Evidence before recruiting claims.
SourcingOS research is designed around reproducible sourcing questions: what a source actually adds, where a search collapses, when the market is genuinely exhausted, which contact-provider signals are usable, and which AI suggestions recruiters accept or reject.
Published methods and benchmark protocols
Unique Contribution Rate
Measure how much net-new evidence-fit discovery a source adds beyond the other sources already tested.
Published methodSearch Exhaustion Framework
Use observable coverage, duplicate pressure, lane yield, and donor-map signals instead of saying “we looked everywhere.”
Published benchmark protocolBoolean Search Benchmark
Compare title, skill, evidence, adjacency, and donor-company queries as distinct candidate-selection mechanisms.
Published market-mapping methodFederal Contract Data as a Sourcing Lane
Use public federal award data to create evidence-backed donor-company maps rather than relying only on recruiter memory.
Sourcing Lab methods and worked search teardowns
These pages publish the methodology or worked example before any measured product result. A methodology page is not presented as evidence that one provider, source, or search strategy has already won.
Search Teardown: RHEL Administrator Near Annapolis Junction With Secret Clearance
A worked example showing how one natural-language brief becomes explicit requirements, discovery expansions, evidence lanes, verification rules, and calibration questions.
The Sourcing Lab · MethodologySourcing Lab Method: How to Compare Contact Data Providers Without Fooling Yourself
A reproducible provider bakeoff for work email, personal email, phone, verification, freshness, overlap, credits, and recruiter-confirmed usability.
The Sourcing Lab · MethodologySourcing Lab Method: Measure Which Source Actually Finds People the Others Miss
A controlled source-contribution experiment using common role briefs, independent lanes, identity resolution, evidence-fit retention, duplication, and incremental yield.
Search Teardown · Worked exampleSearch Teardown: ML Platform Engineer for Production LLM Inference
A worked example for separating model buzzwords from the actual platform, inference, reliability, and production evidence an ML infrastructure role needs.
Search Teardown · Worked exampleSearch Teardown: Senior Data Engineer for a Streaming Platform
A worked example for sourcing data engineers by system architecture, operational ownership, title adjacency, and open technical evidence instead of a shopping list of tools.
Search Teardown · Worked exampleSearch Teardown: ISSO / RMF Talent Near Fort Meade With TS/SCI Requirement
A worked cleared-cyber sourcing example that separates RMF evidence, role/title adjacency, geography, federal context, and clearance discovery from authorized verification.
Search Teardown · Worked exampleSearch Teardown: ICU Registered Nurse With State Licensure Requirement
A worked healthcare sourcing example for specialty evidence, title normalization, location, authoritative licensure verification, and avoiding false equivalence between provider identity and license status.
Worked Talent Map Briefs
These are market-structure hypotheses, not measured candidate-pool counts. They make the title families, donor categories, geography bands, evidence lanes, and scarcity assumptions explicit before SourcingOS reports any live market measurement.
Talent Map Brief: Cleared RHEL / Linux Talent Around Fort Meade
A worked market hypothesis for mapping RHEL/Linux administration talent around Annapolis Junction, Fort Meade, Jessup, Laurel, Columbia, and nearby Maryland corridors.
Worked market hypothesisTalent Map Brief: Cleared DevSecOps & Platform Engineering in Northern Virginia
A worked market hypothesis for cleared cloud/platform talent across the Northern Virginia and Washington federal technology ecosystem.
Worked market hypothesisTalent Map Brief: ML Platform & LLM Infrastructure in the Bay Area
A worked market hypothesis for ML platform, inference, model-serving, and distributed-systems talent in the San Francisco Bay Area.
Worked market hypothesisTalent Map Brief: Healthcare Data & Analytics in Minneapolis–St. Paul
A worked market hypothesis for healthcare analytics, clinical data, payer/provider data engineering, and adjacent talent in the Twin Cities.
Research program
These are planned research questions, not published benchmark results. SourcingOS will only publish product-derived findings when the sample is large enough, the measurement is meaningful, and the reporting can be aggregated without exposing customer or candidate data.
- Source unique-contribution benchmarks by role family
- Duplicate pressure and search-exhaustion behavior
- Which requirements most often collapse a candidate market
- Recruiter acceptance/rejection of AI-generated search expansions
- How hiring-manager calibration changes search yield
- Verified-contact usability by source/provider mix
Research formats
Research rules
- Do not invent sample sizes, benchmark results, or product telemetry.
- Explain the measurement method before drawing a conclusion.
- Separate vendor claims, public evidence, recruiter observations, and SourcingOS-measured results.
- Aggregate product research so individual customers and candidates are not exposed.
- Keep uncertainty, scope, sample composition, and limitations visible.