QA Engineer

Key responsibilities

• Test strategy across the six services sits with this seat, along with the regression suite that keeps every line of business working. Business and Industry, Healthcare and Higher Education each have their own market baskets and their own assessment template, and Catering and Services runs across them.
• Calculation correctness is validated against reference assessments. That covers value to market at item, category, unit, market and account level, straight and sales-weighted averages, and scenario outcomes against the contractual cap where the account is on a fee contract.
• Mapping quality is tested as a measurement rather than a pass or a fail. The measures are precision, recall and coverage, whether the thresholds route the right items to a person, and whether analyst corrections improve later matches.
• The AI controls are tested as well as the AI itself. That means the model version and prompt are pinned, the full prompt and every decision are logged, inputs and outputs are reviewed for prompt injection, and low-confidence results reach a person. Compass has these in writing, which makes them testable requirements.
• Data validation on the Excel upload portal is tested against Compass templates, including malformed and partial submissions from external submitters.
• Access control and approval routing are tested role by role, and the audit trail is tested for completeness. Every decision, override and approval has to be present and has to be queryable.
• Automated coverage is built into GitHub Actions alongside the pipeline’s quality and security gates, and the suite is kept fast enough that people actually run it.
• Defect management runs from this seat, as does user-acceptance testing with the pricing managers. They decide whether a mapping or a recommendation is right, and their time is committed at two to four hours a week, so scheduling it forms part of the role.

Required experience

• Test automation for web applications and APIs is essential, using Playwright, Cypress, Selenium, pytest, Jest or comparable tools, and building a suite that runs in continuous integration.
• Data-validation and calculation testing carries the most weight in this seat, meaning comparison of an application’s output against a known-good source and accounting for every difference, down to the small ones.
• SQL needs to be good enough to verify results directly in the database and the warehouse as well as through the interface.
• Testing role-based access control and multi-step approval workflows should be familiar ground.
• Writing acceptance criteria is part of the seat, and they have to be criteria a developer can build against and a business user recognizes as their own requirement.
• Defect management, along with experience running user-acceptance testing directly with business users.
• Git and continuous integration, including quality gates in a pipeline someone else governs.

Useful additions

• Testing machine-learning output, where the measure is a distribution rather than a single expected value.
• Spreadsheet-heavy domains, and financial or pricing calculation testing.
• Performance and load testing, which is shared with the Benchmarking & Scenario Developer.