Recruiter compensation benchmarks by metro: every percentile, delivered

Two different products carry that label, and mixing them up is how compensation decks go wrong. Occupational wage distributions are employer-reported: one row per area x industry x occupation, carrying estimated employment plus nine points on the wage distribution, produced from a six-panel sample of roughly 1.1 million establishments. Crowd-reported pay profiles are individual-reported: one page per job x city cross-tab, carrying a median, a 10th-90th percentile band, pay components and the profile count behind them. The first answers "what does the market pay here" with survey discipline and no commission detail; the second answers "what does a senior agency recruiter earn in this city with commission" with self-report bias baked in.

Datadory delivers both as typed rows in one schema-consistent extract: BLS Occupational Employment and Wage Statistics (OEWS) as the distributional core, Payscale Research as the components layer, and the demand-side records that tell you whether a metro's pay band is about to move. Every figure ships with the geography, occupation code and industry cut it was computed against, so nothing arrives as a bare number.

The wage payload is nine percentiles, not an average: hourly and annual mean and median plus the 10th-through-90th band (A_PCT10 through A_PCT90). Concentration fields make the metro comparison analytical rather than anecdotal - JOBS_1000 (jobs per 1,000), LOC_QUOTIENT against the U.S. average and PCT_TOTAL share of area employment - and EMP_PRSE and MEAN_PRSE record estimate precision beside every figure they qualify. The May 2025 reference file holds 413,528 rows across 32 documented columns, with companion archives reaching back to 1997.

The crowd layer adds what a government occupational survey deliberately averages away: base-pay spread, bonus, profit sharing, commission and total-pay ranges, years-in-field experience bands from Entry Level through Late Career, employer and skill cross-tabs, the profile count behind each figure and the computation date it was stamped with.

One row per area-by-industry-by-occupation cell, identical anatomy whether the geography is the nation or one metro:

# recruiter benchmark row -- shape preview; value slots fill in your sample
AREA_TITLE   : <metro or nonmetropolitan area -- the join spine>
OCC_CODE     : 13-1071   Human Resources Specialists
NAICS        : <sector through selected 6-digit industry cut>
EMPLOYMENT   : <estimated jobs held in this area-industry cell>
JOBS_1000    : <jobs per 1,000 -- local concentration>
LOC_QUOTIENT : <local concentration versus the U.S. average>
A_MEAN       : <annual mean wage>
A_MEDIAN     : <annual median wage>
A_PCT10..90  : <nine-point wage band, the spread IS the risk measure>
MEAN_PRSE    : <precision of the mean, %>

Six decisions separate a defensible benchmark from a screenshot:

1. Fix the occupation definition first. Map "recruiter" onto the federal taxonomy before pulling wages - O*NET Database (downloadable) defines all 1,016 O*NET-SOC titles across skills, tasks and work context, so the filter you argue about in March is the same one you ran in January.
2. Keep the industry cut. Agency-side and in-house recruiter pay blend into a misleading average unless the 6-digit NAICS detail stays in the extract.
3. Quote percentiles, never the mean alone. The nine-point band is the product; the mean is its least interesting summary.
4. Add concentration context. LOC_QUOTIENT and JOBS_1000 tell you whether high pay reflects scarce local supply or expensive demand.
5. Overlay demand and employer base. BLS Job Openings and Labor Turnover Survey (JOLTS) contributes openings, hires, quits and layoffs from December 2000 onward across roughly 2,061 series with state estimates for all 50 states plus DC; BLS Quarterly Census of Employment and Wages (QCEW) - Employment Services NAICS 5613 counts the staffing employers themselves - establishments, monthly employment levels and total quarterly wages by county, with location quotients and over-the-year changes pre-calculated into the same rows.
6. Split-check against the crowd layer wherever commission matters, and label which series produced which number.

Name the metros, the occupation codes and the industry cut in a sample request and the extract arrives cut to all six decisions, with the field dictionary attached.

A comp model that only reads wages misses half its inputs, and five sibling records finish the picture inside the same industry slice.

Demand pressure. JOLTS measures the flows - monthly openings, hires, quits and layoffs at national, industry and state level - so a rising quits rate in a metro explains a widening pay band better than any wage reprint can.

Employer base. The QCEW NAICS 5613 slice isolates employment placement agencies, temporary help services and professional employer organizations census-style from unemployment-insurance tax reports, with history reaching back to 1990 - the supply side of who is competing for recruiter labor in a county.

