Sourcing candidates who aren't applying

The people you want are not applying

The candidate worth a fee is almost never the one who applied. They are employed, decent at their job, and not reading job boards. Sourcing is the craft of finding them, and it is the half of recruiting under the most pressure from automation, which means doing it with judgment is also how you stay worth your fee.

Four sources, in the order a new agency should work them.

Warm list first: every past conversation is a warm lead, even a year old. Corpus operators are emphatic on this: before hunting fresh names, mine the people who already answered you once. They answered because something in their career was moving, and that motion may have matured.

Targeted search: LinkedIn's free search with boolean strings covers more ground than beginners expect. The craft is narrowing:

("operations manager" OR "plant manager") AND (food AND manufacturing) AND (Ohio OR "greater Cincinnati") -recruiter -hiring

The pattern: title synonyms, industry terms, location, and exclusions for fellow recruiters. Three to five different searches per role, each slicing the problem differently, beats one search scrolled deeply.

The common-denominator map: a trick from the corpus material. Study the client's current team and find the pattern, where their best people came from, which companies, which schools, which adjacent industries. Then search for people from those exact feeders. It works because hiring patterns are real, and it makes your shortlist look eerily well-matched on the first call.

Referral chains: every screen call ends with two questions: who else is excellent at what you do, and who on your team is quietly ready to move. Candidates refer people like themselves. A single warm screen call executed well seeds the next three.

The numbers that predict placements

Recruiting runs on funnel math, and the corpus benchmarks, drawn from in-house tech recruiting programs, are a serviceable starting picture of the ratios. Read them as operator benchmarks, not physics:

| Stage | Benchmark | |---|---| | Outreach to build a healthy pipeline | About two weeks, on the order of 150-200 contacts if sourcing is the main job | | Response rate | Roughly 13-25% on LinkedIn-style outreach, varying by role and seniority; cold email runs far lower | | Interested candidates delivered per week | 3-6 to the hiring manager | | Candidates advancing past the hiring-manager screen | Around 65-80% |

The number to manage weekly is response rate, judged against the channel you are actually running. Current benchmark data splits hard by channel: sequences that combine email, LinkedIn, and phone roughly double replies, running about 18 percent single-channel and 34 percent combined, while pure cold email is the floor at 3.4 percent on average, top-quartile senders near 5.5 (Pin sourcing benchmarks). The corpus rule stays blunt once you know your band. A LinkedIn-style sequence replying under roughly thirteen percent has a messaging problem, so change the template, not the volume. A cold-email-only sequence running single digits is sitting at its channel's normal level, and the fix is the channel or the targeting, not another rewrite of the same note. Track the number in your candidate tab, review it every Friday, and iterate the message the way you would iterate a pitch. Two weeks of data is enough to judge a template.

Personalization at batch scale

You cannot hand-write four hundred messages a week, and you should not send four hundred identical ones either. The corpus method threads it: group the list by niche and role, write one message per group, and personalize a single line per person, the detail that proves a human looked, their company's recent news, a specific skill pairing, a mutual connection. The skeleton carries the pitch. The one line carries the sincerity.

Keep the candidate outreach shorter than you think. Passive candidates respond to messages that respect their time and reveal nothing to their employer:

Hi [Name], your background in [specific thing] caught my attention. I am working a [role] search in [niche] that lines up with it, and the base range is [range]. Not sure you are looking, and no details needed if you are not. Open to a fifteen-minute call this week?

Salary range in the first message. It is the strongest reply driver you have, and it filters mismatched expectations before anyone spends a screen call on them.

AI in the sourcing stack

Use the category, not a verdict: AI-assisted sourcing tools that build Boolean strings, draft sequences, and rank profiles against a job description. The industry consensus through 2025 is that these tools compress the mechanical work dramatically and do not replace the judgment layer, which is deciding who is actually worth a call and getting them to take it (HeroHunt, BCG). Let tools widen the top of your funnel, and keep the selection and the conversation yours. One caution from operator experience: candidates recognize machine-written outreach at volume, and your response rate, the number you now track weekly, is the scoreboard that tells you when the balance tipped.

Sourcing fills the top of the funnel. Whether any of it converts is decided on a thirty-minute phone call, the one where a profile becomes a person.