Where to source engineering candidates

There are only five channels that reliably produce engineering hires: your own referrals, inbound from your job posts, technical communities, recruiters, and warm talent pools. Everything else (mass LinkedIn InMail, generic job boards, career-fair tables) is one of those five run badly. The real question is not which channel is best in the abstract. It is which one yields the most signal per hour at your stage, because a 4-person seed company and a 40-person Series B have completely different sourcing economics.

This post goes deep on those trade-offs: what each channel costs, what it yields, and when it stops working. If you want the broader hiring sequence first, read the founder playbook for hiring after a raise. Here we stay narrow on sourcing.

What are the five real sourcing channels?

Strip away the tooling and every engineering hire arrives through one of these. They differ on three axes that actually matter: cost per hire, time to the first qualified conversation, and how well the channel scales once you exhaust your founders' personal network.

Referrals: your highest-yield channel, and the first to run dry

Referrals convert better than any other channel and cost the least, because a current engineer is staking their reputation on the introduction. A referred candidate is pre-screened on competence and on fit, two things you would otherwise spend weeks discovering. Internal data across many startups puts referral hire rates several times higher than cold inbound.

The catch is supply. A 5-person team has maybe 200 strong second-degree connections, and the good ones get tapped in the first month. By your third or fourth hire from the same network you are scraping. Referrals are the channel you start with and the one you should never rely on past the first handful of hires. Pay the bonus, ask specifically ("who is the best backend engineer you have worked with"), but do not build your plan on it.

Inbound: free until you account for the filtering

Inbound means engineers find your role and apply. It looks free, and the marginal cost of an application is near zero. The real cost is filtering. A well-known startup can get 300 applicants for one backend role, and someone senior has to read them. If a founder spends 4 hours triaging a pile to find 3 worth a call, that is not free at a founder's hourly value.

Inbound only works if you have a reason for engineers to find you: existing traffic, a recognizable brand, or a job post good enough to travel. Most seed companies have none of the three. This is also where a clean public listing earns its keep. The roles.cc board pulls roles straight from each company's own job site and sorts by funding recency, so a fresh raise puts your post in front of engineers actively reading the recent raises page without you buying traffic. If your roles get no applicants, the cause is usually the post, not the market, and this teardown covers the common failures.

Communities: high signal, low volume, slow to pay off

Technical communities (a specific Slack or Discord, open-source projects, niche conferences, a strong meetup) produce some of the best-fit candidates you will ever meet, because you are seeing people do the work rather than describe it. An engineer who maintains a library you depend on is a known quantity in a way no resume matches.

The economics are inverted from inbound: near-zero dollar cost, high time cost, and low volume. You cannot decide on Monday to hire from a community by Friday. It works when a founder is already a credible member of one, and it does not work as an emergency channel. Treat it as a slow compounding source you cultivate before you need it, not a tap you turn on under deadline.

Recruiters: you pay for speed and pre-filtering, only on a hire

A contingency recruiter charges a percentage of first-year salary, typically 15 to 25 percent, paid only when someone you hire starts. On a $190,000 base that is roughly $28,500 to $47,500 (illustrative, not advice). What you are buying is a pre-filtered shortlist and your own time back, which is the scarcest thing a founder has. For how the fee structure works in detail, see how recruiting agency fees work.

The honest trade-off: a good recruiter compresses weeks of sourcing into a shortlist; a bad one floods your inbox with mismatches and burns goodwill. Contingency is no-risk on cost (no hire, no fee) but variable on quality, so vet the recruiter the way you would a candidate. The case for and against bringing this in versus building it internally is laid out in contingency recruiter vs in-house startup hiring.

The warm talent pool: built once, drawn from many times

The fifth channel is the one most founders skip because it pays off later. A warm pool is a standing set of engineers who already raised their hand: they dropped a resume, they applied to a similar role, they expressed intent. The cost is front-loaded (you build it once), and every later draw is close to free. This is the channel that compounds, and the one a recruiting agency exists to maintain on your behalf.

When you work with roles.cc, the pool is the engine. Engineers drop their CV and get a score on it; we match the pool against companies in their post-raise hiring window. You get a shortlist of engineers who are already interested and already screened, and you pay a percentage of first-year salary only on a hire. No retainer, no per-seat tooling, no monthly sourcing fee.

Which channels should you use at each stage?

Stage changes the math more than anything else. Headcount, budget, and brand all move, and the right mix moves with them.

The pattern: lead with referrals while the network is fresh, layer in a pool or an agency the moment referrals thin (usually around hire 3), and only build dedicated in-house sourcing once you have enough open roles to keep a full-time sourcer busy. Hiring a sourcer for two open roles is overhead; hiring one for fifteen is leverage.

How do you compare channel cost honestly?

Compare fully loaded, not by sticker price. The trap is calling inbound and referrals "free" because no invoice arrives, while counting only the recruiter's fee. A worked example, all illustrative, not advice:

• Inbound for one senior role. 300 applicants, founder triage at 4 hours, plus 6 first calls at 30 minutes each. Roughly 7 hours of founder time, much of it on people who will not advance. The dollar cost is zero. The opportunity cost is a full day a founder did not spend building.
• Recruiter for the same role. 15 to 25 percent of a $190,000 base, so about $28,500 to $47,500, paid only if you hire. Founder time drops to interviewing a pre-filtered shortlist, maybe 3 to 4 calls total.
• Warm pool via an agency. Same percentage-of-salary fee on a hire, but the shortlist arrives already interested and already scored, so the close rate is higher and time-to-offer is shorter.

There is no free channel. There are channels you pay in dollars and channels you pay in founder hours, and at a startup founder hours are the more expensive currency.

Run the comparison in the currency that is actually scarce for you right now. Pre-product-market-fit, founder time is everything and a paid channel that returns hours can be the cheapest option on the board. Post-Series-A with a real budget and a brand, inbound finally starts pulling its weight.

One closing rule: never run on a single channel. Referrals dry up, inbound depends on a brand you may not have yet, and communities are slow. The startups that hire well run two or three channels at once and shift the weight as they grow.