jobtrackingapp/docs/redesign/phase-5/p5-07

P5-07 — Existing-report Analytics destination

Source checkpoint: ec6baaeddd3e950b1eb20abb992182deb0747ca9, codex/ui-redesign-phase4. Scope: approved P5-07, Dashboard source definitions, P5-06. No P5-08a, Phase 8 or B07.

Inventory before implementation

There is no standalone Analytics route at the source checkpoint. All existing personal reports live in DashboardView; they are not separate backend capabilities to rebuild. Every read below uses authenticated local principal and the existing tenant-filtered context; non-deleted owner jobs only, except stats.total includes deleted. No admin permission or Pro entitlement is required. No new aggregate or full Jobs download is necessary.

Existing report	Source / actual population	Filters and dates	Representation / actions / empty behavior
Monthly applications / dated responses	GET /jobapplications/analytics	Existing UI 6/12/24 months, month-aligned submitted cohort; responses counted in displayed months only for that cohort	Shared-scale solid/dashed chart, exact monthly table; no date-cohort Jobs link because Jobs cannot represent the same cohort
Current pipeline / raw status distribution	overview.funnel + stats.byStatus	Current non-deleted snapshot; canonical normalization may omit custom stages; raw values retained separately	Bars/counts and raw status links to /jobs?status=...; custom names unchanged, empty canonical series distinct from no raw statuses
Current-stage age	overview.timeInStage	Current occupancy median, most recent matching transition or DateApplied/SavedAt fallback; not completed duration	Stage label, days and sample count; absent sample hides report as before
Source response samples	overview.responseRateBySource	Current Company.Source; top six, includes prospects; responses = flag OR date	All returned sources, response/total and server rate; no invented source drill-down
Company activity	overview.topCompanies	Current snapshot, top eight; same response rule	Counts/rate and supported /jobs?companyId=...; true empty state
Salary groups	overview.salaryInsights	Recorded non-deleted ranges; separate currency/pay period; sample required	Mean midpoint, minimum/maximum and count; missing sample hides report, no FX or inferred salary
Job tags	GET /jobapplications/tags?limit=10	All-time non-deleted tags; no months parameter consumed	Dashboard displays top eight; counts/bars, not verified user skills; no tag drill-down
Monthly tag trends	GET /jobapplications/tag-trends?months=...&limit=5	Selected applied cohort; counts per month per tag	Bars plus exact tables, no hover dependency; true no-series state
Snapshot / attention / onboarding	stats + reminders + existing setup reads	Tracked total, last30 dates, current interviews/offers; ranked reminder feed	Dashboard-specific immediate actions remain on Dashboard, not copied as Analytics widgets

All existing report UI text uses EN/Bokmål translations and locale formatting; raw user/custom labels and returned source names remain raw. Backend reminder prose is not silently translated. Summary and trends are independent resource groups, each with genuine error and Retry; a failed read must not become an empty report. Tags/date-filter API support (from/to) exists but is not exposed by the current UI and is not expanded here: tag-trends/overview cannot share that arbitrary range safely. There are no existing status/source/role/location controls on these reports to consolidate.

Preservation and implementation decision

PRES	Existing behavior → intended P5-07 change	Required evidence
PRES-S01	All eight report families → separate detailed Analytics reading using one shared report owner/calculation	Dashboard regressions plus exact values, period/current scope, partial failures and report inventory tests
PRES-S02	Five Dashboard owner-local flags/onboarding → unchanged Dashboard choices; Analytics details remain readable independently	Hidden Dashboard preferences do not hide Analytics; no Analytics preference writes
PRES-E07–E09	Existing route/provider/auth/theme/locale owners → one authenticated /analytics, Insights item, honest page metadata	Native direct entry, reload/Back, mobile/rail entry, account isolation, all four theme/locale pairs
PRES-J05/J09	Supported status/company report links → same query contract	Native matching IDs after drill-down, browser Back restores Analytics period

The approved plan explicitly permits initial reuse of the report presentation component while retaining Dashboard groups. Use a bounded reporting mode on the existing owner and a thin Analytics route. No parallel metric calculations, providers, requests within a mounted route, new charting library or backend contract. Analytics omits snapshot/attention/onboarding/customization, opens exact-value disclosures and provides more explicit period/snapshot/truncated-list context. Dashboard remains unchanged in functionality and retains access to every report and all five flags.

