components/export-button.tsx · issue #15, status: "done", statusLabel: "Done" },
+ { pp: "C · Visualization", name: "Recharts components, types chosen per metric", ev: <>retention-risk-chart.tsx, risk-alert-chart.tsx, readiness-assessment-chart.tsx, status: "done", statusLabel: "Done" },
+ { pp: "B · Definitions", name: "Tooltip primitive exists; no centralized glossary yet", ev: <>components/info-popover.tsx, status: "partial", statusLabel: "Partial" },
+ { pp: "D · Methodology", name: "How predictions are made — surfaced in-app", ev: <>app/methodology/ route, status: "done", statusLabel: "Done" },
+ { pp: "D · FERPA", name: "RBAC, audit log, FERPA-compliant identity resolution", ev: <>Issues #67, #75, #77, #78 (closed), status: "done", statusLabel: "Done" },
+ { pp: "D · Automation", name: "Self-service upload (PDP, AR, student, course)", ev: <>Issue #86 (closed) · components/upload/, status: "done", statusLabel: "Done" },
+ { pp: "D · Explainability", name: "SHAP narrator — fine-tuning epic in progress", ev: <>Issues #97 — #103 · branch fine-tuning/97-shap-narrator-task-type, status: "partial", statusLabel: "In flight" },
+ { pp: "A · Validation", name: "Upload exists; human-readable validation report missing", ev: <>Addressed by new issue #110, status: "partial", statusLabel: "Partial" },
+ { pp: "C · Filtering", name: "By cohort, term, demographic, credential type", ev: <>Issues #66, #81 (closed), status: "done", statusLabel: "Done" },
+ { pp: "C · Query", name: "Natural-language query against the data", ev: <>lib/prompt-analyzer.ts · issues #17, #61, #88, #90, status: "done", statusLabel: "Done" },
+ { pp: "E · Knowledge", name: "Self-service upload reduces single-person dependency", ev: <>In-app submission runbook missing — addressed by #111, status: "partial", statusLabel: "Partial" },
+]
+
+const TIERS: Array<{
+ pri: "p0" | "p1" | "p2"
+ pillLabel: string
+ h: React.ReactNode
+ twoCol?: boolean
+ cards: Array<{ num: number; tag: string; title: string; body: React.ReactNode }>
+}> = [
+ {
+ pri: "p0",
+ pillLabel: "P0 · Differentiators",
+ h: <>Match the loudest complaints. Datathon-eligible.,
+ cards: [
+ { num: 105, tag: "Pain B · Definitions", title: "Metric definitions glossary with IPEDS & state-compliance cross-walks", body: <>Every KPI surfaces a tooltip with PDP, IPEDS, and state-compliance equivalents. Centralized /glossary page indexed by metric. Markdown source-of-truth, versioned with the code. },
+ { num: 106, tag: "Pain C · Export", title: "Presentation-ready chart export (PNG / PDF)", body: <>Every chart exports as a polished image with title, definition, source, and date stamp baked in. Eliminates the manual Excel-rebuild workflow IR staff perform daily. },
+ { num: 107, tag: "Pain A + D · Lineage", title: "Data lineage view — “where did this number come from”", body: <>Click any number → see source rows, upload event, transformations, and timestamps. The single highest-leverage gap from the session: directly answers trust + governance + differentiation in one feature. },
+ ],
+ },
+ {
+ pri: "p1",
+ pillLabel: "P1 · Governance hardening",
+ h: <>Table-stakes for institutional adoption.,
+ cards: [
+ { num: 108, tag: "Pain D · Transparency", title: "AI Transparency Page", body: <>Per-model disclosure: features used, training data source, homegrown vs. third-party, where data flows when invoked, retention policy. Reviewable independently by institutional IT & legal. },
+ { num: 109, tag: "Pain D · Sensitive populations", title: "Sensitive-population safeguards", body: <>Per-institution feature-exclusion lists. Context warnings on small sub-populations. Audit log entries for any query touching flagged groups. Demoable, not just claimed. },
+ { num: 110, tag: "Pain A + E · Validation", title: "Upload validation report — diff vs. last upload", body: <>Row-level errors, field coercions, dedup decisions, anomaly flags (“3 campuses dropped from this upload”). Readable by non-technical IR staff. Survives the “person retires” scenario. },
+ ],
+ },
+ {
+ pri: "p2",
+ pillLabel: "P2 · Process & institutional fit",
+ h: <>Lower urgency, higher institutional gratitude.,
+ twoCol: true,
+ cards: [
+ { num: 111, tag: "Pain E · Process", title: "Submission runbook generator", body: <>Tool records the upload steps + field mappings that worked, then generates a printable runbook so a successor can replicate without tribal knowledge. Replayable on new files. },
