Recruiting agency: reaching hiring managers in a specific tech stack

The situation

The client was a 9-person tech recruiting agency specializing in Go and Rust backend engineers. Not generalist tech recruiting — specifically, high-scale backend performance roles in the systems and infrastructure space. Their placements averaged $180k–$260k salary, their fee was 25% contingent, and they had roughly 60 active placements per year when I came in.

Their primary customer acquisition channel had been LinkedIn InMail — recruiters spending 5–8 hours/week each prospecting hiring managers on LinkedIn, reaching out with personalized notes, following up manually. It worked, sort of. They were getting ~20 hiring manager meetings/month across the agency, most from the top 2 recruiters. But:

Reply rates on InMail had been declining for 18 months. (This is an industry-wide trend, not a team skill problem.)
The team’s best recruiters spent more time prospecting than recruiting, which was an expensive use of their time.
Junior recruiters couldn’t replicate the senior recruiters’ InMail results, so scale depended entirely on hiring more senior recruiters — expensive and slow.

The founder came to me with a specific question: could email-based outbound supplement LinkedIn prospecting without the recruiters having to do it personally? And could it find stack-specific hiring managers in a way the generic recruiting tools couldn’t?

The “stack-specific” part was the hard part. Tools like ZoomInfo or Apollo give you “VP Engineering at SaaS companies.” They don’t give you “VP Engineering at SaaS companies whose backend is predominantly Go, who are currently hiring senior backend engineers.” That precision is where this agency’s value lived.

The bet

We made two connected bets:

Stack signals from public sources are identifiable at scale. GitHub, job postings, engineering blogs, conference speaker lists, Stack Overflow — all leak information about what language a company runs. If we could automate the signal extraction, we could build a much more precise list than Apollo would produce.
The right opener for hiring managers isn’t “we help companies like yours hire,” it’s a very specific reference to a stack challenge they’re almost certainly wrestling with. Generic recruiter outreach is filtered. Opener about a specific Go concurrency pattern in the context of a specific engineering problem lands differently.

If both bets held, we’d have precise targeting + differentiated messaging — a combination that most recruiting outbound lacks.

Month 1 — Signal mining

Built a stack-detection pipeline. Complicated; worth documenting because this is the kind of thing that gets shipped as a black box and then breaks in month 4.

Sources and what we extracted

GitHub organizations. For public repos, counted language distribution weighted by repo activity in the last 12 months. Company with 60%+ Go in active repos = strong Go signal.
Job postings on Greenhouse, Lever, Ashby. Scraped for language keywords, required experience in specific frameworks (gin, echo, fiber for Go; tokio, axum, actix for Rust), performance-oriented phrasing (“low-latency,” “high-throughput,” “distributed systems”). The phrasing was a better signal than the language alone.
Engineering blogs. Scraped the last 12 months of posts, looked for Go/Rust-specific technical writing from the company.
Conference speaker lists. GopherCon, RustConf, Strange Loop, QCon, Systems Distributed — any engineer from a target company speaking at these was a strong signal.
Stack Overflow company pages. Where present, listed languages and technologies. Less reliable than the other sources but catchable at low cost.

Scoring logic

Each company got a 0–100 Go/Rust score:

30 points for GitHub-based signal
30 points for recent job postings signal
20 points for engineering blog signal
10 points for conference speaker signal
10 points for Stack Overflow listing

Threshold: 50+ points = eligible for the list. Most companies scored under 20. About 2,400 US/UK companies cleared the threshold.

Hiring manager identification

Separately, identified hiring managers at the scored companies:

VPs of Engineering
Directors of Engineering
Heads of Platform / Infrastructure
Engineering managers whose LinkedIn bios mentioned Go, Rust, distributed systems, or similar

Cross-referenced by company. About 5,800 candidate contacts.

Final usable contact list after enrichment (verified email address, recent activity, correct title): ~3,900 contacts.

The whole pipeline ran on a weekly cadence — new job postings, new GitHub commits, new blog posts refreshed the data. Companies moved on and off the list as their signal scores shifted.

Month 2 — Messaging

Four opener variants tested in parallel. The one that dominated:

[First name] — saw [specific engineering artifact: blog post / open-
source commit / talk recording] on [specific topic, e.g. "your team's
approach to tail latency in the Go services at [company]"].

The recruiting pattern I'm seeing in [company-size] teams running that
stack right now is [specific observation about the current Go/Rust
hiring market]. Not pitching a search — curious whether [specific pain
that observation implies] is live for you.

If it is, happy to share what [named peer company with similar stack]
did about it last quarter.

— [recruiter name]
[agency name, 1-line: "We place senior backend engineers in Go/Rust
shops. ~60 placements last year."]

Three features that mattered:

The engineering artifact reference. Required the recruiter (or, more often, a research analyst at the agency) to actually look at what the engineering team had published. Took 5–8 minutes per lead. Not scalable to infinite volume, but the agency didn’t need infinite volume — they needed precision.
The recruiting pattern observation. Generic: “the market for senior Go engineers is tight.” Specific: “teams in the 200–500 engineer range running Kubernetes-native Go services are losing 2–3 candidates per role to [competing company category].” The specific version signals market intelligence the hiring manager doesn’t get from other recruiters.
The peer reference. Naming a comparable company’s solution without revealing details. “Let me tell you what [competitor] did” is both useful (competitors are the most relevant benchmark) and a credibility signal (we work with the right kind of companies).

