Methodology
How the atlas earns trust
JobShift Atlas keeps evidence, interpretation, and uncertainty visible to the reader.
BLS measures and units
National employment projections use BLS Table 1.2 for 2025–2035. The source reports employment, numeric change, and annual average openings in thousands. Stored job counts equal the source value multiplied by 1,000 and rounded to the nearest job. Wages are 2025 median annual dollars. Years and source units remain attached to every normalized record.
O*NET rated fields
Skills, knowledge, and work activities come from O*NET 31.0. Rows are grouped by occupation and element ID: IM supplies Importance on a 1–5 scale and LV supplies Level on a 0–7 scale. Suppressed values and Level rows marked not relevant are not shown. An element appears once per occupation. Profiles without a complete rated table carry an explicit source-unavailable explanation.
Career Explorer ranking
Explorer ranking is deterministic. When active, shared skills contribute 35%, shared work activities 25%, shared tasks 15%, entered theme terms 15%, and desired-direction terms 10%. Only active evidence inputs are included and their weights are normalized. Education and pay are filters, not ranking points. Qualitative alignment labels replace percentage claims.
Comparisons and career paths
Only explicitly approved comparisons are indexable. Reviewed career paths use source-backed skill, task, activity, preparation, and labor-market evidence plus reviewed editorial rationale. Related work does not imply an easy transition, eligibility, hiring success, or improved pay.
Publication gates
A published occupation must have a verified BLS and O*NET identity mapping, complete dated BLS labor fields with explicit units, a valid relative AI exposure category, source provenance, unique occupational copy, and valid metadata. Mapping exceptions remain out of the public index. Generated data validation runs before build and publication.
Interpretation and AI
Source facts, deterministic derived values, editorial explanations, and optional AI analysis are separate layers. BLS relative AI exposure is not automation probability, worker-replacement risk, an employment forecast, a wage forecast, or a productivity forecast. Morgan answers are constrained to the page evidence and identify uncertainty.