Sourcing dashboards are unusually easy to game, because the activity is visible and the outcome is slow. A sourcer can produce an impressive week — 900 profiles reviewed, 300 messages sent — that yields nothing a hiring manager wants to interview. Six weeks later the role is still open and the dashboard is still green.
Five metrics fix that. Each one is a ratio or a rate that gets worse when quality drops, which is the property a useful metric needs.
1. Qualified candidates per week
Definition: candidates you sourced whom the hiring manager agrees are worth interviewing, per active role, per week.
This is the north star, and the word doing the work is agrees. If the sourcer decides what counts as qualified, the metric is self-scored and drifts upward. Make the grading explicit: a weekly ten-minute review where the hiring manager marks each sourced profile yes/no, ideally in the ATS so it is captured.
Healthy range (our estimate): 3–8 per active role per week for a specifiable role in a normal market. Below 2 means the top of the funnel is broken — targeting, market size, or the brief. Above 10 consistently often means the bar is too loose, not that sourcing is exceptional.
The secondary benefit is diagnostic: when hiring-manager agreement is low, you have surfaced a brief problem early, before you have contacted 200 people with the wrong pitch.
2. Reply rate
Definition: replies ÷ candidates contacted, counting any human response including declines.
Reply rate is the cleanest available proxy for whether your targeting and message are relevant. Declines count because a decline still tells you the person read it and it was plausibly aimed at them.
Healthy range: 15–35% for genuinely personalised outreach to a well-targeted list; under 10% signals a targeting or message problem, not a market problem. Detailed benchmarks by channel, seniority, and touch count — plus the levers that actually move the number — are in our outreach benchmarks.
Watch the interaction with volume: reply rate almost always falls as send volume rises, because targeting loosens. A team whose reply rate is stable while volume doubles is either exceptional or miscounting.
3. Sourced-to-interview pass-through
Definition: sourced candidates who reach a real (non-screen) interview ÷ sourced candidates who replied positively.
This catches the failure mode reply rate hides: outreach so broad and flattering that people reply, then fail the first conversation. It is the metric that distinguishes "we generate interest" from "we generate fit".
Healthy range: 20–40%. Persistently under 15% usually means the screening bar sits after outreach rather than before it — you are qualifying with recruiter calls, which is the most expensive possible place to do it.
If you use an autonomous agent, this is the metric that tests whether the calibration loop is working. Track it per role and expect it to climb across the first two to three grading rounds; if it doesn't, the agent is matching keywords rather than your bar.
4. Recruiter hours per qualified candidate
Definition: total human hours spent sourcing a role ÷ qualified candidates produced. Include searching, screening, writing, following up, and rework.
Nobody tracks this and it is the single most decision-relevant number in the function, because it converts sourcing into money. At a fully loaded ~$55/hour, an hour saved per qualified candidate across 100 candidates a year is $5,500 — larger than most licence negotiations.
Healthy range: practitioner reports put manual sourcing at roughly 2–4 hours per qualified candidate; teams running agents report well under one hour, with the residual concentrated in briefing and shortlist review. Both figures are self-reported, so measure your own baseline for two weeks before believing any vendor's claim.
Measuring it is easier than it sounds: ask recruiters to log sourcing time in blocks for two weeks. You are not building a timesheet culture, you are establishing a baseline to test tools against — the same baseline our methodology weights most heavily when ranking.
5. Cost per sourced hire
Definition: (sourcing software + sourcing labour) ÷ hires that originated from outbound sourcing.
The comparison this enables is the one finance actually asks: how does outbound compare with an agency fee at 15–25% of salary, or with job-board spend? For a $150K role, an agency placement is $22K–$37K. If your fully loaded cost per sourced hire is $4K, in-house sourcing is not a cost centre; it is the cheapest channel you have.
Two rules keep it honest. Attribute a hire to sourcing only if first contact was outbound — not if the person later applied through the careers page. And include the software the sourcing motion actually consumed, prorated, not the whole HR stack.
The three metrics to stop reporting
| Vanity metric | Why it misleads | Report instead |
|---|---|---|
| Profiles viewed / sourced | Measures browsing. Rises when qualification gets lazier. | Qualified candidates per week |
| Messages sent | Measures send volume. Rises exactly when personalisation falls, damaging reply rate and employer brand. | Reply rate, and replies per week |
| Database size ("800M+ profiles") | Vendor inventory, not your outcome. Coverage in your niche and contact accuracy are what bind. | Verified-contact rate on a sample of your own roles |
The database-size point deserves emphasis because it drives purchasing decisions. Aggregate profile counts are unaudited vendor claims, and profile counts include stale and duplicate records. The test that matters is small and cheap: take 50 people you have already identified in your niche, run them through the vendor's contact finder during a trial, and measure how many emails are both found and valid. Our contact-data accuracy guide has the full protocol.
How to instrument this in a week
Day 1 — define the funnel stages in your ATS: sourced → contacted → replied → hiring-manager approved → interviewed → hired. Most ATSs support custom stages or tags; the point is that "sourced" and "approved" are distinct, recorded events.
Day 2 — set the source-of-origin field so outbound-sourced candidates are distinguishable from inbound at reporting time. Everything in metric five depends on this one field being reliable.
Days 3–5 — baseline recruiter hours with rough time logging in blocks. Two weeks is better, one week is enough to argue with a vendor.
Week 2 — start the weekly hiring-manager grading ritual. Ten minutes per role. This is the highest-leverage habit in the whole system, and it is free.
Once the five numbers exist, tool evaluation stops being a matter of taste. You can run a pilot against a control and know within a month whether the product moved qualified candidates per week and hours per qualified candidate — which is exactly the test our buyer's framework is built around, and the evidence you will want at renewal.