Sourcing Lab Method: How to Compare Contact Data Providers Without Fooling Yourself
This is the SourcingOS test protocol for a future contact-provider bakeoff. It intentionally publishes the method before publishing any winner so the result cannot be reverse-engineered around a preferred vendor.
Last reviewed: 2026-09-06
Ask SourcingOS about this page
1. Freeze a representative test set
Use a privacy-safe set that reflects the actual recruiting markets being served: technical, cleared/federal where permitted, healthcare, executive, and common corporate roles. Do not cherry-pick profiles known to work well with one vendor.
2. Score fields independently
Measure work-email coverage, personal-email coverage where policy permits, phone coverage, verification state, freshness, provider confidence, and social/profile identity separately. A provider that returns any value should not receive the same score as one that returns a recently verified usable value.
3. Measure truth with recruiter outcomes
Where lawful and operationally available, record bounce, successful delivery, connect, wrong-person, disconnected, recruiter-confirmed correction, and candidate opt-out. Keep these outcomes tied to the specific observation and provider.
4. Track overlap and marginal value
Calculate how often provider B adds a usable contact that provider A did not. A waterfall provider is valuable only if its incremental yield justifies its cost, latency, and rights profile.
5. Publish limitations with the result
Provider quality varies by geography, seniority, industry, recency, and contact type. Report sample composition, time window, failure definitions, and confidence intervals or ranges where appropriate instead of declaring a universal winner.