# EduInsights data and downloads Source: https://eduinsights.ai/data.md Canonical: https://eduinsights.ai/data Current snapshot: `2026-08-25-longitudinal-r28-a1` Generated from the published release and EduInsights public authorities. The public data page explains the current release, its source families and ontology, and the authenticated download path. Documentation and curated MCP inspection are public; release-file downloads require sign-in. ## Current release - Snapshot: `2026-08-25-longitudinal-r28-a1` - Publication snapshot: `ipeds-publication-2026-08-25-ipeds-2025-26-v1` - Created: 2026-08-25T18:24:00.895627+00:00 - Schemas: `marts`, `core`, `ref`, `bridge` - Published semantic marts: 54 - Rows across published files: 640,996,866 ## Source families - **Accreditation and Professional Standards References** — Version varies; status: `not_in_release` - **American Community Survey Public Use Microdata Sample** — 2024 1-year national person file; status: `included`; release table: `marts.acs_degree_occupation_context` - **Anthropic Economic Index** — 2026_06_26; status: `included`; release table: `marts.detailed_occupation_ai_use` - **Baccalaureate and Beyond Longitudinal Study** — publication 2022241, B&B:16/20 first-look tables A-1 and A-3 through A-7; status: `included`; release table: `marts.nces_bandb_national_cohort_context` - **BLS Employment Projections (National Employment Matrix)** — 2024-34; status: `included`; release table: `core.occupation_outlook` - **BLS Occupational Employment and Wage Statistics (OEWS), all areas** — may-2025; status: `included`; release table: `marts.oews_area_occupation` - **CareerOneStop Occupational License Download** — bulk archive dated 2024-10-14; retrieved and validated 2026-08-19; status: `included`; release table: `core.occupational_license_snapshot` - **Carnegie Classifications 2025** — 2025; status: `included`; release table: `core.institution_carnegie` - **Census Post-Secondary Employment Outcomes (PSEO)** — V4.13.0 / 2025Q4, graduation cohorts through 2021 by participating state; status: `included`; release table: `marts.pseo_program_outcome` - **College Scorecard** — 2026-06; status: `included`; release table: `core.scorecard_institution` - **Common Data Set (per-institution)** — 2025-26; status: `research` - **Credential Engine Registry** — Version varies; status: `not_in_release` - **DAPIP - Database of Accredited Postsecondary Institutions and Programs** — 2026-08-01; status: `included`; release table: `marts.program_accreditation` - **Institution catalog extraction (program pages and course descriptions)** — per-unitid (see data/raw/catalog_extraction//current); status: `not_in_release` - **IPEDS — Integrated Postsecondary Education Data System** — manifest-v2 baseline: 2014-15 through 2023-24, plus explicitly separate newer provisional releases; status: `included`; release table: `marts.program_award_cell` - **Microsoft Working with AI** — commit c94a07c52fb1 (2025-12-22); status: `included`; release table: `marts.microsoft_occupation_ai_use` - **National Labor Exchange Research Hub** — Version varies; status: `not_in_release` - **National Survey of College Graduates** — 2023; status: `included`; release table: `marts.nscg_degree_occupation_alignment` - **NC TOWER** — 2026-08-15 snapshot of NCCCS and UNC public files; status: `included`; release table: `marts.nc_tower_program_outcome` - **NCES/BLS CIP 2020 - SOC 2018 Crosswalk** — 2020; status: `included`; release table: `marts.field_occupation_edge` - **O*NET Occupational Information Network database** — 31.0; status: `included`; release table: `core.occupation_tasks` - **Occupational Requirements Survey** — 2018 and 2023 final; 2025 preliminary; status: `included`; release table: `marts.occupation_requirement` - **Projections Central long-term state occupational projections** — snapshot-date; status: `included`; release table: `marts.state_occupation_projection` - **Quarterly Census of Employment and Wages** — 2025; status: `included`; release table: `marts.qcew_regional_industry_context` - **Quarterly Workforce Indicators** — R2026Q2 / V4.14.0 / qwipu_nc_20260512_1027; status: `included`; release table: `marts.qwi_recent_workforce_context` - **Stanford AI Index Report (underlying chart data)** — 2026; status: `included`; release table: `marts.macro_ai_signal` - **USAJOBS** — Version varies; status: `not_in_release` Status is part of the contract: `included` is in the current release, `not_in_release` is documented but absent, `research` is contextual research rather than a