Topic 5 — Sourcing candidates

Goal: Find and attract the right candidates, including those not actively looking.

Lesson 5.1 — Why anyone goes looking at all

Yara's third week at Northwind Robotics, and she's staring at the applicant pool for a senior backend role Marcus has been chasing for two months. Forty-one applications. She reads every one. None of them are close. Marcus wants someone who's run distributed systems at scale, and the people who can do that are, almost without exception, already doing it somewhere good and not thinking about her job ad.

That's the whole problem in one screen. The best engineers are rarely refreshing job boards.

What Yara is about to learn is sourcing: proactively going out to find candidates instead of waiting for them to apply. An inbound application is someone raising their hand. Sourcing is you walking into the room and tapping the right shoulder.

It matters more in tech than almost anywhere, because of one fact: most strong engineers are passive candidates — employed, content enough, not applying anywhere. If your pipeline is inbound-only, you are fishing in the small pond of people who happen to be unhappy this month, and you are systematically missing the best ones.

An inbound-only pipeline doesn't fail loudly. It just quietly never surfaces the person you actually needed.

Imani, the senior recruiter mentoring her, puts it plainly when Yara shows her the forty-one rejects: "Stop reading the inbox. Go build the pipeline." Sourcing is the engine that fills the top of the funnel — and for the rest of this topic, you're learning to run it.

Lesson 5.2 — The channels (and which one quietly wins)

There are more places to find candidates than Yara expected, and she learns fast that each one behaves differently.

LinkedIn is the dominant sourcing platform, full stop — it's where the largest searchable pool of professional profiles lives. The free version lets you search; the paid LinkedIn Recruiter tier adds advanced filters and InMail, the messages you can send to people you're not connected to. This is where most of Yara's hours go.

Inbound applicants are the people already in your ATS (applicant tracking system) — the pool Lesson 5.1 warned about relying on alone, but still real candidates worth working.

Referrals are the channel Imani tells her to never skip. When an engineer recommends a former colleague, that candidate arrives pre-vetted and pre-warmed. Referrals are consistently the highest-quality, highest-conversion source most teams have — a trusted person vouched for them, and they're far likelier to reply and accept.

Niche communities are where specialists actually gather. For developers that's GitHub and Stack Overflow; for designers, Dribbble and Behance; plus conferences and the right Slack and Discord groups. The same role lives in different watering holes.

Your own network and "silver-medalist" candidates — strong people who came second for a past role — are gold. They were already good enough to nearly hire, and you've already built rapport.

When Yara maps Marcus's backend req against this list, the plan writes itself: referrals from his current team first, then GitHub for people doing the exact work, then LinkedIn to round it out. Different role, different mix.

Lesson 5.3 — Search by what they can do, not what they're called

Yara's first instinct is to type "Senior Backend Engineer" into LinkedIn and hope. She gets a flood of mismatches — and misses people whose title is "Member of Technical Staff" or "Software Engineer III" doing exactly the work she needs.

Here's the trap: titles vary wildly between companies, but skills don't. A staff engineer at one shop is a senior at another and a "tech lead" at a third. Searching by job title alone is searching by a label that means nothing consistent.

So you search by skills and experience instead — and this is exactly where knowing the tech (Topic 3) pays off. Yara doesn't need to write Python, but knowing that Marcus's stack is Python, PostgreSQL, and Kubernetes lets her target the real signal: the languages, tools, and seniority that define the job rather than the noun on someone's badge.

This is also why Boolean search exists, which is the next lesson. The reason it works is everything in this one: you're describing a capability, then letting logic find everyone who has it regardless of what their title happens to say.

Lesson 5.4 — Boolean and X-ray: the precision tools

Imani sits Yara down for the skill she promised would change her week.

Boolean search combines keywords with logic operators — AND, OR, NOT — plus quotes and parentheses, to make a search precise instead of a shotgun blast. The operators do exactly what they say:

• AND — both terms must appear (narrows).
• OR — either term counts (widens; great for title variants).
• NOT — exclude a term (cuts noise).
• "quotes" — match an exact phrase.
• (parentheses) — group an OR-cluster so the logic stays clean.

Put together, a real string looks like this:

("software engineer" OR developer) AND Python AND (fintech OR banking) NOT recruiter

Read it aloud: someone called an engineer or developer, who does Python, in fintech or banking, and is not a recruiter. That NOT recruiter is there because otherwise you drown in other recruiters' profiles. Boolean is a learnable skill. An hour of practice and Yara is writing strings that cut a thousand results down to forty good ones. On LinkedIn she then layers filters on top: location, years of experience, current and past companies, listed skills.

X-ray search is the companion move. It uses Google's site: operator to find public profiles on a platform through Google instead of the platform's own search — for example:

site:github.com "backend" "Python"

One important reality, because a paying recruiter needs the truth: LinkedIn X-ray largely broke in January 2024, when LinkedIn de-indexed profile fields (headline, experience, skills) from Google. Searching site:linkedin.com now returns far thinner results than it used to. The practical consequence for tech roles: GitHub became the strongest X-ray surface, with 150M+ developers whose public work is right there to read.

