Data Factories and API-First Setup

Prefer factory functions that accept overrides and return complete objects (createUser(overrides))—never rely on static fixtures.
Seed state through APIs, tasks, or direct DB helpers before visiting the UI; UI-based setup is for validation only.
Ensure factories generate parallel-safe identifiers (UUIDs, timestamps) and perform cleanup after each test.
Centralize factory exports to avoid duplication; version them alongside schema changes to catch drift in reviews.
When working with shared environments, layer feature toggles or targeted cleanup so factories do not clobber concurrent runs.

Source: Murat Testing Philosophy, blog posts on functional helpers and API-first testing.