DB management for a global fashion brand — untangling orders, stock and returns from a dozen marketplaces.

The chaos

A global fashion manufacturer sold through own stores, wholesale partners and more than a dozen marketplaces. Each channel exported different schemas — SKUs that did not match internally, returns attributed to the wrong warehouse, stock double-counted after promotions.

Finance could not reconcile revenue by channel. Ops could not answer which marketplace triggered a spike in returns. Leadership had dashboards built on spreadsheets that disagreed with each other.

Approach

Under NDA we mapped every inbound feed, defined canonical product and order entities, and built a governed warehouse with lineage — every row traceable to source file, channel and ingest time.

• Ingestion layer — scheduled pulls and SFTP drops normalised to staging
• Matching rules — internal SKU, EAN, marketplace IDs with confidence scores
• dbt transforms — tested models for orders, stock, returns, margin by channel
• Reconciliation jobs — flag orphan rows, duplicate stock movements, VAT mismatches

What changed for the client

One database leadership trusts. Channel P&L without manual VLOOKUP marathons. Returns routed to the correct fulfilment node. New marketplace onboarding becomes configuration, not a three-month BI project.

NDA note

Client name, exact channels and dashboards are confidential. This summary reflects the problem class and solution shape only.