Data roadmap
Where the data comes from next
The tiered data wishlist, phased into a build plan. Option levels are bundles of phases; each phase below is its own matrix of datasets — descriptions, sources, costs, effort, and who has to act. Budget figures are estimates; the two do-first audits (library licenses, Lightcast academic access) can zero out the purchase tier.
Option levels
Level 1 — Core
P0 + P1 + P2 · ~$500–2,000/yr
Live GA demand trends; tech-skills tab; dollar-priced skills; sequenced degree + purchasable certificate paths; credentials shelf; state-verified external providers; freshness discipline; proprietary collection started.
requires: Nothing — free data + one small order.
Level 2 — Personalized
+ P3 · Likely $0 marginal (in-contract); worst ~$24,500/yr
Everything in Level 1 grounded in the learner’s real record: you’ve-taken-X, audit-backed gaps, real syllabus matching, verified employers/sections.
requires: Registrar/OIT/career-center cooperation.
Level 3 — Gold evidence
+ P4 · Staff governance only
Education→earnings proof at institutional scale (SLDS); OMSCS aggregate outcomes; anticipatory employer demand; near-real-time mirror.
requires: The relationships Level 2 builds.
Level 4 — Enterprise depth
+ Lightcast (5.1) · ~$30–50k+/yr
Vendor-maintained postings/skills/career-pathway depth.
requires: Only if the 90-day usage review says the free stack is insufficient.
Level 5 — Full depth
all of P5 · ~$45–180k/yr worst-case
Firmographics/prospect enrichment, credential registry, B2B content catalogs.
requires: A validated B2B motion.
Phases
Build the demand loop and the trust layer on data we already pay zero for. Postings get persisted instead of discarded; O*NET turns on its ignored Technology Skills surface; every derived artifact becomes self-describing with CI rot checks; usage instrumentation starts recording what features people actually touch. The two do-first external actions (library audit, Lightcast academic email) run here because their answers shape the purchase gate.
- Spec:
- data-integration-scope groups 1–3; exponential-expansion group 0 + instrument tasks
- What improves:
- A live GA demand time-series, a tool-level occupation tab, and silent rot made impossible.
- Data budget:
- ~$500/yr
| ID | Dataset | Description | Used for | Cost (est.) | Effort | Owner action |
|---|---|---|---|---|---|---|
| 1.8 | Dataset freshness gates + CI rot checks | Every derived/static artifact records generated-at, max-age and input vintage; CI fails on silent rot | Trust layer for every other row: vintage labels visible everywhere; committed artifacts cannot rot silently | FREE | S | none |
| 1.1 | O*NET Technology Skills + Work Context | Per-occupation technology (tool/software) skill categories and work-context measures from the USDOL occupational database | New occupation detail tab (category→example groups); technology terms added to the occupation→course matching text so tool-teaching GT courses surface | FREE (existing key) | S | none |
| 1.2 | Job postings corpus + nightly snapshots | Persisted, deduplicated postings from 6 providers (Adzuna, USAJOBS, JSearch, Freehire, Remotive, Arbeitnow): provider, title, company, location, salary bounds, text | Substrate for demand trends, salary premium and employer graph; GA window determined by location gazetteer (remotes excluded) | $0–25/mo (Neon storage growth) | M | none |
| 1.3 | Skill→O*NET element reconciliation | Deterministic alias map (plus low-volume LLM fallback) mapping posting-derived skill terms onto O*NET element names | One skill ontology across postings analytics, occupation skill lists and the gaps radar — makes market signal flow into recommendations | $0–10/mo (small LLM batches on unresolved set) | M | none |
| 2.1 | Library A-Z license audit | Self-serve check whether GT Library already licenses Lightcast, Data Axle/Reference Solutions, IBISWorld-class databases | Decides whether Tier 5 purchases are ever needed; may free firmographics + the entire postings/skills stack | FREE (30 min) | XS | ANYONE today — DO FIRST |
| 2.2 | Lightcast academic license | Academic/CTE-licensed access to posting analytics + skills taxonomy (universities frequently comped) | May free the turnkey stack; informs the purchase memo with negotiating knowledge | Commonly FREE for academic use | XS | SEND the email — DO FIRST |
| 7.1 | Usage instrumentation of all analytics views | Track usage of premium/trend/outcomes/offering views | Supplies the evidence every Tier-5 buy/don’t decision consumes | FREE | S | build first (task 7.2) |
machine-sortable: docs/data-wishlist-matrix.csv · spec-of-record: openspec/changes/