Functional behavior
Authenticated loading, dashboard KPIs, brands, pipeline, tasks, command palette, theme switching, and accounting flows.
OrbelithExplore the productManual QA case study
Testing Orbelith
A focused production QA pass that turned functional, financial, localization, and data-consistency findings into verified fixes.
- Role
- QA Engineer / Product Tester
- Date
- 8 August 2026
- Coverage
- 16 documented checks
- Result
- 5 / 5 defects passed retest
Test objective
Validate the product as a connected system
The pass targeted authenticated workflows and the consistency of the data behind them. It used the live owner workspace with demo records and avoided destructive actions, payment execution, outbound messages, exports, and third-party credentials.
Data consistency
Cross-page comparison of task counts, invoice classification, overdue totals, and pipeline metrics against visible records.
Product quality
Empty states, English and Russian localization, desktop and 390 px mobile behavior, plus unauthenticated API boundaries.
Approach
A complete QA loop, not a screenshot exercise
Each finding moved from observable behavior to reproducible evidence, impact-based severity, a product-level correction, and focused retesting.
10
Passed initially
5
Defects found
1
Environment block
0
Production records changed
Defect analysis
Five findings, five verified fixes
The defects ranged from misleading product feedback to a financial classification error that affected several dependent views.
BR-03High
Past-due invoices were missing from overdue totals
Pending invoices past their due date were not consistently classified as overdue across finance surfaces.
Fixed · retest passed
Introduced one effective-status rule across Finance, Accounting, reports, client views, and scheduled digest logic.
BR-01Medium
Filtered search showed the wrong empty state
A no-match brand query displayed the first-use “no brands yet” state even though four brands existed.
Fixed · retest passed
Separated first-use and filtered no-result states and added a clear-search action.
BR-02Medium
Pipeline percentages implied conversion
The interface labeled current stage occupancy as conversion, which could lead to a false funnel interpretation.
Fixed · retest passed
Renamed the metric to current stage share and added explanatory context.
BR-04Low
Russian invoice count used the wrong plural
The singular state rendered “1 счетов” instead of the correct localized form.
Fixed · retest passed
Added locale-aware plural rules and checks for 1, 2, 5, 11, and 21.
BR-05Low / Medium
Open-task KPI included hidden subtasks
The top-level KPI and the visible board used different counting scopes.
Fixed · retest passed
Aligned the KPI with the parent-task scope shown by the board and export.
Highest-impact finding
Financial correctness required one shared rule
The overdue defect was not limited to one label. The correction had to propagate through Finance, Accounting, reports, brand details, the client portal, and scheduled digest logic.
BR-03 · High
- Observed
- Past-due invoices still marked pending were excluded from overdue totals.
- Root cause
- Several consumers trusted the stored status instead of deriving overdue state from status and due date.
- Retest evidence
- Accounting and Finance both reported six overdue invoices totaling 3,521,300 L.
Production retest
Evidence across product surfaces
The retest checked the visible outcome and the product rule behind it. Interface language is English; names inside demo records may remain multilingual.
BR-01 · Passed
A no-match query now shows the correct recovery path
Brands · filtered empty state
BR-02 · Passed
Pipeline percentages describe current stage share
BR-05 · Passed
Open-task KPI matches the visible board scope
Security boundary
Request checks backed the manual pass
Existing Playwright request tests were used where a browser session was not required.
- Protected brand and task APIs returned 401 without a session.
- OAuth connection start routes rejected anonymous callers.
- OAuth callbacks rejected a tampered brand identifier in state.
Outcome
Evidence before claims
The final result is a closed, traceable QA loop: defined coverage, reproducible findings, root-cause corrections, and production retest evidence.
✓
All five documented product defects were fixed and passed focused production retesting.
✓
One high-severity financial accuracy risk was corrected across every affected surface.
✓
Search, pipeline, task, accounting, and localization behavior now use consistent product rules.
✓
Responsive behavior passed at a 390 × 844 mobile viewport.
✓
Request-level security checks passed for protected APIs and OAuth state handling.
✓
No customer data was created, changed, deleted, or exported during the test pass.
Evidence note
This is a self-initiated case on the author-owned product. Screenshots were captured from the live owner workspace with demo records and personal account details excluded. The case does not use former-employer data or present reconstructed summaries as exported Jira artifacts.