published release table, and `unavailable` means release state could not be established. ## Semantic marts - **`marts.acs_degree_occupation_context` — Acs degree occupation context:** Weighted population degree-field/occupation context. ACS 2024 one-year PUMS weighted descriptive context linking publisher degree-field codes to 2018 SOC occupation, state, age group, and sex. Cells under 30 public-use records have measures suppressed by an EduInsights minimum-cell policy (not a Census flag), and replicate-weight variance is not yet calculated. This is never evidence about a named institution or program. Grain: (source_version, state_fips, first_field_of_degree_code, soc6, age_group, sex_code). Rows: 429,899. - **`marts.aei_conversation_cluster_history` — AEI conversation cluster history:** ConversationCluster AI-use observation. March 2025 Anthropic Economic Index privacy-bucketed conversation clusters at native cluster grain, with the publisher's hierarchy descriptions, prevalence values, collaboration/Thinking ratios, and explicit O*NET task mapping cardinality. A cluster-to-task edge can fan out across occupations. Bucketed prevalence is observed Claude.ai context, not an individual-use estimate or population-wide AI exposure. Grain: (cluster_observation_id, task_id, onet_soc_code). Rows: 788. - **`marts.aei_detailed_occupation_history` — AEI detailed occupation history:** DetailedOccupation AI-use observation history. Complete published Anthropic Economic Index global detailed-occupation history. Claude.ai and first-party API are separate key dimensions, as are release and month; the API channel excludes Claude Code. Values are observed Claude-use empirical_calibration, not universal AI exposure, automation capability, productivity, displacement risk, or a forecast. O*NET specialization suffixes are native and must not be collapsed to SOC-6 without an explicit weighting method. Grain: (source_release_id, aei_source, onet_soc_code, period_start). Rows: 2,758. - **`marts.aei_task_history` — AEI task history:** Task AI-use observation history. Complete Anthropic task-facet history retaining release, Claude.ai versus first-party API channel, period, geography, publisher category/metric, coverage, and source-row provenance. Older releases identify tasks by text and can map one source observation to multiple occupation-specific O*NET task rows; inspect match_method and mapping_cardinality. Missing publisher rows are not zero. This is empirical_calibration of observed Claude use, not universal AI exposure, task capability, or an automation forecast. Grain: (observation_id, task_id, onet_soc_code). Rows: 1,700,596. - **`marts.detailed_occupation_ai_use` — Detailed occupation AI use:** DetailedOccupation AI-use observation. Latest Anthropic Economic Index occupation facet at O*NET-SOC specialization grain, kept separately for Claude.ai and first-party API usage. Automation/augmentation and collaboration values describe observed Claude interactions in the published population; they are empirical_calibration, not universal AI capability, displacement risk, or a forecast. Do not collapse specializations to SOC-6 without an explicit weighting method. Grain: (aei_source, onet_soc_code, period_start). Rows: 1,408. - **`marts.field` — Field:** One row per CIP-6 field of study: taxonomy identity, latest canonical IPEDS completions by award level, institutions offering it, and a career outlook rolled up over the field's mapped occupations (edge-weighted wage / growth / AI exposure, with aggregate_soc bridging for BLS combined codes). Career columns are structural_prior — curricular preparation, NOT observed destinations; career_wage_coverage says how much edge weight has wage data. Grain: (cip6). Rows: 2,143. - **`marts.field_award_benchmark` — Field award benchmark:** CIPField x credential level benchmark. National earnings/debt benchmarks per CIP-6 x Scorecard credential level, aggregated across institutions from program_outcomes_cip6. Medians are weighted by each institution's earnings/debt cohort size (weighted lower median); the underlying values are CIP-4 grain (earnings_grain='cip4'). Use this INSTEAD of aggregating program_award_cell yourself — never SUM or average program medians without cohort weights. Grain: (cip6, credlev). Rows: 7,216. - **`marts.field_occupation_edge` — Field occupation edge:** Structural prior linking fields of study to occupations (CIP2020xSOC2018 crosswalk + EduInsights fallback for high-volume unmapped CIPs). PROHIBITED INTERPRETATION per the ontology: this is