Increasing the Conversion Rate on an E-commerce Website
Industry Vertical and the Company:
An E-commerce company that deals in recreational toys.
The e-tailer was facing the problem of low conversion rate (currently 3%) and low average transaction value (>80% transactions less than $ 50 when it could potentially go up to several hundred dollars).
Potential buyers are in different states of willingness-to-purchase (WTP) that range from “not interested” to “will buy right away”. The objective is to help them move to the latter state. But the challenge is that these states are not visible to the marketer.
To meet this challenge, our approach was to use math to model visitor’s hidden state of willingness-to-purchase (WTP), and to increase the willingness as the she moved along the purchase path. This was somewhat similar to well-known AIDA (Attention, Interest, Desire and Action) model but with the increased flexibility to move both back and forth. Two models used concurrently were:
- Information stock model where visitor accumulates more information as she moves on the purchase path
- A Hidden-Markov Model to probabilistically predict the state of WTP and movement between different states of WTP to jointly model the purchase incidence and the value of purchase.
Applying Bayesian approach, we estimated the model using the open source OpenBUGS.
Our solution could estimate a visitor’s states of WTP. Based on this information, we devised a set of rules for the e-tailer to expose the visitor to different marketing communications to help her make a purchase, and thus almost doubling the conversion rate.