Only the existing 6/12/24 month control applies to activity/tag trends; labels explicitly exclude current snapshot and all-time tags. Analytics may persist that supported selection in its URL for direct entry/refresh/Back, with unsupported values safely defaulting to 12. No new filter is implied.

Metric definitions and unavailable work

• Window percentage = sum of returned dated-response monthly counts / sum of returned applied monthly counts × 100. Denominator zero means no percentage. It is a recorded application cohort/window, not eventual response probability or all employer replies in the window.
• Source/company percentages use returned response flag OR response date counts divided by current non-deleted jobs in each group, including prospects. Top lists are truncated and not population totals. Current mutable Company.Source is not original import provenance.
• Current-stage age is the server median occupancy age with fallback, not completed-stage duration. Salary average is arithmetic mean of each recorded range's midpoint (a single bound substitutes the missing bound), separated by currency/period.
• Raw status counts are not a historical conversion funnel. Never infer an Unknown count by subtracting canonical stages from another population.
• Roles, locations, sectors, historical conversions, average response time, AI/match/CV/document performance and aggregate admin statistics require separate defined reads/events. Do not activate previously unused DTO averages/medians with an unsupported interpretation.

Implementation and hierarchy

AnalyticsPage is a thin consumer of DashboardView's reporting mode. Activity comes first with exact monthly values initially open, then a two-column current pipeline/raw-status comparison, current-stage ages/source samples, separate salary groups, and company/tag reports. On phones these stack in the same DOM order; exact tables wrap within their width. Native tag-value disclosure also starts open for detailed reading. Every disclosure can still be collapsed/expanded by keyboard; no hover-only information.

The shared owner retains all Dashboard behavior by default. Analytics does not load its reminder feed or setup checklist, read/write Dashboard flags, show its snapshot strip, or offer Dashboard customization. Its five report calls are stats, overview, analytics, tags and tag-trends, once per initial mount across two existing resource groups. Period changes reload only the trends group. Dashboard still makes its six existing calls and retains all controlled groups. There is no duplicate report tree or parallel metric calculation in either mounted route.

Analytics displays all ten tags already returned by the existing bounded read (Dashboard remains top eight). The top-six source/top-eight company/top-five trend limits and exact denominators are visible. Tag sparklines normalize within each tag series, with totals and exact monthly values available; they are patterns within a tag, not an absolute cross-tag height comparison. No source/role/location/date-cohort drill-down is invented. Current company and raw status links retain their supported query fields; their Jobs result IDs are checked natively.

Deliberately unchanged: report calculations, shared-scale SVG plotting, source/company response samples, occupancy-age bars, separate salary summaries and supported Jobs links. No existing report is deleted or replaced with an unsupported percentage. Analytics differs through activity-first ordering, initially expanded exact-value tables/raw-status breakdown, URL-backed period selection, explicit scope and all ten returned tags; it does not duplicate Dashboard's daily attention/onboarding/snapshot widgets. Admin aggregate reporting remains in its authorized Admin owner and is neither queried nor migrated here.

/analytics uses the existing private Shell authentication gate, account-plan/provider tree and lazy route owner. It adds exactly one route and one Insights item; all original routes/aliases remain. Its page title/subtitle/breadcrumbs are localized. Dashboard's historical Home → Analytics → Overview breadcrumb is corrected to Home → Dashboard; its command still opens Dashboard and its hint now describes Dashboard instead of Analytics. All six existing commands remain: no new command catalogue or business action.

Related shared shell boundary

The first rendered review showed the new Insights/current row partially clipped below the expanded sidebar's initial scroll viewport, above the fixed Account group. This is a direct consequence of the new destination, not authorization to redesign the shell. AppShell now uses the appropriate existing permanent/temporary Drawer ref to reveal its actual current item with native scrollIntoView({block:'nearest',inline:'nearest'}) when route/collapse/breakpoint/open state or measured header height changes. It does not reset later manual scrolling, move focus, resize/group items or alter callbacks/permissions. Separate refs avoid targeting the hidden keep-mounted mobile copy. A native assertion checks the whole current row against every scrollable ancestor, and existing rail/drawer/focus regressions remain required.