curricular preparation, NOT observed graduate destinations or employment probabilities. Weights are equal-split structural shares. Grain: (cip6, soc6). Rows: 5,767. - **`marts.institution` — Institution:** One row per IPEDS institution: identity, Carnegie 2025, campus family (OPEID6 rollup policy: attribute, never sum, OPEID6-grain values across member campuses), admissions/cost/outcomes context, and institutional accreditation summary. IPEDS release stage and Source Release identify whether the current canonical roster is provisional. Grain: (unitid). Rows: 5,985. - **`marts.institution_enrollment` — Institution enrollment:** Institution enrollment facet. IPEDS fall enrollment by facet (all students / undergraduate / graduate / degree-seeking slices) with gender and race/ethnicity breakdowns. Facets overlap — filter to one facet_code before summing across institutions; facet_code '1' is the all-students total. Release stage remains visible because canonical data may use a provisional-only newest cycle. Grain: (unitid, reporting_year, facet_code). Rows: 113,833. - **`marts.institution_enrollment_latest_final` — Institution enrollment latest final:** Institution enrollment latest-final convenience view. Explicit point-in-time convenience table for the newest accepted final or revised enrollment observation. Grain: (unitid, facet_code). Rows: 159,716. - **`marts.institution_enrollment_longitudinal` — Institution enrollment longitudinal:** Institution enrollment Observation. Canonical fall-enrollment history selected by the active Publication Snapshot with final-family preference and provisional-only fallback, joined to the contemporaneous Institution Snapshot and retaining Reference Period, Source Release, coverage, and row provenance. Grain: (unitid, collection_cycle, facet_code). Rows: 1,355,204. - **`marts.institution_enrollment_provisional_context` — Institution enrollment provisional context:** Institution enrollment provisional context. Newer accepted provisional enrollment is kept separate from canonical final history as an explicit audit surface even when those same rows serve as the canonical fallback. Grain: (unitid, collection_cycle, facet_code). Rows: 113,833. - **`marts.ipeds_canonical_source_release` — IPEDS canonical source release:** Canonical IPEDS Source Release selection. One accepted Source Release per component-cycle. Selection precedence is late-revised, revised, final, then provisional; is_provisional_fallback explicitly marks the last case. Grain: (component, collection_cycle). Rows: 134. - **`marts.ipeds_completion_demographic_latest_final` — IPEDS completion demographic latest final:** Completion demographic latest-final observation. Newest final/revised IPEDS completion-demographic cell at its institution-total or award-level scope. It is point-in-time demographic context without a CIP/program key; do not reinterpret it as program-specific evidence. Grain: (unitid, completion_scope, award_level_code, facet_code). Rows: 741,598. - **`marts.ipeds_completion_demographic_longitudinal` — IPEDS completion demographic longitudinal:** Completion demographic observation. Snapshot-selected IPEDS completion demographics. C-B rows are institution totals with no CIP or award key; C-C rows add award-level scope but still have no CIP key. Never attribute these counts to a named program or CIP. Sex, race/ethnicity, gender-identity, and age facets can overlap; retain facet definition, coverage, release, and provenance. Grain: (unitid, completion_scope, award_level_code, facet_code, collection_cycle). Rows: 5,828,885. - **`marts.ipeds_component_coverage` — IPEDS component coverage:** IPEDS Component Coverage. Every expected component-cycle is accounted for by an accepted Source Release or an explicit non-observed Coverage State. Grain: (component, collection_cycle). Rows: 144. - **`marts.ipeds_constant_dollar_series` — IPEDS constant dollar series:** IPEDS Constant-Dollar Series. Derived monetary series that keeps source-native nominal values alongside constant values, named price index, index release, and base Collection Cycle. Grain: (unitid, measure_id, collection_cycle). Rows: 136,234. - **`marts.ipeds_distance_enrollment_composition_latest` — IPEDS distance enrollment composition latest:** Latest canonical institution distance-enrollment composition. It selects the newest accepted Collection Cycle, including a provisional fallback, rather than ranking publisher status dates or Reference Period labels. Use the longitudinal composition surface for change over time. Grain: (unitid, survey_period, student_level_code). Rows: 21,810. - **`marts.ipeds_distance_enrollment_composition_longitudinal` — IPEDS distance enrollment composition longitudinal:** Institution distance-enrollment composition by collection cycle. One row keeps the published total and each distance-enrollment category with its own coverage state. The total remains available when a category was not collected; unavailable cells are never converted to zero. Same-cycle institution attributes define the governed comparison population. OPEID-6 groups are federal-aid administrative families, not evidence that member colleges share governance or operate as one institution. Grain: (unitid, survey_period, student_level_code, collection_cycle). Rows: 383,810. - **`marts.ipeds_distance_enrollment_latest_final` — IPEDS distance enrollment latest final:** Institution distance-enrollment latest-final observation. Newest final/revised IPEDS distance-enrollment observation for each native institution, survey-period, student-level, and metric cell. This is an explicit point-in-time convenience surface; use the longitudinal mart for change and retain its source release and coverage state. Grain: (unitid, survey_period, student_level_code, metric_code). Rows: 359,976. - **`marts.ipeds_distance_enrollment_longitudinal` — IPEDS distance enrollment longitudinal:** Institution distance-enrollment observation. Snapshot-selected IPEDS fall or 12-month distance-enrollment evidence at the publisher's student-level and metric grain. Exclusive-distance, some-distance, no-distance, and location counts are distinct source metrics; do not sum overlapping categories without their questionnaire definition. Zero is observed, while not_collected, not_applicable, and suppressed remain distinct with release and row provenance. Grain: (unitid, survey_period, student_level_code, metric_code, collection_cycle). Rows: 2,783,710. - **`marts.ipeds_distance_offering_latest_final` — IPEDS distance offering latest final:** Institution distance-offering latest-final observation. Newest final/revised IPEDS distance-offering flag for each native institution, education-level, offering-kind, and questionnaire-variable cell. It is point-in-time institutional context, not evidence of current named-program modality. Grain: (unitid, education_level, offering_kind, source_variable). Rows: 92,846. - **`marts.ipeds_distance_offering_longitudinal` — IPEDS distance offering longitudinal:** Institution distance-offering observation. Snapshot-selected IPEDS institutional distance course/program offering flags at the questionnaire's education-level and source-variable grain. Flags describe institutional offering context, not student participation and not a named CatalogProgram. Preserve value labels, applicability, coverage, release, and row provenance. Grain: (unitid, education_level, offering_kind, source_variable, collection_cycle). Rows: 771,951. - **`marts.ipeds_distance_program_latest_final` — IPEDS distance program latest final:** ProgramFieldGroup distance-availability latest-final observation. Newest final/revised IPEDS C-DEP field/award distance-availability cell. This point-in-time convenience surface remains broader than a named catalog program and must retain native CIP vintage and publisher program level. Grain: (unitid, cip_code, cip_version, program_level_code, metric_code). Rows: 16,754,020. - **`marts.ipeds_distance_program_longitudinal` — IPEDS distance program longitudinal:** ProgramFieldGroup distance-availability observation. Snapshot-selected IPEDS C-DEP distance-program availability at UnitID × native CIP/vintage × publisher program/award level × metric × cycle. It is direct evidence for the published field/award availability cell, not proof of a particular named CatalogProgram. Preserve CIP level, applicability, coverage, release, and row provenance. Grain: (unitid, cip_code, cip_version, program_level_code, metric_code, collection_cycle). Rows: 88,751,892. - **`marts.ipeds_measure_comparability` — IPEDS measure comparability:** IPEDS measure Comparability State. Direct, mapped, or break state between releases. Mapped comparisons expose their transformation; Trend arithmetic is prohibited across a break. Grain: (measure_id, from_source_release_id, to_source_release_id). Rows: 110. - **`marts.ipeds_measure_longitudinal` — IPEDS measure longitudinal:** IPEDS curated measure Observation. Curated, definition-backed IPEDS measures with native units, Reference Periods, Source Releases, coverage states, and Publication Snapshot identity. Grain: (unitid, measure_id, collection_cycle). Rows: 556,964. - **`marts.ipeds_publication_snapshot` — IPEDS publication snapshot:** IPEDS Publication Snapshot. Versioned release-selection manifests. is_active identifies the snapshot used to materialize the current serving artifact; older identifiers remain pin-able. Grain: (publication_snapshot_id). Rows: 1. - **`marts.ipeds_source_release` — IPEDS source release:** IPEDS Source Release. Accepted append-only IPEDS publications with independent Collection Cycle, Reference Period, release stage, publication date, and archive identity. Grain: (source_release_id). Rows: 134. - **`marts.macro_ai_signal` — Macro AI signal:** macro AI context. Macro AI context for narrative framing: Stanford AI Index 2026 indicators (economy, education, policy, public opinion) plus the Anthropic Economic Index augmentation-vs-automation split of Claude usage. Not education-entity data — no unitid/cip6/soc6 keys; join nothing to it, cite it. Grain: (year, indicator, series). Rows: 850. - **`marts.microsoft_occupation_ai_use` — Microsoft occupation AI use:** Occupation AI-assisted-work observation. Microsoft Working with AI v1.1 metrics for U.S. Bing Copilot conversations from 2024-01-01 through 2024-09-30, at detailed 2018 SOC grain. This is product-specific empirical_calibration of observed assistance and must not be described as universal AI exposure, automation, replacement risk, or a forecast. Grain: (soc_version, soc6). Rows: 785. - **`marts.nc_tower_program_outcome` — Nc tower program outcome:** NC TOWER ProgramFieldGroup outcome observation. NC TOWER direct observed cohort outcomes for participating NCCCS and UNC campuses, restricted to the publisher's All-gender/All-race rows for serving. Campus-to-UnitID and CIP-vintage resolution remain unresolved; -1 is suppression, never zero. Campus labels and publisher program rows must not be silently joined to the institution/program spine. Grain: (source_version, source_file, source_row_number). Rows: 1,755,783. - **`marts.nces_bandb_national_cohort_context` — Nces bandb national cohort context:** NCES B&B national cohort context. Baccalaureate and Beyond published aggregate estimates with publisher standard errors at publication-table, characteristic, and measure grain. National cohort statistics only — no institution, CIP, or SOC key; interpret_with_caution and rounds-to-zero publisher flags remain visible, and a blank estimate is never zero. Grain: (table_number, characteristic_sequence, measure_id). Rows: 1,629. - **`marts.nscg_degree_occupation_alignment` — Nscg degree occupation alignment:** Weighted graduate degree-field/job alignment context. NSCG 2023 weighted degree-field, principal-job, job-relatedness, earnings, and salary context. Cells under 30 public-use respondents have measures suppressed by an EduInsights minimum-cell policy, not an NCSES publisher flag. NSCG publisher classifications are not silently recast as CIP/SOC, and this is not a named-institution outcome. Grain: (source_version, survey_year, highest_degree_type_code, highest_degree_field_code, principal_job_occupation_code, highest_degree_job_relatedness_code, sex_code, race_ethnicity_code). Rows: 37,667. - **`marts.occupation` — Occupation:** One row per detailed SOC-2018 occupation: OEWS wages (May 2025), BLS Employment Projections 2024-34, Anthropic Economic Index exposure (observed Claude usage, empirical_calibration — not universal AI exposure), O*NET job zone, task-level AI rollups (n_tasks, ai_task_coverage, ai_task_penetration_wtd — importance-weighted mean AEI penetration over rated tasks), and per-source coverage flags. When wages/outlook are missing check aggregate_soc — BLS may publish the occupation only inside a combined code. Grain: (soc6). Rows: 879. - **`marts.occupation_requirement` — Occupation requirement:** Occupation requirement observation. BLS Occupational Requirements Survey estimates at publisher occupation, requirement, category, datatype, and series grain. The 2018 and 2023 waves are final; 2025 is preliminary. Successive ORS releases are not a time series, and 2018 O*NET-SOC 2010 codes remain distinct from 2018 SOC codes. Grain: (reference_year, series_id). Rows: 236,468. - **`marts.occupation_task` — Occupation task:** One row per O*NET task within an occupation: task statement, importance/frequency/relevance ratings, and Anthropic Economic Index task penetration (observed Claude usage share matched to the task — empirical_calibration, not universal AI capability). Rows with in_onet_ratings=false are