One Collector, Three Data Shapes: HTML, Directories & JSON

The challenge Web-data jobs vary in layout, but rebuilding retrieval, pagination, deduplication, validation, and exports for every source wastes time and creates inconsistent QA. The reusable approach I built one configuration-driven collector and tested it against three independently defined source shapes: a product catalog with detail pages, an organization directory, and a paginated JSON service. Only the schema and selectors changed; the retrieval, verification, evidence, and delivery pipeline stayed the same. Exact-reference results The product example matched 3 rows and 21 business values exactly. The organization directory matched 3 rows and 12 values exactly. The JSON example matched 3 rows and 18 values exactly. Every delivery passed its output verifier. Change and failure handling A controlled comparison correctly identified one added record, one changed field, and one missing record. When a page failed, the run was marked incomplete and change conclusions were suppressed. An interrupted run resumed from verified cached responses without duplicating records. Client value This makes a new supported source faster to configure while keeping coverage, source evidence, and acceptance criteria visible. Each real source still receives a feasibility sample before a full quote. This is an independent synthetic portfolio demonstration, not paid client work.