Gem's 1.2M-Hire Study: Outbound Wins 8x, US Has 1 Sourcer per 86

If you run a recruiting team in late 2026, you are almost certainly staffed for the wrong workflow. Gem's new benchmarks report, built on 1.2 million hires, says outbound-sourced candidates get hired at 8x the rate of inbound applicants. Meanwhile Greenhouse says applications per posting have hit 254 and applications per recruiter have jumped 412%. The correct response is not a better inbound filter. It is fewer inbound hours.

What Gem's 1.2M-hire study actually says

Direct sourcing produces 11% of hires from just 2.6% of applications, a 4x yield and roughly an 8x hire likelihood versus inbound. That is the headline from Gem's 2026 Recruiting Benchmarks Report, which analyzed 165 million applications, 15 million candidates, and 1.2 million hires from June 2021 to May 2025.

The rest of the study is the mechanism behind that ratio:

• Job boards and company marketing generate about 90% of applications but only about half of hires.
• Referrals convert at 11x inbound. Internal mobility converts at 32x.
• Recruiters now manage 13.4 open roles at once, up 40% from 2021.
• 46% of sourced hires come from rediscovered candidates already sitting in the CRM or ATS, up from 26% in 2021.
• Only 8% of applicants get past the initial screen; 0.5% receive an offer. That is one hire per 200 inbound applications.

The uncomfortable read is that the inbound channel now consumes the majority of recruiter hours to produce the minority of hires, and the ratio is getting worse every quarter.

8x

The AI application flood, in one paragraph

Inbound conversion cratered because applying got cheap, not because candidates got worse. Greenhouse CEO Daniel Chait calls this the "AI doom loop," and the data he cites explains why any inbound-side fix is doomed.

Greenhouse now hosts about 175,000 live jobs at an average of 254 applicants per posting, and per-recruiter application volume is up 412%. Candidates are spending about $20 on auto-apply tools that submit hundreds of Greenhouse applications on their behalf. LinkedIn separately reports applications running at roughly 11,000 per minute, up more than 45% year on year. Chait's own summary is the cleanest framing in circulation:

Everyone's using their own AI to solve their own problem, but it's making the whole system worse.

The implication most teams miss: any employer response that lives inside the inbound channel (smarter ATS filters, AI screeners, voice-agent interviews) is a participant in the loop, not an exit from it. Outbound is the only channel where the employer picks the counterparty, which is what breaks the escalation.

Why the 8x gap will widen, not close

The 8x number is a statement about the collapse of self-selection on the inbound side, not about candidate quality, and the mechanics guarantee the gap grows as long as auto-apply is cheaper than a recruiter's time.

Inbound offer rates fell to 0.5% while $20 auto-apply tools arrived; the denominator got polluted. Every quarter that a candidate can submit more applications for the price of lunch than they could hand-write in a year, the outbound multiplier gets larger by construction. If you are planning 2027 headcount off Gem's 8x today, assume the real 2027 number is higher.

There is a related second-order effect on trust signals. With only 8% of applicants passing initial screen and resumes now trivial to fabricate, hiring managers on the September 2026 HN "Who is Hiring?" thread are openly asking for a GitHub link alongside the résumé. Public code history is the cheapest costly signal available: expensive to fake, free to verify. The same logic explains why Greenhouse's "My Dream Job" flag, where a candidate can nominate one role per month across the platform, correlates with a roughly 5x hire rate across the nearly 500,000 submissions since launch. Both are intent signals surviving in a channel that lost its ability to price attention.

244 applications per job is a staffing problem, not a triage problem

At 244 applications per posting across 13.4 open reqs, a single recruiter is looking at more than 3,200 applications per quarter to produce a small handful of inbound hires. Reallocating even 20% of those hours to outbound is not a preference. It is arithmetic.

The derived math: matching Gem's 11%-of-hires-from-outbound benchmark requires roughly 85 targeted outbound touches per hire, versus 200 inbound applications per hire at the current 0.5% offer rate. That is a 2.4x recruiter-hour efficiency gain per placed hire before you count time saved on screening rejects. And it treats every outbound touch as manual, which no serious team does anymore.

This is where a plain-English search layer earns its keep. Instead of writing boolean strings or ranking 244 résumés, you describe the person you actually want ("staff backend engineer, has shipped a payments system at a series-B fintech, based in the Mountain time zone") and get a ranked shortlist back. That is the exact gap Refolk closes: ask for the people you want across GitHub, LinkedIn, and the open web, and get them back every time.

The rediscovery insight most teams miss

46% of sourced hires now come from candidates already in the company's CRM or ATS, up from 26% in 2021. Nearly half of your outbound wins are people you already met.

The mechanism is boring and important. Every rejected finalist, every silver-medalist, every "we love you, wrong timing" candidate from the last three years is sitting in your ATS with structured notes attached. In 2021 those records were worth roughly a quarter of sourced hires. In 2025 they are worth nearly half. The gap is your competitors quietly working their own graveyards while you refresh LinkedIn Recruiter.

