← Customer intelligence collection

LG Electronics Personalization & value

The right category is only half the answer.

Category-aware spend predictions that connect product relevance with a suitable price level, giving teams a more useful starting point for personalized merchandising.

Spend predictionCategory contextPersonalizationModel calibration
My contribution

Built the category-spend capability and connected its outputs to the customer intelligence workflow.

Delivery status

Production customer model

Category spend & price preferences
Likely category
Spend estimate
Price preference
Product selection
From a business question to a working system.Workflow overview

What price level fits this purchase?

Expected spend conditional on a category purchase, plus a price-preference tier where appropriate, to guide product selection and category audience planning.

Conceptual workflow

The business problem

A customer's overall value does not describe what they would spend in a particular category. Treating every category alike can distort audience priorities and lead to poorly matched product recommendations.

What I built

Developed a category-aware spend model that builds on customer purchase context and category pricing patterns.

Translated predictions into category-specific spending estimates and entry, mid-range, or premium preferences where the data supports that distinction.

Integrated category spend with customer value and next-purchase ranking to make audience planning sensitive to both relevance and value.

How I evaluated it

Compared predictions with subsequent category purchases. Checked aggregate spend, within-category ranking, and whether predicted price tiers corresponded to different observed spending patterns.

Built to operate

Recurring scoring integrated with upstream customer models, with checks before publishing new outputs and a prepared retraining and evaluation workflow.

What the work enables

Historical validation showed more informative category-level spending estimates than a single customer-wide spend value, and meaningful differences in observed spend between predicted price tiers.

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

Let’s talk about the work