# Connected dashboard walkthrough

23 October 2026. User requested opening the DeepBrain dashboards and integrating a connected decision/interview walkthrough into the published study guide.

## What changed

The existing /dashboards/ chapter now contains eight connected decision stops and 37 individual operational/explorer chart explanations. Each chart has a decision question, checked fictional finding, strongest caveat, next evidence/action, spoken answer, metric links, exact-population note and data/SQL/Metabase links. A crosswalk covers all 32 definitions and the 27 shared metric views without treating them as unrelated stories.

The interactive selector and previous/next controls show one decision at a time. Context panels follow the decision or opened chart. Separate per-stop practice notes/completion use the existing DeepBrain study-work export/import, preserving old chapter notes. Selecting a new stop cancels speech; read-aloud reads the selected decision. Charts remain available as accessible expandable explanations. Print styling reveals every decision, and existing before-print logic opens details.

The guide links the customer workflow, strategy, metric definitions, PRD, research boundaries, economics, evidence map, SBI answers and practice sessions. Each original decision exercise links back to its corresponding chart explanation. Fifty-three Metabase text explanations across all eight registered DeepBrain dashboards link back into the tour. Layouts, tabs, filter mappings, saved query formulas and databases are preserved. The complete Markdown guide now also includes the later analysis and field-dictionary chapters.

## Verification

All 37 saved question results were refreshed using the actual Metabase query API and independently matched to SQLite. Sixteen numerical assertions protect the highlighted findings, including task counts, recovery, review errors, zero-activity users, assignment interference, MRR, evaluation regression and the negative held-out prediction result. All eight dashboard layouts, tabs and parameter mappings were checked against the pre-change snapshot. Canonical SORI source hashes stayed unchanged.

Browser checks: eight decision selections each show exactly one section and correct context; per-stop notes and completion survive reload; keyboard Enter advances the tour; a chart link opens the correct explanation and context; short-answer copy returns the expected text; import updates the visible per-stop response and export retains both test records. Temporary test responses were cleared. Additional local/live/link/download and browser results are recorded in walkthrough_validation.json, walkthrough_browser.json and live_validation.json after final verification.

Source of authored content: content/metabase_walkthrough.json. Render with scripts/build_walkthrough.py, normally invoked by build_website.py. Refreshed synthetic evidence is in qa/walkthrough_evidence.json; it contains no credentials or real company/customer records. Numerical prose is tied to this fixed practice snapshot, not a live data integration. If the dataset or examples change, refresh and review both the evidence and authored findings.

No company product tests, new factual performance claims or historical resume accomplishments were added. Actual device audibility, physical mobile/screen-reader behavior and native print pagination remain unverified.