Unified Retail Metrics & APIs

01 / Business Challenge

• Leadership reviews were undermined when two teams presented the “same” KPI with different definitions.
• Many analysts re-created the same joins and extracts, inflating warehouse spend.
• Shared tables changed without a clear version or retirement path for consumers.
• Excel, BI tools, and internal apps needed stable, documented outputs—not ad hoc database access.
• Store and employee data required consistent redaction and access rules.

02 / Our Approach

How we executed this engagement in practice. The phases below describe the delivery rhythm we use across ServiceNow, custom engineering, and mobile programs.

We worked with the client’s data engineering team to publish trusted, tested datasets (using dbt for documentation and quality checks). On top of those datasets we shipped versioned read APIs and CSV exports with FastAPI, scoped API keys by business domain, and explicit API versions when definitions changed. Finance helped lock canonical KPI meanings. We added caching for the busiest endpoints so seasonal reporting stayed responsive and affordable.

• Published dbt-documented “gold” datasets for sales, inventory, and margin.
• Implemented FastAPI services with OpenAPI, rate limits, and structured error responses.
• Centralized row- and column-level security patterns for store and employee attributes.
• Built lightweight admin tooling to rotate keys and map consumers to allowed datasets.
• Added integration tests that fail CI when contract schemas drift from dbt outputs.
• Delivered analyst onboarding guides and example notebooks against the stable APIs.