Recent approaches in retail banking marketing optimization
summary
Banks generally run dozens to hundreds of propensity models (logistic regression and simple deep nets) on fragmented customer 360s, then feed scores into an optimizer that uses linear programming plus business rules and channel capacity limits to pick offers. Projects add a modeled expected lifetime revenue per contract to put scores on a common scale, simulate outcomes, and measure conversion uplift, but runtimes vary (minutes to several hours) and few banks achieve full real time end to end deployment because of governance and organizational misalignment.
AI-written summary · about
chapters (9)
- 0:00Retail banking marketing intro
- 1:50Retail banking marketing challenges
- 4:40Current banking marketing practices
- 10:42Optimization trends and motivations
- 12:27Technical components
- 15:38AI models and optimization
- 21:00Optimization constraints
- 24:11Real-world optimization complexities
- 27:02Key takeaways
related talks
35mBank evolution to data science-driven business model
58mNew-type Scoring Models: Analyzing Customer Actions
40mNext Best Action: Maximizing Customer Value
45mKnow Your Customer
39mMachine Learning in Financial Credit Risk Assessment
32mMaking Models Work: How an optimized ML lifecycle can drive better ROI
32mData science vs real world: friends or foes
34mMaximizing Credit Card Profitability with ML–Driven CLV