For engineers, GitHub doesn't just list a skill — it shows the actual code. That's a signal a LinkedIn headline can't fake.

Lesson 5.5 — Modern tools, and what they don't replace

By month two, Yara hears the other recruiters mention HireEZ, SeekOut, AmazingHiring, and Gem, and worries she's behind for not using them.

She isn't. These are AI sourcing tools that aggregate candidate profiles from LinkedIn, GitHub, Stack Overflow and 40+ sources into one searchable place, and some will even auto-generate a Boolean string from a job description you paste in. They're real accelerators — they save hours and surface people you'd have missed.

But they support judgment; they don't replace it. A tool can hand Yara a ranked list of 200 "matches"; deciding which fifteen are right for Marcus's team, and writing the message that makes them reply, is still her. The majority of US employers now use some AI in hiring — recent surveys put it around half today and climbing fast, with resume-screening adoption already above 80% — so the tooling is becoming table stakes. And precisely because everyone has it, the human edge is the judgment and the outreach, not the search button.

The lesson Imani drives home: learn Boolean by hand first (Lesson 5.4). Then the tools amplify a skill you have, instead of hiding one you don't.

Lesson 5.6 — Outreach is half the job

Yara finds Devon Asante on GitHub — a backend engineer whose open-source work is exactly Marcus's stack. Perfect on paper. Now she has to make him reply, and Devon, like every strong engineer, gets a dozen recruiter messages a week and ignores almost all of them.

Finding the person is only half of sourcing. Outreach is the other half, and it's where the response rate is won or lost. The principles are essentially sales applied to talent — which is exactly the muscle Yara built as an SDR:

• Personalize. Reference something specific to them so it's obviously not a blast. Yara opens with Devon's actual GitHub project, not "I came across your profile."
• Lead with their gain. What's in it for this person — the problem they'd get to solve, the scale, the team. Not "we're hiring."
• Keep it short. Messages under ~400 characters get materially higher reply rates; the shortest ones outperform the longest by a wide margin. Respect their time.
• Sound human. A warm, real voice beats corporate copy. You're starting a relationship, not running an ad.

The size of the gap is what makes the case. Exact reply rates float around with role, seniority, and how cold the list is, so don't anchor on any single headline number you see in a vendor blog — but the direction is consistent everywhere: personalized, relevant outreach commonly doubles the reply rate of a generic blast. That multiple is the lever, not the precise percentage. And LinkedIn enforces a floor from its side: Recruiter accounts must keep a minimum InMail response rate of roughly 13% (LinkedIn's own published threshold, measured over a rolling window of sent messages) or face restrictions. Blasting low-quality messages doesn't just waste time — it can get you throttled.

The lever in sourcing is the quality of one message, not the volume of a hundred. Blast volume is the amateur's mistake.

Lesson 5.7 — The follow-up most recruiters skip

Devon doesn't reply to Yara's first message. Her old SDR instinct kicks in: one message is easy to miss, and silence is not a "no."

So she follows up — politely, once, four days later, adding one new useful detail rather than just "bumping this." A thoughtful second touch lands a real share of the replies you'd otherwise never get, because the first message simply got buried under standups and PRs.

The discipline is the balance. One or two follow-ups, spaced out, each adding something new. Beyond that you're not persistent, you're a nuisance, and a candidate's lasting impression of Northwind is being set by you, right now, before they've spoken to a single human there. Yara's second message to Devon mentions the specific scaling problem his GitHub project would map to. He replies the next morning.

Worked example — Sourcing Devon end to end

Marcus needs a senior backend engineer. The inbound pool is empty of real fits. Watch Yara run the engine.

Channel mix (5.2). She starts with referrals — asks Marcus's two current backend engineers for names. Two leads, no immediate yes. She moves to GitHub, because for this role public code is the strongest signal.

Search (5.3–5.4). She ignores titles and targets capability. Her X-ray string:

site:github.com "backend" "Python" "Kubernetes"

On LinkedIn she runs a parallel Boolean string, then filters by location and 5+ years, and excludes agency recruiters with NOT recruiter.

Surface (5.5). She also drops the job description into Gem to auto-generate a string and catch anyone she missed — then she decides who's actually right. Devon, from the GitHub search, is clearly the strongest.

Outreach (5.6). Short, personalized, candidate-first: she names his open-source project, ties it to a concrete scaling problem at Northwind, keeps it under 400 characters, and signs off human. No reply for four days.

Follow-up (5.7). One polite second message with a new detail. Devon replies the next morning, curious.

One open req, no inbound fits, and Yara has a top passive candidate in conversation — purely because she went and found him, then wrote something worth answering. That's the whole top of the funnel, run by hand.