AEI-matched tasks absent from the current O*NET ratings file. Grain: (soc6, task_id). Rows: 19,368. - **`marts.oews_all_published_observations` — Oews all published observations:** Occupation employment and wage publisher observation. All published May 2025 OEWS rows across national, state, metro/nonmetro, and all-area packages, including overlapping area/industry surfaces at their publisher grain. Do not sum across source packages. Retain NAICS, area type, suppression, top-coding, source member, row, and SHA-256 when selecting evidence. Grain: (source_release, source_package, source_file, source_row_number). Rows: 651,048. - **`marts.oews_area_occupation` — Oews area occupation:** Occupation employment and wage observation by area. May 2025 BLS OEWS cross-industry detailed-occupation observations for the national, state, metropolitan, and nonmetropolitan packages. Geography type is part of the key. Wage/employment suppression and top-coding remain visible; a suppressed value is not zero. This mart intentionally excludes industry cuts, which remain in oews_all_published_observations. Grain: (source_release, geography_type, geography_code, soc6). Rows: 223,555. - **`marts.program_accreditation` — Program accreditation:** Programmatic (specialized) accreditation from DAPIP, rolled up to the current grant per institution x accreditor x program. cip_family anchors the program to a CIP-4 family with cip_map_confidence (high/medium/low) — an EduInsights mapping, not a DAPIP field. Institutional accreditation lives on marts.institution instead; records that could not be matched to an IPEDS UnitID remain in core.program_accreditation only. Grain: (unitid, accreditor_id, program_id). Rows: 14,877. - **`marts.program_award_cell` — Program award cell:** One row per program at the product's scoring grain. Completions are direct IPEDS observations; earnings/debt are inherited_from_cip4 (College Scorecard field-of-study measures at CIP-4, shared across the institution's CIP-6 children — never SUM them across programs of one institution). IPEDS release stage and Source Release remain visible because the newest canonical cycle may be provisional fallback. Grain: (unitid, cip6, award_level_code, reporting_year). Rows: 273,404. - **`marts.program_award_cell_harmonized` — Program award cell harmonized:** Harmonized ProgramAwardCell Series. Separately labeled CIP-harmonized completion series. Native observations remain in program_award_cell_longitudinal; mapping weights, ambiguity, and crosswalk release are always disclosed. Grain: (unitid, native_cip6, harmonized_cip6, award_level_code, collection_cycle). Rows: 3,031,138. - **`marts.program_award_cell_latest_final` — Program award cell latest final:** ProgramAwardCell latest-final convenience view. Explicit point-in-time convenience table for the newest canonical final/revised completions Collection Cycle. Use the longitudinal table for changes or Trends. Grain: (unitid, cip6, cip_version, award_level_code). Rows: 601,227. - **`marts.program_award_cell_longitudinal` — Program award cell longitudinal:** ProgramAwardCell Observation. Canonical IPEDS completion observations selected by the active Publication Snapshot: late-revised, revised, then final are preferred; provisional is used only when no final-family release exists for that component-cycle. Retains native CIP vintage, Collection Cycle, Reference Period, Source Release, coverage, and row provenance. A reported zero-zero row is an Observation but does not establish a Program Award Cell; positive first- or second-major activity does. Grain: (unitid, cip6, cip_version, award_level_code, collection_cycle). Rows: 3,031,138. - **`marts.program_award_cell_unharmonized` — Program award cell unharmonized:** Unharmonized ProgramAwardCell Residual. Diagnostic residual: longitudinal completion cells whose native CIP version has no crosswalk mapping to the harmonization target. These cells are absent from program_award_cell_harmonized by construction — the build asserts longitudinal = harmonized + unharmonized, so nothing drops silently. Grain: (unitid, cip6, cip_version, award_level_code, collection_cycle). Rows: 0. - **`marts.pseo_earnings_evidence` — Pseo earnings evidence:** ProgramFieldGroup earnings observation. Census PSEO V4.13.0 earnings at the publisher's institution/state or aggregate, degree, CIP level, cohort, geography, and horizon grain. CIP-4 and CIP-2 values are inherited broader-field context, not direct named-program outcomes. Preserve applicability, coverage/status flags, and OPEID-to-UnitID