B. Analyzing the Mismatch

This finding combines NLx national postings and USAJOBS federal postings to ask how well O*NET's occupation-task assignments line up with observed task demand. The question is narrower than simple O*NET coverage: when postings support an off-list occupation-task pair, does that task look broadly general, shared within an occupational cluster, or concentrated enough to flag a candidate addition to a postings-informed O*NET?

The motivation is simple: O*NET is useful because it gives occupations a common task language, but its occupation-task lists are necessarily selective. Postings provide a second view of work as employers describe it. Where validated posting evidence repeatedly places an O*NET task outside its official occupation list, the result is not automatically an error, but it is evidence that a more permissive many-to-many task map may describe work better than a single official home.

The headline occupation denominator is the corrected one: all O*NET canonical tasks for the occupation plus validated off-list tasks. The off-O*NET share is a lower bound because the off-list numerator only includes pairs that were sampled and passed the two-model review.

Why this is provisional. Real task-occupation pairs identified from postings and these review procedures provide evidence on possible candidate maps for occupations, but additional verification has not taken place. Users should also note the limits of occupational coding based on titles.

What The Evidence Can And Cannot Say

The strongest evidence is a combination of four conditions: the task appears in postings for an occupation, the occupation-task pair passed two-model review, O*NET does not list the exact pair, and task spread points to broad generalization, sibling-cluster allocation, or concentrated review candidates.

The page does not claim every noncanonical row is an O*NET error. It identifies places where postings support a more permissive many-to-many task allocation than a single official home. Mixed-spread and low-volume tasks remain leads for review, not stable taxonomy recommendations.

Table 1. Review-retained off-list occupation-task pairs by evidence status

These counts are extracted from the updated workbook's Off_list_pairs tab and sum to 13,893 validated off-list pairs. The table separates task-pair validation from occupation-coding reliability, so readers can see which candidate pairs are strongest before interpreting the task evidence.
No Results

Median off-O*NET task share

35.0%

Review-retained off-list pairs

13,893

Tasks with retained noncanonical homes

5,283

Specialized review candidates

5,393

Examples First

The case-study workbook separates two gates. Gate 1 asks whether the occupation coding looks reliable. Gate 2 asks whether the off-list task pair passed the two-model task review. The best candidate changes clear both gates; weaker examples are still useful, but should be read as audit leads.

Software Developers

O*NET now: 19 canonical tasks.

Postings-informed view: 35 validated off-list additions; 25 are specialized or cluster-shared candidates.

Why stronger: coding is reliable, and examples such as system troubleshooting support and software architecture have high RCA in NLx and USAJOBS.

Show task contrast
Current O*NET examples
  • Analyze user needs and software requirements to determine feasibility of design within time and cost constraints.
  • Confer with systems analysts, engineers, programmers and others to design systems and obtain information on limitations and requirements.
Postings-informed candidates
  • Provide technical guidance or support for the development or troubleshooting of systems.
  • Develop system engineering, software engineering, system integration, or distributed system architectures.
  • Analyze problems to develop solutions involving computer hardware and software.
Dental Assistants

O*NET now: 16 canonical tasks.

Postings-informed view: 20 validated off-list additions; 14 are specialized or cluster-shared candidates.

Why stronger: coding is reliable, and tasks such as sterilizing instruments and preparing patients are occupation-near work.

Show task contrast
Current O*NET examples
  • Prepare patient, sterilize or disinfect instruments, set up instrument trays, prepare materials, or assist dentist during dental procedures.
  • Expose dental diagnostic x-rays.
  • Record treatment information in patient records.
Postings-informed candidates
  • Clean and sterilize instruments, equipment, or materials.
  • Prepare patients for and assist with examinations or treatments.
  • Clean, maintain, and sterilize instruments or equipment.
Firefighters

O*NET now: 30 canonical tasks.

Postings-informed view: 40 validated off-list additions; 13 are specialized or cluster-shared candidates.

Why cautious: this is substantively useful, but the LCA data used for occupational-coding validation do not include the firefighter occupation.

Show task contrast
Current O*NET examples
  • Rescue victims from burning buildings, accident sites, and water hazards.
  • Dress with equipment such as fire-resistant clothing and breathing apparatus.
  • Move toward the source of a fire, using knowledge of fire types, building materials, and property layout.
Posting-evidenced audit leads
  • Participate in firefighting efforts.
  • Communicate fire details to superiors, subordinates, or interagency dispatch centers, using two-way radios.
  • Conduct fire, safety, and sanitation inspections.

Occupation-Level Gap Share

This table asks where reviewed posting evidence adds the most to an occupation's official O*NET task list. "Gap share" means validated off-list tasks divided by official O*NET tasks plus validated off-list tasks. It is a lower-bound audit measure, not a complete occupational task profile.

Median gap share

32.6%

75th percentile

51.3%

Occupations above 50%

259

Review-retained off-list pairs

13,168
Terms. O*NET = official task count. Off-list = validated posting pairs not listed by O*NET. Gap share = off-list / (O*NET + off-list). Specialized = concentrated review candidates, not confirmed errors.
No Results

Task Buckets

Terms. Generic = ≥5 SOC-2 sectors. Mixed = 3-4 sectors. Sector-concentrated = 1-2 sectors. Links = retained off-list occupation-task pairs.
No Results
Browse task-label examples
Terms. Category = spread label. Effective sectors = concentration-adjusted SOC-2 spread. Occs RCA>=1 = occupations where the task is at least as common as baseline.
No Results

Workbook Downloads

O*NET case-study workbook

Reader-facing examples with official O*NET tasks, postings-evidenced additions, coding verdicts, task-pair validation, RCA fields, and recommended candidate changes.

Full O*NET mismatch workbook

Standalone audit workbook with occupation summaries, off-list pair evidence, task summaries, coding reliability, dictionary, and audit fields. Use this as the public review artifact.

Audit Downloads

Show SQL for evidence-status table
Show SQL for headline summary
Show SQL for occupation-gap table
Show SQL for task-bucket table
Show SQL for task-label examples