Two practical implications:

1. Before you spin up a cold sourcing project, run the same search across your CRM. If the role is a repeat of something you hired for in 2023, the shortlist is likely already indexed.
2. The candidates you rejected in 2022 have three years of new work history attached. Their GitHub graph, promotions, and job changes since then are the freshest signal you own.

The US is structurally under-sourcered

In Refolk's index of professional profiles, the United States has roughly 110,806 people with a "Recruiter" or "Technical Recruiter" title and only 1,281 with a "Sourcer" or "Technical Sourcer" title. That is one sourcer per 86 recruiters, in a labor market where Gem says outbound produces 11% of hires from 2.6% of applications.

1,281

The concentration is striking. The top employers of sourcers in Refolk's index include:

• Rippling
• Anthropic
• MongoDB
• Verkada
• Zoox
• EvolutionIQ
• Fetch

Five of the top ten sourcer locations are in California. These are not accidents. They are the companies that already believe Gem's thesis and staff for it, and they are clustered where the target candidates live. Everyone else is running a 2021 org chart against a 2026 funnel.

The industry has three ways out:

1. Hire more sourcer titles. This is slow and the talent pool is small (1,281 people).
2. Convert recruiters into hybrid "AI-native recruiter" roles that absorb sourcing. This is what most vendor pitches actually mean.
3. Push sourcing into software so a single recruiter can run outbound at a former sourcer's throughput.

Options 2 and 3 are the only realistic paths at scale, and both require a search layer that speaks plain English instead of boolean. Describe the person, get the shortlist, move to outreach. When one recruiter can produce 85 targeted outbound touches in the time it used to take to triage 244 inbound résumés, the sourcer-to-recruiter ratio stops mattering.

What to actually do this quarter

Stop grading recruiters on inbound throughput and start grading them on outbound-sourced hires as a share of total hires. That is the one metric change that forces the workflow to follow Gem's numbers.

A concrete reallocation plan:

1. Cap inbound triage time per req at a fixed hour budget. Anything past 244 applications gets a bulk decision, not a résumé read.
2. Move the freed hours to CRM rediscovery first (46% of sourced hires) before cold sourcing.
3. Adopt a plain-English sourcing tool so a recruiter can run outbound without a dedicated sourcer, and log every outbound touch to the ATS so the graveyard compounds.
4. For engineering roles, require a GitHub or portfolio link in the outbound message and screen on public artifacts before scheduling. The HN "no GitHub, no read" heuristic is a rational Bayesian filter, not gatekeeping.
5. Report weekly on outbound-sourced share of hires. When it crosses 15%, cut inbound job-board spend and reinvest in outreach tooling.

None of this requires a new headcount plan. It requires accepting that the inbound channel is a polluted denominator and that the employer's only durable advantage is picking the counterparty. Gem's 1.2M hires already ran the experiment. The 8x number is the answer.

FAQ

Is Gem's 8x outbound advantage really about better candidates?

No, and that is why the number is durable. The 8x gap widened because inbound self-selection collapsed under $20 auto-apply tools, not because outbound candidates got smarter. Inbound offer rates dropped to 0.5% while outbound conversion stayed roughly stable. As long as applying stays cheaper than a recruiter's screening time, the multiplier grows mechanically. Planning off the current 8x is conservative.

If I only have one recruiter, where should the outbound hours go first?

Straight into your own ATS and CRM. Gem's data shows 46% of sourced hires now come from rediscovered candidates, up from 26% in 2021. Every finalist you rejected in the last three years has updated GitHub, promotions, and job changes attached, which is the freshest signal you already own. Cold sourcing is the second move, not the first, and a plain-English search layer like Refolk lets one recruiter cover both without a dedicated sourcer.

Why is "no GitHub, no read" showing up on HN hiring threads?

Because it is a cheap costly signal in a channel where résumés are now trivially fabricated. With only 8% of applicants passing initial screen and generative tools producing plausible résumé text for pennies, hiring managers need an artifact that is expensive to fake and free to verify. Public code history clears both bars. The same logic explains why Greenhouse's "Dream Job" intent flag correlates with a 5x hire rate: any surviving signal of real intent is now worth an order of magnitude more than a polished résumé.

How do I justify cutting inbound triage time to my hiring managers?

Show them the derived math. At 244 applications per posting and 13.4 open reqs, a recruiter is triaging over 3,200 applications a quarter for a 0.5% inbound offer rate. Matching Gem's outbound benchmark requires roughly 85 targeted outbound touches per hire versus 200 inbound applications per hire, which is a 2.4x efficiency gain per placed hire before you count screening time. Hiring managers who see that arithmetic tend to stop asking why the résumé pile is not being fully read.