InfoWatch senior test engineer Mikhail Shalepo has shared a comprehensive internal regulation that connects manual and automated testing into one coherent quality process. The article uses the metaphor of an “orange” to describe the overlapping area of functional verification where the two testing approaches must deliver a single, reliable signal about product state.
Shalepo notes that when automation volume, environments, and mandatory runs remain small, teams often rely on shared context. As the product grows, this implicit understanding disappears, creating the need for explicit rules on selecting automation candidates, validating new tests, defining mandatory regression sets, classifying failures, and recording manual compensation.
Core Principles of the Regulation
The document distinguishes between process rules—who decides, what result is recorded, and who owns the next step—and specific engineering decisions for individual features. It establishes that manual testers review the purpose of an automated scenario rather than its code, and that every failure must receive an owner and follow-up action before a run is considered resolved.
The regulation is split into two blocks: development and verification of new functionality, and full regression testing before release. Both blocks follow the same three-stage structure—preparation, execution, and completion—producing clear artifacts at each step.
Feature Development Block
Before a sprint, manual testers and automation engineers meet to discuss scenarios worth automating. Decisions are turned into backlog tasks instead of remaining informal promises. During the sprint, the manual testing team acts as the customer of automation meaning, while smoke tests must pass before any build reaches manual verification. At sprint end, coverage status is recorded directly in TMS Scale to keep the single source of truth up to date.
Pre-Release Regression Block
Before regression begins, teams align on the full scope of checks, supported configurations, and exit criteria. During execution, automated results are triaged jointly with developers; every failure is classified, assigned an owner, and linked to a task or defect. Manual compensation is documented only when an automated signal cannot be restored quickly. At the end of regression, the team produces a connected set of artifacts that allow full reconstruction of the release decision without relying on chat history.
Shalepo emphasizes that the regulation does not guarantee zero defects or automatically shorten regression time. Its value lies in making the basis for release decisions observable and repeatable across the team.