Files and ownership

• src/components/DashboardView.tsx: one existing report/data owner with bounded Analytics composition; same calculations/status semantics/error hook.
• src/views/AnalyticsPage.tsx: thin route consumer, no independent data owner.
• src/App.tsx: route, Insights item and page metadata only; authentication/providers/callbacks untouched.
• src/layout/AppShell.tsx: current-item visibility in the existing scrolling navigation only, justified above.
• src/components/QuickCommandDialog.tsx: Dashboard hint only; command actions/catalogue unchanged.
• src/i18n/translations.ts: nine localized UI labels/context/error messages per language, no raw status/source mutation.
• src/analytics-presentation.test.tsx, e2e/analytics-presentation.spec.ts: scoped regressions. No backend/schema/configuration/CI/dependency files changed.

Regression protection

Five new frontend cases cover complete report/value access independent of Dashboard hidden flags, URL period/Back and no snapshot refetch, genuine snapshot failure with independent trends/Retry, empty/unsupported-period safety without an invented denominator, and exact existing status/company destinations. Existing Dashboard/daily-control-loop/workflow-trust-signals, command and shell tests remain in force. Initial regression run intentionally failed because the new route consumer did not yet exist; a report-name fixture was then aligned to the actual existing translated headings, without changing report names to satisfy tests.

Two new browser cases use the actual production UI export and disposable local API:

1. Native ordinary owner, six jobs including current/sparse/custom statuses, twelve raw tag categories, missing optional metadata, separate recorded salary data and an application outside the shorter window. Compare all six exact monthly rows, 1/4 dated responses versus 1/6 source response sample, and the 24-month 1/5 cohort; period Back/refresh, matching custom-status/company Jobs IDs, unchanged hidden Dashboard flags, expanded/current-row visibility, rail/drawer entry, keyboard disclosure/focus, EN/Bokmål × light/dark desktop/phone, intermediate/minimum/wide widths, and account switch to a fresh owner plus one application. Cross-owner direct job read returns 404.
2. Controlled transport holds/503s for genuine loading and independent resource failures, Retry and zero-history recovery in the actual rendered app. These are synthetic failure fixtures, not production-data/load proof; native values and ownership come from case 1.

Locale fixture setup requires a real 204 preference save from a blank document before mounting the next Analytics view, following P5-06's documented preparation. It does not test or repair concurrent preference writes during Dashboard loading. No retry, forged identity, mocked success or permission bypass is used.

Validation observations retained

The initial full browser run found one unchanged light/English desktop attachment-confirmation case failing to focus Delete after Shift+Tab from Cancel. The trace proves the initial Cancel focus assertion passed and the keyboard event occurred shortly after opening; the unchanged ConfirmProvider also unconditionally focuses Cancel in its transition onEntered callback. A transition/focus race is a plausible explanation, not an instrumented proof of callback timing. Neither this provider nor the inherited test was changed, skipped or given a retry. This observation is P5-07-O01, separate from Analytics; a later passing run does not certify the race fixed. Preserve its sanitized initial summary and compare the final full run independently.

Rendered review also caught a reused Dashboard-specific snapshot error description on Analytics. The final localized Analytics description now identifies current job reports and independently loading activity/tag trends; both unit and browser assertions cover it. Final validation and error-state capture are repeated after that correction.

Accessibility and rendered evidence

The detailed destination uses the existing shared PageSection headings, native links/disclosures, row/column-header tables, semantic theme/status colors and ViewStateNotice error/Retry owners. Noninteractive charts and metric rows are not focusable. Canonical status color is supplemented by labels and values; raw custom status names remain available separately. Monthly and tag values do not require hover, SVG interpretation or color perception. Keyboard checks exercise disclosure toggles/focus, supported report navigation and mobile route-heading focus; unchanged shared dialog/command/shell tests remain part of the full suite.

Scoped axe scans cover the populated English phone page in light and dark mode using the existing WCAG A/AA tags, without disabling rules. The final sanitized summary records violations and incomplete checks separately. These scans do not certify all states/locales or full screen-reader operation. Spoken names, chart/table reading and live/state announcements remain an open manual P5-15 acceptance item; no manual screen-reader, native touch, forced-colors or exhaustive zoom session is claimed.