Sender identity

Each of the 9 recruiters had their own sender domain (lookalike of the agency primary). The emails came from the individual recruiter who would actually handle the search if the engagement happened. No agency-generic sender.

This matched how hiring managers actually buy recruiting services — they hire individual recruiters, not agencies. Aligning outbound sender identity with buying identity was important.

Volume

Each recruiter handled ~50 emails/week from the pipeline. Total: ~450 emails/week across the agency. Small relative to most outbound programs. Large relative to what this agency had been doing with InMail.

Month 3 — Early results

Positive reply rate: 5.4%
Meetings booked: 22 (for the month)
Meetings converted to active searches: 7
Searches that led to placements: still in flight (searches run 6–10 weeks on average)

By end of month 3, the outbound motion had matched the agency’s prior InMail baseline (20 meetings/month) while freeing ~25 hours/week of senior recruiter time that had previously gone to prospecting.

Months 4–5 — Scale

Months 4 and 5 normalized at:

~450 emails/week
5.8% positive reply rate
~26 meetings/month (up from 20 pre-engagement, and the InMail motion continued in parallel at lower intensity, so total meetings rose to ~38/month)
12 new searches initiated per month (up from 6)

Placements attributable to outbound-sourced searches: 14 over the 5-month engagement (fees ranging from $45k to $65k each).

Cost-per-placement from outbound (engagement fees + infrastructure amortized + research analyst time) worked out to about 40% of the equivalent cost-per-placement from the InMail motion (primarily because the senior recruiter time saved was the dominant cost line).

What was specific to this engagement

The agency had a genuinely specialized vertical. Go/Rust backend is a real specialization with observable public signals. A generalist “senior engineering roles” recruiting agency wouldn’t have this signal surface.
The founder supported the shift in how recruiter time was allocated. Moving senior recruiters off prospecting and onto placements required internal restructuring. Some agencies can’t make this shift.
The signal pipeline was non-trivial to build. It’s one of the more technically involved list-building projects I’ve run. Agencies without technical capacity (or without me to do it for them) would struggle to replicate this cleanly.

What was generalizable

For any recruiting agency — or any vertical agency — considering outbound:

Vertical specialization is a signal advantage. The narrower your specialization, the more specific your list can be, and the less you compete against generic outreach.
Public stack signals are extractable at scale. GitHub, job postings, engineering blogs — none require expensive data licenses. The work is in the extraction, not the acquisition.
Individual-recruiter sender identity beats agency-generic. Hiring managers buy from recruiters, not agencies. Match the sender identity to the buying identity.
Opener specificity is the whole game. Generic recruiting outreach is filtered. Specific references to engineering artifacts lands.
Free up senior recruiter time. If outbound just adds to their load, it doesn’t produce the economic lift. The motion works because it moves prospecting off the senior recruiters.

The honest caveats

The stack signals aren’t perfect. About 12% of the list in month 3 turned out to have meaningfully misclassified stacks (e.g. company with one public Go repo but primarily running on a different language internally). Recruiters had to cross-check before sending.
The research analyst time (5–8 minutes per lead to verify the artifact reference) is a real cost. At higher volumes the agency would need multiple research analysts.
The 14 placements in 5 months is the attribution I’m most confident in. There were 3 additional placements where the hiring manager had multiple touch points (InMail + email + referral) and attribution was ambiguous. Could be higher; could be lower.
One recruiter on the team didn’t take to the new motion. He preferred manual LinkedIn prospecting and performed below the agency average on the outbound. Not every team member will adapt.

Why this mattered beyond placements

The unexpected benefit was that the signal pipeline became the agency’s sales intelligence tool. Recruiters started consulting the signal data before outbound discussions with hiring managers, which changed the tone of those calls.

Instead of “tell me about your hiring needs,” recruiters could open calls with “I saw you shipped [thing] last quarter, and I suspect you’re about to hit [specific hiring pain].” Hiring managers felt seen. Conversations started on more equal footing.

That shift isn’t captured in placement numbers. It showed up in close rates on in-progress searches — the agency’s search-to-placement ratio improved by ~8 percentage points during the engagement, which the founder attributed partly to the informed opening of every hiring manager call.

When this approach fits

Right fit:

Specialized recruiting (or specialized professional services) with a narrow, verifiable vertical
Buyer persona that responds to depth and specificity rather than convenience
Team that can shift senior-practitioner time off prospecting and onto delivery
Willingness to invest in signal pipeline infrastructure

Wrong fit:

Generalist recruiting (the signal pipeline has nothing specific to target)
High-volume, low-fee placements where cost-per-lead has to be pennies
Teams that can’t or won’t reorganize who does prospecting

For the right kind of specialist recruiting agency, this motion is a 2–3x lift on meeting volume with roughly half the senior-recruiter time invested. For the wrong kind, it’s an expensive project that doesn’t pay back.

The test question: “would our hiring managers recognize a signal-specific cold email as different from every other recruiter’s cold email?” If yes, the motion will work. If no, the agency isn’t specialized enough for it to matter.