Analytics — QA Plan

TKN-24 targeted regression commands

Run from the repository root. These commands are references for report, route-hydration, and insight regressions, not execution results or production benchmarks. Documentation cleanup does not run validation; the implementation owner records exact latest results separately.

pnpm -C apps/api exec vitest run --config vitest.unit.config.ts src/analytics/analytics.service.unit.spec.ts src/analytics/analytics.controller.unit.spec.ts src/analytics/analytics-reports.controller.unit.spec.ts
TEST_DB_WORKERS=1 pnpm -C apps/api exec vitest run --config vitest.integration.config.ts src/analytics/analytics-reports.int.spec.ts
NEXT_PUBLIC_E2E_TEST_MODE=true pnpm -C apps/web exec vitest run --config vitest.integration.config.ts src/components/analytics/reports-screen.int.spec.tsx src/components/overview/v2/overview-v2.int.spec.tsx src/components/overview/v2/revenue-panel.int.spec.tsx src/components/overview/sidebar-nav.int.spec.tsx
pnpm -C apps/api exec vitest run --config vitest.unit.config.ts src/insights/insight-consolidation.unit.spec.ts src/insights/insights.service.unit.spec.ts src/insights/insights.controller.unit.spec.ts
TEST_DB_WORKERS=1 pnpm -C apps/api exec vitest run --config vitest.integration.config.ts src/insights/insight-consolidation.int.spec.ts
NEXT_PUBLIC_E2E_TEST_MODE=true pnpm -C apps/web exec vitest run --config vitest.integration.config.ts src/components/payments/payment-tabs.int.spec.tsx 'src/app/[lang]/(protected)/dashboard/tasks/tasks-page.int.spec.tsx' src/components/overview/v2/action-center.int.spec.tsx src/components/overview/schedule/weekly-calendar.int.spec.tsx
pnpm -C apps/web exec vitest run --config vitest.unit.config.ts 'src/app/[lang]/(protected)/dashboard/schedule/schedule-page.unit.spec.ts'
make test-e2e-web   # includes e2e/specs/owner-reports.spec.ts

The PostgreSQL command requires DATABASE_URL pointing to a prepared, idle, disposable _test database. Existing setup clones/drops worker databases and uses deterministic scoped reporting fixtures; do not point it at application data.

TKN-24 acceptance matrix

Prior local verification baseline

Before the three-area/action consolidation, TKN-24 recorded existing API unit/controller tests 25 PASS, new controller tests 43 PASS, real-Postgres report tests 10 PASS, and focused web suite 49 PASS (including 28 report tests). API/web TypeScript checks, focused ESLint, and Nx API/web builds also passed at that prior baseline. These are historical implementation results, not latest consolidation verification.

The prior cmux browser smoke covered report projections across en/he/ru, class/date/method filters, a CSV click, future/partial lifecycle cells, multiple currencies, and the OFF baseline. Its former UI arrangement is not evidence for the current three-area navigation or reports-only snapshot placement.

That prior browser stayed at 743×994, ignored requested widths, and produced viewport-only screenshots. The current consolidation browser surface is 800×600 despite the requested viewport; neither establishes complete responsive coverage. API smoke uses a private local clone without migrations, not production. At the prior baseline, full-demo seed failed on unrelated missing subscriptions.discount_percent; scoped reporting fixtures recovered that smoke. No production smoke or performance benchmark is claimed.

Consolidation verification — 2026-09-09

Actual API smoke exercised both local rollout branches: with no overrides, status was 200 { enabled: false }, report data/new CSV were 404, and a legacy CSV remained 200. With isolated test-process overrides only, all seven report CSV projections returned 200 with UTF-8 BOM. A charge created in August but settled in September and a September refund produced ILS 100 collected / 20 refunded / 80 net; USD 50 remained a separate currency. Filtered attendance returned one attended booking, 100% attendance and 10% weighted capacity utilization. Insights returned 200 in both branches.

Actual built-web browser smoke opened revenue/attendance/retention direct URLs, followed area history, exercised a CSV download button, verified Hebrew RTL and Russian filter hydration, opened the Payments debt tab and the real not_completed&type=manual_refund&search=all task state, and visually confirmed the settled Overview Action Center without a Studio KPI snapshot. The cmux surface stayed at 800×600 and screenshots were viewport-only; complete responsive coverage is not claimed.

The smoke used a privately cloned local test database and scoped fixtures. The broad base seeder still failed on the unrelated missing subscriptions.discount_percent; no migration or source workaround was applied. Local test-process overrides did not create, modify, or enable remote PostHog flags. No production smoke, accounting reconciliation, or performance benchmark was performed.

Legacy summary regression checklist

The checks below concern the pre-existing summaries/exports, not the stricter new report contract.

Auth

Revenue summary

Revenue trend

Members summary

Members activation

At-risk

Classes

Workouts summary

Legacy CSV export

Agent surface

Performance

• Suggested measurement targets, not observed production results: p95 < 500ms for orgs up to 1000 members; investigate coach overview above 1s for 50-coach orgs.
• Repeated requests should not regress (no per-call cache, but indexes may be warm). Local reporting smoke establishes behavior, not scale.

Legacy negative inputs

• Pass months=abc → parsed as NaN, defaults applied.
• Pass sample=-1 → behavior depends on SQL LIMIT; document and ideally clamp to ≥ 0.
• Pass days=10000 → query returns all historical data; performance scales linearly.