Sourcing Fundamentals
How to build a multi-channel pipeline that doesn't depend on one job board, how to write outreach that gets replies, and how to measure source quality honestly.
Sourcing is the unglamorous half of recruiting. The companies that hire well treat it as a portfolio: multiple channels, measured for quality, with outreach that respects the candidate's time. The companies that don't end up paying 20–25% agency fees forever.
The pipeline model
Think of sourcing as a funnel with a known conversion rate at each stage. Reverse-engineer from your hiring plan: if you need to hire 10 engineers in a quarter and your offer-to-hire is 80%, you need ~12 offers. Working backwards through typical conversion rates tells you how many top-of-funnel prospects you need.
- Prospects identified1000
- Contacted600
- Replied120 (~20%)
- Phone screen60
- Onsite20
- Offer8
- Hire6
Industry benchmarks (LinkedIn Talent Insights, Ashby Benchmarks) put cold outreach response at 15–25%, and prospect-to-hire at ~0.5–1%. Your numbers will differ — track them per channel and per recruiter.
The five channels
| Channel | Best for | Typical cost | Lead time |
|---|---|---|---|
| Inbound (careers page + job boards) | Volume roles, brand-name companies | $0–$500 per posting | Continuous |
| Referrals | Highest quality, all roles | $1k–$10k bonus | 1–4 weeks |
| Outbound sourcing (LinkedIn Recruiter, GitHub, etc.) | Specialist roles, passive candidates | $10k–$15k/seat/yr | 2–8 weeks |
| Community (events, OSS, Slack/Discord) | Specialist + senior roles | Time + sponsorships | Months — long-tail |
| Agencies / contingent | Hard-to-fill or urgent senior roles | 20–25% of first-year salary | 4–8 weeks |
Multiple studies (Jobvite, LinkedIn) show referred candidates have 2–3x higher offer-accept rates and ~25% longer tenure. Build a real referral program: clear bonus, fast feedback to referrers, and a visible 'open roles' digest.
Outreach that gets replies
- 1SpecificityReference one concrete thing about the person's work — a repo, a talk, a project. Not 'I see you're at Acme'.
- 2Why them, why nowTie their work to the role in one sentence. Not 'your background looks great'.
- 3The role in one paragraphWhat the team is, what the role owns, why it matters. Skip the corporate boilerplate.
- 4Comp transparencyPosting bands has been shown to lift reply rates 20–40% (Indeed Hiring Lab, 2024). Some US states now require it.
- 5Low-friction ask‘15 minutes this week or next?' beats ‘apply via this link'.
- 'Hi {{firstName}}, your profile looks impressive'
- Three-paragraph company pitch
- Vague role title
- ‘Let me know if you're interested'
- Sent at 10pm on Friday
- ‘I saw your talk on streaming joins at QCon'
- One paragraph on the team and what they own
- Specific level + band ($X–$Y)
- ‘15 min Wed or Thu?'
- Sent Tue–Thu, 8–11am candidate's time
Sourcing for diversity
Diverse hiring outcomes start with a diverse top of funnel. The intervention is structural: change where you source, who reviews early, and the language you use — not quotas at the offer stage.
- Audit current pipeline composition by source — by gender and (where legal) race
- Add 2+ communities to outreach (e.g., AnitaB, /dev/color, Latinas in Tech, Black Tech Nation, Tech Returners)
- Run job descriptions through bias scanners (Textio, Datapeople) — drop gendered language and inflated requirements
- Require diverse slate at onsite (e.g., Rooney Rule style) — but never tokenize
- Mask names/schools at resume review where legal (UK/EU more often than US)
- Track pipeline composition at every stage to find the drop-off
Forcing demographic targets at the offer stage exposes you to legal risk under US Title VII (post-2023 SFFA implications for private employers are still developing) and demoralizes hires who suspect they were tokens. Intervene at sourcing and review, not at offer.
Source quality, not source volume
‘We got 1000 applicants from LinkedIn' is not a win if zero converted. Track source quality with the same rigor a sales team tracks lead source.
| Metric | Definition | Use it for |
|---|---|---|
| Source-to-screen | % of prospects that pass phone screen | Channel quality |
| Source-to-offer | % of prospects that get an offer | End-to-end fit |
| Source-to-hire | % of prospects that become hires | Channel ROI |
| Quality of hire (6/12 mo) | Performance rating of hires by source | True signal — lag indicator |
| Cost per hire by source | All-in cost ÷ hires | Budget allocation |
When to use agencies
- 1Use them when…Role is genuinely scarce (e.g., niche regulated specialist), urgent (board-mandated), or executive-level where confidentiality matters.
- 2Don't use them when…You have time, the role is common, and the issue is that your in-house sourcing process is weak. Fix the process, don't paper over it.
- 3Negotiate feesContingent fees are 20–25% of base; retained 25–33% paid in stages. Push for fee floors instead of percentages on senior roles.
- 4Run them like a vendorClear scorecard, weekly review, no exclusive-on-everything contracts. Measure them on quality, not submissions.
Common mistakes
- Single-channel dependence (one job board, one recruiter)
- Volume-as-vanity: bragging about applicant count with no conversion
- Outreach templates with no personalization beyond the merge field
- No comp band in the first message
- Treating referrals as 'nice to have' instead of the highest-ROI channel
- Letting agencies set the bar — they optimize for placements, not your culture
- Not measuring quality of hire by source after 6 months
- LinkedIn Talent Solutions — Sourcing benchmarks — LinkedIn
- Ashby — Hiring benchmarks — Ashby
- Greenhouse — Inclusive hiring guide — Greenhouse
- Indeed Hiring Lab — Pay transparency impact — Indeed Hiring Lab
- Textio — Inclusive language research — Textio
- SHRM — Talent acquisition resources — SHRM
- Jobvite — Recruiting benchmark report — Jobvite
Read next
All playbooksThe minimum operating system for fair, fast, and predictive hiring at any company size.
A scorecard turns 'I liked them' into 'they demonstrated X'. Here's how to write one that calibrates a whole loop, reduces bias, and survives legal scrutiny.
What an ATS actually does, the features that matter, how the big vendors compare, and the implementation pitfalls that cost six months.