Entity Resolution (Ecommerce)

Entity resolution in ecommerce is the data-engineering discipline of reconciling records that refer to the same real-world entity across multiple systems. A single customer order, in a typical mid-market commerce stack, generates record fragments in: Shopify (order, line items, discount applications), the payment processor (transaction, processor fee), the 3PL (pick ticket, shipment, freight cost), the ERP (revenue recognition entry, COGS allocation), the returns platform (RMA, restocking outcome), and several analytics and marketing systems (attribution, LTV bucket). Each system has its own ID space and its own record format. To answer a question as basic as 'what was the net margin on order X' requires joining those fragments together. Entity resolution is the systematic solution: a process that identifies cross-system identifiers (Shopify order_id ↔ Stripe payment_intent_id ↔ ERP sales_order_number ↔ 3PL shipment_id) and produces a single canonical record per entity. It is non-trivial because: (1) systems do not always share IDs natively, requiring rule-based or ML-based matching on attributes like timestamp, amount, customer identifier; (2) systems update at different cadences, so the canonical record must handle race conditions and eventual consistency; (3) data quality varies — addresses are typed differently, customer names are inconsistent, SKUs are renamed mid-quarter. Semantic entity resolution applies natural-language and ML techniques (embeddings, fuzzy matching, supervised classifiers) on top of rule-based foundations to handle the harder cases. The output of entity resolution is a 'golden record' — a single trusted version of each entity that downstream analytics and enforcement can use without per-query reconciliation. For a profit firewall, entity resolution is the foundation: real-time margin enforcement requires sub-second access to the joined view of cost (from ERP), price (from commerce), discounts (from commerce + apps), freight (from 3PL or carrier), and FX (from treasury). Without robust entity resolution, the margin engine is either making decisions on incomplete data or blocking on cross-system queries — both of which are unacceptable at checkout latency. Mid-market merchants typically begin to feel entity-resolution pain at $10M+ revenue, when the proliferation of commerce channels and back-office systems exceeds the capacity of spreadsheet-based reconciliation. The mature pattern is a dedicated data layer (Snowflake, BigQuery, or similar) with a master-data-management approach to identifiers, plus event-streaming for real-time downstream propagation.