Distress events. California EDD WARN Notices carries notice-level rows - employer, layoff or closure type, affected employee count, county, effective date - the grain at which contracting recruiting shops actually show up.

Visa constraints. MyVisaJobs aggregates Department of Labor H-1B LCA filings into employer-level reports ranked by petition volume with average salaries, filterable by title, occupation, city and state - decisive when the candidate needs sponsorship.

Ceiling checks. The USAJOBS API returns structured announcement records including grade and salary range for currently open federal positions worldwide - a real-time ceiling reference for government-adjacent recruiter roles.

Function metrics. SHRM HR Benchmarks rolls organization-level operating metrics - recruiting workload and talent-management measures among them - into percentile distributions by industry and size band, answering questions about function efficiency rather than geographic pay.

Six records, six different questions - and choosing among them decides what your benchmark can claim. The table below lays out the layers; the demand-versus-pay trade-off between the two BLS flagship records is argued line by line in JOLTS vs BLS OEWS.

Four limits belong in any methodology note, and all four favor teams that know them in advance.

Wages exclude the variable half. The occupational figures cover base wages only - no overtime, tips or bonuses - which is exactly why commission-heavy agency recruiting roles read thin against them and why the crowd layer's bonus and commission components exist at all. Quote the right series for the role shape.

Self-reported profiles skew. Crowd figures come from people looking up their own pay, tilted toward respondents with an axe or a negotiation in progress, while the occupational series is employer-reported and probability-sampled. Expect the two to disagree about the same metro by construction, not by error - anchor the published number to the occupational percentiles and treat the crowd band as directional corroboration with components attached.

"Recruiter" is not a federal category. Recruiting sits mostly inside 13-1071 Human Resources Specialists, blended with HR generalists; the industry cut is the only lever that pulls agency recruiters out cleanly. Say which cut you used.

Reference dates run behind live markets. The current occupational reference is May 2025, so a mid-2026 offer decision leans on year-old wage levels - fine for structure, riskier in fast-moving metros. Between references, JOLTS flows and the weekly-read ASA Staffing Industry Research measures (ASA Staffing Industry Research) carry the momentum signal, and every Datadory extract ships with its vintage stamped so nobody discovers the gap during review.

Not at the same occupation-by-metro grain, but four open series get close enough for country-level work. ONS People in Work organizes hundreds of UK labour-market datasets across earnings and employment types with regional breakdowns; Destatis Germany Labour Market Statistics publishes German employment and business-services staffing series with Bundesland breakdowns; EURES European Job Mobility Portal exposes roughly 2.05 million live vacancies across 31 states searchable in 24 languages; and World Bank Jobs Data covers hundreds of labor indicators for roughly 265 economies with annual observations from 1960 to 2025.

For market context above the wage line, World Employment Confederation Statistics sizes private employment services across roughly 40 countries and six service segments - market size, placements, penetration rates and agency counts. Penetration rate is the number that tells you whether agency recruiters are scarce or abundant in a given country, which is the supply-side driver a metro comp model usually leaves implicit.

Five personas get outsized value, and each works a different corner of the stack.

Compensation and talent-acquisition analysts set offer bands and annual merit structures metro by metro, quoting percentiles with the reference named. Staffing-firm operators price bench recruiters and expansion markets by pairing wage bands with QCEW establishment counts - what the local competition pays and how many shops there are to lose talent to. Market researchers and consultants build city-rank tables clients republish, where precision fields and industry cuts decide whether the deck survives scrutiny. Data scientists and ML engineers train compensation models on percentile columns rather than flattened averages, joining demand flows and employer counts as features. Journalists, academics and students cite traceable figures - occupation, geography, reference date - back to one row of one record.

The persona-by-persona workflow breakdown lives on data scientists use cases for human resource & employment services.

Because the hard part was never knowing these records exist - it is keeping five heterogeneous series joinable. Occupation codes resolve onto one taxonomy, geography names arrive keyed to identifiers, industry cuts stay intact instead of collapsing into blends, percentiles ship as nine columns rather than one summary statistic, precision and vintage ride beside every value they qualify, and the crowd layer's components land aligned to the same job and location keys.

Metro recruiter pay is one question inside a thirty-record pool. Start with the human-resource-employment-services data hub, which indexes every record in the slice and shows where the pay benchmark sits among them. Then go to the products themselves: BLS Occupational Employment and Wage Statistics (OEWS) for the distributional core and Payscale Research for the components layer. The human-resource-employment-services data guide maps the wider pool with quality scores and use cases, and JOLTS vs BLS OEWS settles when to quote flows and when to quote wages.