Sourcing Lab Method: Measure Which Source Actually Finds People the Others Miss
Database-size claims do not tell a sourcer whether a source will add useful candidates. This protocol measures incremental discovery under a shared role brief.
Last reviewed: 2026-09-06
Ask SourcingOS about this page
1. Use the same role artifact everywhere
Freeze strict requirements, flexible criteria, geography, exclusions, and the evidence threshold before testing. If each source receives a different brief, the comparison is not meaningful.
2. Preserve lane-specific discovery
Record raw discoveries by source before applying the global cap. This prevents the first source in the orchestration order from monopolizing the retained slate and hiding later-source contribution.
3. Resolve duplicates before calculating uniqueness
Normalize identities carefully across sources and require confidence before merging. False identity merges artificially suppress unique contribution; missed duplicates artificially inflate it.
4. Calculate retained and unique yield
For each source, report discovered, evidence-fit retained, unique retained, duplicate retained, and cost/credits. Unique contribution should reflect candidates that meet the common evidence threshold and were not found by the other tested lanes.
5. Repeat across role families
A source may dominate software engineering and add little in healthcare or cleared recruiting. Repeat the protocol across role families and geographies before making stack-level conclusions.