internet users may be exposed to a sequence of different ad campaigns, i.e., display ads, search, or referrals from multiple channels, before led up to any final sales conversion and transaction. For both campaigners and publishers, it is fundamentally critical to estimate the contribution from ad campaign touch-points during the customer journey (conversion funnel) and assign the right credit to the right ad exposure accordingly. However, the existing research on the multi-touch attribution problem lacks a principled way of utilizing the users' pre-conversion actions (i.e., clicks), and quite often fails to model the sequential patterns among the touch points from a user's behavior data. To make it worse, the current industry practice is merely employing a set of arbitrary rules as the attribution model, e.g., the popular last-touch model assigns 100% credit to the final touchpoint regardless of actual attributions. In this paper, we propose a Dual-attention Recurrent Neural Network (DARNN) for the multitouch attribution problem. It learns the attribution values through an attention mechanism directly from the conversion estimation objective. To achieve this, we utilize sequence-to-sequence prediction for user clicks, and combine both post-view and post-click attribution patterns together for the final conversion estimation. To quantitatively benchmark attribution models, we also propose a novel yet practical attribution evaluation scheme through the proxy of budget allocation (under the estimated attributions) over ad channels. The experimental results on two real datasets demonstrate the significant performance gains of our attribution model against the state of the art.

Digital Marketing Attribution: Understanding the User Path

“…In our case, the number of players maps to the number of channels or campaigns, so that this restrings a lot the number of campaigns that we can compute attribution for. The research community has also discussed other alternatives using Markov Chain [5] or recurrent neural networks [4]. However, these approaches have a significant limitation in the interpretability of the obtained results, making them invalid to support the decision-making process that is the ultimate purpose of attribution models.…”

“…The research community has proposed alternative data-driven models to Shapley value. These works propose the utilization of neural networks [4], Markov chains [5,6], survival analysis [7,8], regressions [9] or econometric models [10]. Finally, some works have analyzed the impact of specific channels, in particular, display ads, in the global attribution [11,12].…”

Dynamic Knapsack Optimization Towards Efficient Multi-Channel Sequential Advertising

“…In a digital age, to drive conversion, advertisers can reach and influence users across various channels such as display ad, social ad, paid search ad (Ren et al, 2018). As illustrated in Figure 9, the user's decision to convert (purchase a product) is usually driven by multiple interactions with ads.…”

“…Apart from the myopic approaches, there exists some literatures considering the long-term effect of each ad exposure. Multi-touch attribution (MTA) (Ji & Wang, 2017;Ren et al, 2018;Du et al, 2019) study the credits assignment to the previous ad displays before conversion. However, these methods only attend to figure out the contribution of each ad exposure, while not providing methods to optimize the strategies.…”

The differences of conversions in working with target audience at different decision-making funnel levels

“…In the classical form, the idea is realized in the form of hierarchical models of communication. The most famous of them is AIDA (William W. Townsend, 1924), its variations AIDAS, DIBABA, AIMDA and others [1,2]. From a marketing point of view, modern communication theories suggest that along with rigid hierarchical models, there are some forms implying horizontal distribution of information (hierarchical and combined communication models; models using the Share-AISAS level (Dentsu, 2006)) [3,4].…”