+ { num: 112, tag: "Pain F · Datathon coordination", title: "Institution-grouping helper — shared SIS + shared goal", body: <>Operational artifact, not a user feature: cross-reference AASCU’s SIS list with stated institutional goals; output a candidate cohort matrix for the fall datathon. },
+ ],
+ },
+]
+
+const pillClass = (pri: "p0" | "p1" | "p2") =>
+ pri === "p0" ? styles.pill0 : pri === "p1" ? styles.pill1 : styles.pill2
+
+const statClass = (s: "done" | "partial") =>
+ s === "done" ? styles.statDone : styles.statPartial
+
+export default function AASCUFullPage() {
+ return (
+ + Two AASCU intermediaries described, in their own words, why the Postsecondary Data Partnership dashboard fails the institutions they support — and where a tool built around provable, presentable, governed outputs would land. +
+The intermediaries don’t need another dashboard. They need outputs they can present upward, prove the provenance of, and govern responsibly when sensitive student populations are involved.
+Two AASCU intermediaries described pain in three layers: PDP dashboard quality — inaccurate cohort numbers, buried definitions, wrong chart types, no data export — forcing IR staff into manual Excel rebuilds; AI/governance requirements — FERPA-plus expectations including data lineage, transparency, and explicit safeguards for sensitive student populations; and institutional process gaps — submission knowledge that lives with one person and varies wildly across campuses.
+Our tool already addresses a meaningful share of layer one — CSV export, sane chart types, NLQ, methodology page — and is in-flight on layer two via the SHAP-narrator work. The biggest unaddressed gaps are a definitions glossary with IPEDS / state-compliance cross-walks, presentation-ready chart export, a data-lineage view that proves where each number came from, and AI transparency + sensitive-population safeguards as a precondition for institutional adoption.
+Andres (IR-focused) and Dr. Prateek (data-engineering-focused) emphasized different layers, but converged on a common diagnosis: institutions can’t trust, can’t present, and can’t govern what PDP gives them today.
+Why on earth would they invest any time in learning this new tool?+ — Intermediary, on PDP’s compounding friction · 21-min recording, 03:42 +
codebenders-dashboard.Mapping each pain point to the closed PRs and shipped components. Status reflects the state of the codebase as of this report.
+I just went by each data point and copied that number into the Excel spreadsheet.+ — Dr. Prateek, on rebuilding PDP outputs by hand · 04:58 +
Each issue links back to a specific quote or recurring complaint from the discovery session. Out-of-scope items are listed at the end and intentionally not filed.
+{c.body}
++ A two-page condensation of the AASCU intermediary discovery session and the eight issues filed against the codebenders-dashboard backlog. +
++ Intermediaries describe pain in three layers: PDP dashboard quality (charts, definitions, exports, accuracy), AI & data governance (lineage, transparency, sensitive populations), and institutional process (knowledge silos, inconsistent submission rituals). +
++ The tool already addresses meaningful parts of layer one and is in-flight on layer two via the SHAP narrator. The biggest unaddressed gaps are a definitions glossary, presentation-ready chart export, a data-lineage view, and AI transparency + sensitive-population safeguards. +
+{row.desc}
+{iss.desc}
++ Two AASCU intermediaries described, in their own words, why the Postsecondary Data Partnership dashboard fails the institutions they support — and where a tool built around provable, presentable, governed outputs would land. +
+The intermediaries don't need another dashboard. They need outputs they can present upward, prove the provenance of, and govern responsibly when sensitive student populations are involved.
+Two AASCU intermediaries described pain in three layers: PDP dashboard quality — inaccurate cohort numbers, buried definitions, wrong chart types, no data export — forcing IR staff into manual Excel rebuilds; AI/governance requirements — FERPA-plus expectations including data lineage, transparency, and explicit safeguards for sensitive student populations; and institutional process gaps — submission knowledge that lives with one person and varies wildly across campuses.
