← Back to Projects
Data AnalyticsApplied Math

Stochastic Modeling of Customer Purchase Behavior

Absorbing Markov chains to analyze and optimize retail checkout flows from 100K+ transactions.

Stochastic Customer Modeling
  • Feature-engineered seven customer behavior states from a 100K+ transaction dataset to model the full purchase journey.
  • Performed statistical analysis to estimate the empirical Markov chain transition probabilities, determining a baseline purchase conversion rate of 41.8%.
    • Customer State Heatmap
  • Conducted "what-if" simulations by boosting key transitions (browse→checkout, checkout→purchase) by 5–20%, projecting up to a 6.9% increase in overall conversions.
  • Formulated UX and marketing recommendations — single-page checkout, guest checkout flow, exit-intent offers — to reduce drop-off and enhance purchase pathways.