The final axe summary records zero detected violations in each scan and one incomplete color-contrast node in each: .MuiBadge-invisible, whose content was too short for the checker to determine whether it is text. That is an incomplete result, not an automated contrast pass or closure of the broader manual accessibility requirement. Separate keyboard/disclosure and rendered focus evidence pass within their tested scope.

Evidence is generated by the actual production UI export against a disposable local API. Native populated/sparse accounts contain synthetic data, not real user information. Loading/503 screenshots are explicitly controlled transport fixtures. No production request, credentials, cookies/storage snapshots, raw API bodies or traces are committed. The manifest records capture time, viewport, actual full-page dimensions, period, fixture type and individual rendered inspection. There was no pre-existing Analytics screen to manufacture a before comparison; P5-06's retained Dashboard is the prior report-layout reference.

State / evidence	Rendered review scope
Desktop light EN, light Bokmål, dark EN, dark Bokmål	Activity-first hierarchy, exact six-month values, side-by-side current/canonical versus raw statuses, restrained semantic surfaces and readable labels
Phone light EN, light Bokmål, dark EN, dark Bokmål	390×844; stacked reports, wrapping labels/actions and exact values, with no document overflow in the measured browser fixture
Minimum 320, tablet 768, wide 1920	320×740, 768×1024, 1920×1080; twelve-month default after fresh route entry, bounded content width and responsive navigation; no 1024-specific Analytics image is claimed
Keyboard exact values	Six monthly API values match the visible rows; Enter toggles disclosure and a shared ≥3px focus outline remains visible
New owner, one application	No invented percentage for an empty cohort; 0/1 is valid for a real application with no recorded response; missing optional metadata does not become fabricated reports
Loading, snapshot error, trends error, zero history	Progress is distinct from empty data; genuine 503 messages/Retry preserve the successful independent resource; no false conversion value

The focused full-page screenshot includes fixed-shell stitching at its current scroll offset; it proves the data disclosure/focus, not that stitched header placement. Default captures and native current-item visibility assertions separately validate the shell. Detailed tag/month tables naturally make the page longer; disclosure controls retain the ability to condense them. The unequal company/tag column heights on desktop reflect actual detail data, not empty placeholder widgets. Rebalancing those groups is a future presentation decision, not permission to remove report values.

Performance, compatibility and rollback

Analytics performs five existing bounded aggregate reads, not a request per job or a full Jobs crawl; shared shell reminder polling remains unchanged. Its 6/12/24-month series and top-six/eight/ten/five source/company/tag/trend bounds remain the server contracts. Tags are re-read with the existing trends group when the period changes, although their totals stay all-time; optimizing that owner is not needed for this slice. Local fixtures prove rendering and behavior, not production load certification.

All 31 original route patterns are retained and /analytics adds the 32nd. Authentication/providers and Dashboard preference keys remain compatible. If presentation is rolled back, retain /analytics and its meaningful shared report consumer (or a direct compatibility entry to the retained reports); do not blindly remove the new route and produce a 404. No database migration, irreversible data write, dependency or deployment change is involved.

Validation commands and baseline

Inherited: backend 934; frontend 398; browser 79 actual passes + unchanged P2-D01; rate-limit 9, ingress 6, PDF 1, configuration 3. Commands below run through rtk proxy; frontend commands use job-tracker-ui. Native browser and API security suites use separate disposable local API lifetimes sequentially, never production. Browser export configuration is DOTNET_ROOT=C:/Users/Connor/.dotnet, DOTNET_HOST_PATH=C:/Users/Connor/.dotnet/dotnet.exe, PLAYWRIGHT_STATIC_EXPORT=true, PLAYWRIGHT_PREBUILT_EXPORT=true; production export is built with NEXT_PUBLIC_API_BASE_URL=http://localhost:5302/api, CI=false. The browser run has zero configured retries.

