Why multi-vendor exists
Enterprise clients want a single point of contact for staffing — they do not want to manage 12 small agencies. So they sign a Master Service Agreement with one prime vendor (you) at, say, 22% margin, and you subcontract 60–70% of the actual sourcing to 5–10 sub-vendors at 12–15% margin. You take the 7–10% spread for the operational orchestration: deduping candidates, calibrating quality, managing SLAs, presenting unified shortlists, and handling the contractual risk.
The model works when the operational orchestration is tight. It collapses when you cannot tell which candidate came from which vendor, when you pay the wrong vendor for a hire, or when the client realizes they are seeing the same candidates from your "exclusive pipeline" that they saw from another agency.
Vendor calibration: the most underrated step
Most prime vendors onboard sub-vendors with an MSA and a Slack invite. The result is predictable: Vendor A submits CVs that match the JD exactly, Vendor B submits anyone with the right keywords, Vendor C submits whoever they had on the bench. Three different quality bars, all flowing into your shortlist.
Calibration fix: send every new sub-vendor 5 historical requirements with a known "correct" shortlist (the candidates you actually shortlisted and the ones you rejected, with reasons). Have them screen the requirements and submit their top 5 per role. Compare to your historical correct shortlist. Vendors who get 4/5 right on multiple requirements join your active panel. Vendors who get 2/5 do not. This 1-week onboarding loop saves 6 months of bad submissions.
SLA tracking that actually motivates vendors
Most prime vendors track SLAs in a Google Sheet that no vendor ever sees. The vendor has no incentive to hit the SLA because they cannot tell whether they are or not. Two changes: (1) publish a weekly leaderboard to all sub-vendors showing their SLA compliance vs the panel average, and (2) tier vendor allocation — top 30% of vendors get first-look on new requirements (a 24-hour exclusive window), middle 40% get the standard window, bottom 30% get the requirement only after 48 hours.
This single tiering change typically lifts top-vendor submission speed by 40% within 60 days because the top vendors realize their position in the queue is earned weekly, not granted permanently.
Candidate dedup at ingestion
When two vendors submit the same candidate (same email, same phone, same resume hash), you have to dedup at ingestion — not at shortlist time. Deduping at shortlist time means you have already done the screening work twice, and the dispute over "which vendor gets the credit" gets political.
The rule that works: first vendor to submit a candidate (by timestamp) gets the credit. Subsequent submissions of the same candidate are flagged as duplicates and silently dropped. The second vendor sees in their dashboard that their submission was deduped and against which earlier submission. No one debates it because the timestamp is immutable.
The unified shortlist problem
When you present a shortlist of 5 candidates to the enterprise client, the client sees one ranked list — not five vendors' top picks. The orchestration challenge: you need to compare 50 candidates from 10 vendors using the same evaluation criteria so the final ranking is defensible.
Manual approach: a senior recruiter on your side re-screens every submission against the same 42-point evidence rubric. Time-intensive but works. AI approach: an evidence-based screening model (Gen 4 from the AI Candidate Screening guide) re-scores every vendor submission against your client's structured intake. Same evaluation criteria across all 50 candidates regardless of which vendor sourced them. Final unified ranking takes 15 minutes instead of 5 hours.
The contractual leak that catches first-time prime vendors
Single biggest financial leak: ambiguous "candidate ownership" clauses in the sub-vendor MSA. If a vendor submits Candidate X to you, you reject them, and 6 months later you place Candidate X for a different client without going through the original vendor — does the vendor get a fee?
Most standard MSA templates leave this ambiguous, and you find out the answer when you get a legal notice. Fix: explicit candidate ownership window (typically 6 months from submission), explicit exclusion list (candidates the vendor does not own — direct applicants, candidates from your own database, candidates introduced by other vendors first). Spell it out before the first requirement, not after the first dispute.