IncreasingE-commerceConversionsbyIdentifyingHigh-PurchasePropensityforOurClient’sOnlineJewelleryChain

How we worked the problem

1. 01

   Data Mart Creation: We constructed a Data Mart offering a unified view of each customer. This included basic variables such as age, location, pages visited, events clicked, past transaction details, queries, and complaints. It also incorporated derived variables like the customer value index, engagement index, time spent on the website, and the number of media channels used.

2. 02

   Ensemble Model Development: We condensed over 600 event and 40 transactional variables down to an efficient number of variable components. With these, we developed an ensemble prediction model for each segment using logistic regression, random forest, and support vector machine models. This model was designed to predict visitors with high sales potential.

3. 03

   Armed with insights from our model, the client segmented their website visitors and customized marketing activities for each group. These activities included offering greater incentives to those with lower purchase propensity scores, introducing subscriber retention offers, and assigning concierges to re-engage subscribers at risk of churn.