Multi-Channel Optimal Path Sequencing Through Bayesian Deep Learning
Multi-Channel Optimal Path Sequencing Through Bayesian Deep Learning

Abstract: 

Marketing leads are exposed to numerous channels, and it creates a complex cross-channel relationship that makes the effectiveness of the campaign difficult to comprehend and execute. Are there path-sequences that are better at driving leads than others? Build RNN based Deep Learning model and Conduct Path Analysis and Produce weights for channel contribution for each path?

We will learn optimal path sequencing in the context of Marketing Channel Attribution and Sequencing modeling. While attempting to attribute credits to a channel, it is important to take into account Channel interactions, the number of impressions on the channel and the order in which the channel was touched in a lead’s journey.

Further, it is difficult to isolate how much credit should be attributed towards each channel. This can have a big impact on how budget/efforts should be allocated towards any particular channel.

While Deep Learning has shown significant promise towards model performance, it can quickly become untenable particularly when data size falls short of problem space. One such situation regularly appears when modeling with RNNs. RNNs can quickly memorize and over-fit (the problem is further aggravated when data size is small to medium).

However, on the other hand, _Bayesian_ techniques are more robust to missing data, noise and data-size but it lacks order or sequence information.

The presentation exposes the shortcomings of RNNs and how a combination of RNNs with the Bayesian approach can not only overcome this shortcoming but also improvise the sequence-modeling behavior of RNNs. We will learn this in the context of Marketing Channel Attribution and Path modeling.

The presentation will go on to explain how we can visualize the latent space and perform ‘Next best Action’ on the potential leads, thereby maximizing the impact of channel-based treatment towards any desired outcome.

● R-ecosystem
● Tensorflow version 2+
● tensorflow-probability v0.9 ,
● tfdatasets
● R-interface to tensorflow

https://github.com/vishalhawa/odsc
https://drive.google.com/open?id=1QNLpbpXup4Tw4lClqYlgBae-NPt1lsik

Bio: 

Vishal Hawa is a Principal Scientist at The Vanguard Group, where he works closely with marketing managers to design attribution, propensity, and attrition modeling. Vish has over 15 years of experience in the retail and financial services industries. He has training in executive management from the Wharton School and holds postgraduate degrees in information sciences, statistics, and computer engineering from the Indian Statistical Institute.