Design identity resolution rules for a multi-brand retailer

Real World Scenario

A retailer operates three brands with separate e-commerce sites, a loyalty app, and in-store POS. Customers use different emails per brand but share one phone number. Marketing needs a single golden profile without incorrect merges.

Expected Answer

• Inventory all data sources with primary identifiers: email, phone, loyalty ID, device ID, and anonymous cookie • Define match rules in priority order: verified loyalty ID > phone + last name > email exact match • Use probabilistic matching with confidence thresholds; route low-confidence matches to review queues • Implement survivorship rules for conflicting attributes (e.g., most recent address wins) • Separate household vs individual profiles if B2C models require it • Test match accuracy with labeled golden datasets before production activation • Monitor false merge rate and provide unmerge workflows with downstream propagation

Follow-Up Questions & Answers

Click to expand — each follow-up includes a direct, interview-ready answer

Handle it in layers. First: Inventory all data sources with primary identifiers: email, phone, loyalty ID, device ID, and anonymous cookie. If that fails: Define match rules in priority order: verified loyalty ID > phone + last name > email exact match. For edge cases in "How do you handle shared family email addresses wi…", Use probabilistic matching with confidence thresholds; route low-confidence matches to review queues. Identity resolution is the foundation—bad merges poison every downstream segment and agent personalization. Document the decision in an ADR and align with enterprise standards.

Architect Perspective

Identity resolution is the foundation—bad merges poison every downstream segment and agent personalization. Start conservative with match thresholds; false merges are harder to undo than false splits.