+Our tool already addresses a meaningful share of layer one — CSV export, sane chart types, NLQ, methodology page — and is in-flight on layer two via the SHAP-narrator work. The biggest unaddressed gaps are a definitions glossary with IPEDS / state-compliance cross-walks, presentation-ready chart export, a data-lineage view that proves where each number came from, and AI transparency + sensitive-population safeguards as a precondition for institutional adoption.
+Andres (IR-focused) and Dr. Prateek (data-engineering-focused) emphasized different layers, but converged on a common diagnosis: institutions can't trust, can't present, and can't govern what PDP gives them today.
++ Why on earth would they invest any time in learning this new tool? ++ — Intermediary, on PDP's compounding friction · 21-min recording, 03:42 +
Mapping each pain point to the closed PRs and shipped components. Status reflects the state of the codebase as of this report.
+components/export-button.tsx · issue #15retention-risk-chart.tsx, risk-alert-chart.tsx, readiness-assessment-chart.tsxcomponents/info-popover.tsxapp/methodology/ routecomponents/upload/fine-tuning/97-shap-narrator-task-typelib/prompt-analyzer.ts · issues #17, #61, #88, #90+ I just went by each data point and copied that number into the Excel spreadsheet. ++ — Dr. Prateek, on rebuilding PDP outputs by hand · 04:58 +
Each issue links back to a specific quote or recurring complaint from the discovery session. Out-of-scope items are listed at the end and intentionally not filed.
+Every KPI surfaces a tooltip with PDP, IPEDS, and state-compliance equivalents. Centralized /glossary page indexed by metric. Markdown source-of-truth, versioned with the code.
Every chart exports as a polished image with title, definition, source, and date stamp baked in. Eliminates the manual Excel-rebuild workflow IR staff perform daily.
+ Open issue +Click any number → see source rows, upload event, transformations, and timestamps. The single highest-leverage gap from the session: directly answers trust + governance + differentiation in one feature.
+ Open issue +Per-model disclosure: features used, training data source, homegrown vs. third-party, where data flows when invoked, retention policy. Reviewable independently by institutional IT & legal.
+ Open issue +Per-institution feature-exclusion lists. Context warnings on small sub-populations. Audit log entries for any query touching flagged groups. Demoable, not just claimed.
+ Open issue +Row-level errors, field coercions, dedup decisions, anomaly flags ("3 campuses dropped from this upload"). Readable by non-technical IR staff. Survives the "person retires" scenario.
+ Open issue +Tool records the upload steps + field mappings that worked, then generates a printable runbook so a successor can replicate without tribal knowledge. Replayable on new files.
+ Open issue +Operational artifact, not a user feature: cross-reference AASCU's SIS list with stated institutional goals; output a candidate cohort matrix for the fall datathon.
+ Open issue +A two-page condensation of the AASCU intermediary discovery session and the eight issues filed against the codebenders-dashboard backlog.
+ +Intermediaries describe pain in three layers: PDP dashboard quality (charts, definitions, exports, accuracy), AI & data governance (lineage, transparency, sensitive populations), and institutional process (knowledge silos, inconsistent submission rituals).
+The tool already addresses meaningful parts of layer one and is in-flight on layer two via the SHAP narrator. The biggest unaddressed gaps are a definitions glossary, presentation-ready chart export, a data-lineage view, and AI transparency + sensitive-population safeguards.
+Validation report on upload addresses the institutional side; PDP-side accuracy is out of scope.
+Tooltip primitive exists. Centralized glossary and cross-walks not yet built.
+Recharts components, types chosen per metric. Done.
+CSV shipped (#15). PNG/PDF export with embedded definitions is the next step.
+Issues #67, #75, #77, #78 closed. FERPA basics covered.
+Methodology page exists. SHAP narrator in flight. Lineage and transparency disclosures unbuilt.
+Self-service upload (#86) helps. Runbook generator would close the loop.
+Hover tooltips on every KPI · centralized /glossary page · markdown source-of-truth.
Title, definition, source, date stamp baked in. Eliminates the manual Excel rebuild.
+Click any number → source rows, upload event, transformations, timestamps. Highest-leverage gap.
+Per-model disclosure — features, training data, provider, data flow, retention. Reviewable independently.
+Per-institution feature exclusion · low-sample-size context warnings · audit log entries.
+Row-level errors, coercions, dedup decisions, dropped-campus flags. Readable by non-technical IR staff.
+Capture what worked → printable runbook · replayable on new files · survives staff turnover.
+Operational artifact — cross-reference AASCU's SIS list with stated goals; output cohort matrix.
+