← Customer intelligence collection

LG Electronics Recommendation systems

Make the next recommendation personal.

A next-purchase model that translates customer history into ranked product-category recommendations for more relevant cross-sell and outreach.

Predictive rankingCross-sellTemporal validationProduction ML
My contribution

Hands-on model development and delivery as part of the customer intelligence platform.

Delivery status

Production customer model

Next-basket recommendations
Purchase context
Category ranking
Relevant shortlist
Outreach planning
From a business question to a working system.Workflow overview

What will this customer want next?

A ranked set of product categories for each known customer, usable for category audiences, cross-sell planning, and personalized recommendations.

Conceptual workflow

The business problem

A broad product catalog gives marketing teams more choices than any email or recommendation slot can hold. A shared best-seller list cannot reflect what each customer already owns or may need next.

What I built

Built personalized category ranking from purchase history and ownership context, turning customer signals into a shortlist teams can act on.

Connected category relevance with purchase readiness and category-level spend, so audience selection can consider more than product affinity alone.

Preserved prediction history and evaluated recommendations against purchases that happened later, with best-seller rankings as a practical comparison.

How I evaluated it

Compared stored predictions with subsequent registered purchases. Examined ranking quality across purchase channels and customer histories, rather than relying on a single development score.

Built to operate

A recurring scoring workflow refreshes recommendations after new purchase activity and preserves historical outputs for evaluation.

What the work enables

Historical testing showed stronger next-category ranking than a best-seller baseline, with additional audience-selection value when paired with customer purchase readiness.

Selected details are generalized to respect confidentiality. Internal metrics, customer records, and proprietary implementation details are omitted.

Let’s talk about the work