Check	Command after rtk proxy	Final result / inherited comparison
Backend	C:/Users/Connor/.dotnet/dotnet.exe test JobTrackerApi.Tests/JobTrackerApi.Tests.csproj -c Release --no-restore --logger 'console;verbosity=minimal'	934 passed, 0 failed/skipped; unchanged
Frontend	npm.cmd test -- --runInBand --silent	403 passed in 80 suites; +5 Analytics cases, no removed tests
Browser	npm.cmd run test:e2e -- --reporter=line,json	81 actual passes (+2 Analytics), unchanged P2-D01 expected failure; 0 unexpected/skipped/retries, 18m12s. Reporter says “82 passed” because it includes the expected failure; this record does not
TypeScript	npm.cmd run typecheck	Passed; historical TS2571 is not current
New E2E fixture typing	npm.cmd exec tsc -- --noEmit --incremental false -p ../.artifacts/p5-07-e2e-tsconfig.json	Passed; ignored local config extends UI tsconfig and includes the new fixture, no committed compiler/CI change
Lint	npm.cmd run lint	Passed, zero warnings
Frontend production export	npm.cmd run build	Passed compilation, TypeScript and static generation after final error-description change
API Release	C:/Users/Connor/.dotnet/dotnet.exe build JobTrackerApi/JobTrackerApi.csproj -c Release --no-restore -warnaserror	Passed, 0 warnings/errors
Rate-limit	npm.cmd run test:rate-limits -- --reporter=line	9 passed; unchanged
Ingress	npm.cmd exec playwright test -- --config=playwright.ingress.config.ts --reporter=line	6 passed; unchanged; local application checks do not close production F04
PDF admission	npm.cmd run test:pdf-admission -- --reporter=line	1 passed; unchanged
Effective Compose/config	python scripts/test-ingress.py	3 passed; unchanged
Supply-chain fixtures	python scripts/test-supply-chain.py	3 passed; unchanged
Ollama evaluation fixtures	python scripts/test-ollama-evaluation.py	5 passed; unchanged
Extension capture URL	node --test browser-extension/tests/capture-url.test.mjs	3 passed; unchanged
Secret scan	python scripts/supply-chain.py scan-secrets	Passed against staged final candidate; no high-confidence credentials
Whitespace	git diff --check and git diff --cached --check	Passed against final candidate
Document/evidence validation	python .artifacts/p4-04-check-doc-links.py plus local screenshot/manifest and sanitized JSON assertions	140 current/root Markdown documents resolve local targets; all 18 inspected PNGs/dimensions and 82 distinct outcomes validate, zero retries/skips; summaries exclude raw rendered HTML/credentials

The initial full browser summary preserves 80 actual passes, unchanged P2-D01 expected failure and one unexpected attachment-focus failure, before the final error-copy correction. It is not reported as a pass. The final complete run is reported independently; passing reruns do not close P5-07-O01. No existing test was removed, skipped, renamed or given a weaker assertion.

The final browser summary records all 79 inherited successful cases still passing, both new Analytics cases passing and the frozen-package case still failing under its original P2-D01 annotation. Final screenshots were retained before the API-only suites replaced working test output; all 18 final files match the individually inspected bytes and their manifest records actual 2026-10-06 capture dates.

F04, P2-D01, P5-02 manual screen-reader acceptance, incomplete axe/manual checks, native Board touch/full-result scaling and P5-06-O01 preference-save 500 remain open. P5-07 does not repair the preference endpoint or certify concurrent locale writes during loading.

No new roles/locations/sectors, historical conversions, response-time averages, AI recommendations or CV/document performance are added. Those require reliable separately approved data contracts. Aggregate Admin reporting, new charts/frameworks, arbitrary analytical date filters and Phase 8 dashboard customization remain out of scope.

Acceptance and stopping point

P5-07 presentation/preservation acceptance is satisfied within its approved scope: 934 backend, 403 frontend (+5), 81 actual browser passes (+2) plus unchanged P2-D01 expected failure; rate-limit 9, ingress 6, PDF 1 and configuration 3. Existing reports, Dashboard controls, original routes and ownership boundaries remain protected. Both builds, TypeScript and lint pass after the final application change; staged repository checks are recorded above. No baseline count was reduced, unexpected failure suppressed or report silently removed.

Scoped keyboard and rendered light/dark EN/Bokmål/responsive checks pass; the two incomplete axe checks and broader manual screen-reader acceptance remain open. P5-07-O01's initial inherited focus failure remains a separate observation despite a passing final full run. This is a local feature-branch presentation checkpoint, not production or universal accessibility approval.

No P5-08a implementation has begun. The recommended next ticket is P5-08a only after its G-WORKSPACE mobile-progress/focus prerequisites are established; package/CV gates and production F04 remain separate. No B07, package/tailoring or CV work is authorized by this Analytics checkpoint.