← Portfolio
Proprietary

Hybrid CRM offer engine at Bci

Industry
Banking
Type of work
Decision engine design
Period
2017

Confidential project: code and data are not public.

Architecture

Rules expert judgment Triggers birthdays, policyexpiry, etc. ML scores propensity models Hybrid engine rules + thresholdson the scores Teradata DW offers per customer 400K+ customers evaluated daily Shiny threshold tuning Bci · 2017 · design and first implementation

Goal

Decide every day which offer to show each customer, combining the business's commercial judgment with the propensity models that already existed.

What I did

  • Designed the engine and built its first implementation.
  • Rules were coded directly from expert judgment; ML contributed scores from the existing propensity models.
  • Added triggers on customer events, such as a birthday or an insurance policy expiring, to make offers at the right moment.
  • Offers were pulled from the Teradata data warehouse.
  • Decision thresholds were tuned with a Shiny application.

Results

400K+
customers evaluated daily

Stack

TeradataSQLR ShinyXGBoost