resolution state; suppressed, unavailable, not participating, and zero are distinct. Grain: (source_release, source_file, source_row_number, horizon_years). Rows: 1,869,252. - **`marts.pseo_industry_flow_evidence` — Pseo industry flow evidence:** ProgramFieldGroup graduate-industry flow observation. Census PSEO V4.13.0 graduate employment flows at native institution/state or aggregate, degree, CIP level, cohort, geography, NAICS level, and horizon grain. This is an empirical field-to-industry relationship for participating cohorts, not a named-program fact. Preserve applicability, coverage/status flags, and identifier-resolution state; do not treat suppression or non-participation as zero. Grain: (source_release, source_file, source_row_number, horizon_years). Rows: 126,488,250. - **`marts.pseo_institution_participation` — Pseo institution participation:** PSEO institution participation assessment. Release-specific Census PSEO participation derived from the institution lookup and partner/release metadata, not from the presence or absence of an earnings/flow row. OPEID is preserved as text and resolves to UnitID only at cardinality one; ambiguous participant OPEIDs remain one native row with a null UnitID. coverage_state distinguishes observed participation, not_participating, and unresolved identifiers; statewide coverage estimates retain their estimate year. Grain: (source_release, opeid, unitid). Rows: 6,041. - **`marts.pseo_program_outcome` — Pseo program outcome:** ProgramFieldGroup normalized outcome observation. Normalized query view over Census PSEO earnings and industry-flow evidence without changing publisher grain. Filter outcome_family before choosing measures, and retain cohort, geography, CIP/degree level, NAICS level, applicability, coverage, and identifier-resolution fields. Broader CIP evidence is inherited context, never a direct CatalogProgram outcome. Grain: (outcome_family, source_release, source_file, source_row_number, horizon_years). Rows: 128,357,502. - **`marts.qcew_regional_industry_context` — Qcew regional industry context:** Regional industry employment context. BLS QCEW 2025 quarterly state/county total, NAICS-sector, and NAICS-4 establishment observations at all-size grain. Disclosure, location-quotient, and over-the-year coverage states remain separate; publisher zero-fill for protected employment/wage cells is normalized to NULL while disclosed establishment counts remain available. This is regional establishment/industry context, not occupation demand or graduate outcomes. Grain: (geography_code, ownership_code, industry_code, aggregation_level_code, year, quarter). Rows: 2,744,436. - **`marts.qwi_recent_workforce_context` — Qwi recent workforce context:** Regional workforce-flow context. Most recent twelve quarters from the bounded North Carolina QWI slice at all-demographic/all-firm characteristic state, metro, and county NAICS-4 grain. Publisher status flags remain measure-specific. This is employer/workforce-flow context, not occupation demand or named-program outcomes. Grain: (geography_type, geography_code, industry_code, ownership_code, year, quarter). Rows: 302,709. - **`marts.semantic_catalog` — Semantic catalog:** Catalog of the served marts: one row per mart in this release, naming the entity it describes, what one of its rows represents, and how to read it. The catalog intentionally does not catalog itself, so it has no row of its own. Grain: served mart table name. Rows: 53. - **`marts.state_occupation_projection` — State occupation projection:** State occupation projection observation. Projections Central long-term state occupation projections. Each state retains its own base/projected-year pair and declared source SOC vintage; taxonomy_mapping_method exposes the explicit bridge to SOC 2018. Missing estimates and unresolved codes are not zero. AvgAnnualOpenings is the publisher's combined openings measure; separate exits/transfers are not available and are not inferred. Grain: (source_release_id, state_fips, base_year, projected_year, source_soc6). Rows: 35,352. ## Interpretation and access - Begin with the [ontology reference](https://eduinsights.ai/data/ontology.md) and the table grain, not a convenient name match. - Use [MCP tools](https://eduinsights.ai/connect/tools.md) to resolve IDs, inspect the schema, retrieve program profiles, or run authenticated read-only SQL. - Keep structural crosswalks, observed outcomes, and research context separate in conclusions. - Cite the source family and snapshot `2026-08-25-longitudinal-r28-a1`.