Prediction Solution for Web Portal Visitors
Scope

One of the major EU-based manufacturing companies tasked Auriga to develop a solution analyzing visitor’s behavior on the web portal and purchase history to predict prospective purchase, increase engagement and boost sales.

Projects Highlights
  • 55K users records, 7K web-pages, 5K positions in store.
  • Churn prediction: refactoring and improvement of existing approach.
  • Data processing transferred from Azure + BigQuery to AWS Redshift.
  • Accuracy for recommender system increased (0.83).
  • Data processing accelerated by 10x (1 hour -> 5 minutes).
  • Data inconsistency decreased (16% -> 5%).
  • Fully automated cloud CI/CD.
Achieved Benefits
  • Recommender system: developed from the scratch in 3 months from start.
  • Increased level of customer’s engagement in sales process and communications (>10%).
  • Increased sales turnover irrespective of seasonal and other effects.
  • Scheduling and data processing pipeline fully automated.
  • Notification system launched.