The Warm-Intro Path to a Passive Candidate: Map, Rank, Ask

You have found the exact person you want to hire, their profile is a clean match, and they ignore cold messages the way everyone with options ignores cold messages. This guide is for in-house recruiters, sourcers, talent leaders, and founders doing their own hiring, and it walks the full distance from one named passive candidate to a mapped, ranked set of warm paths and a scripted, permission-based ask. Published advice on this stops at "leverage mutual connections." This is the procedure that turns that phrase into an introduction.

The whole method rests on one asymmetry. A cold message to a passive candidate lands somewhere around 1-5% response. A warm introduction runs 21-34% open or response, and by some measures far higher. Everything below is about earning that difference without burning the person whose credibility makes it possible.

Why a warm path beats a cold message here

A warm introduction to a passive candidate converts far above cold outreach because the introducer is staking their own reputation, and the candidate reads that stake before they read your pitch. The numbers vary by context but the direction is consistent across every source.

Read the figures below as directional. They come from different worlds - sales, fundraising, hiring - and no one has published a single clean "warm intro to a named passive candidate" recruiting benchmark. What holds across all of them is the gap.

The hiring-specific proxy worth internalising is the referral funnel. Of every 10 candidates who receive a referral notice, 8 respond, 6 apply, 4 are interviewed, and 1 is hired, drawn from 1.1 million referrals across 744,580 users. A referral is a warm intro's blunter cousin, and even that blunt version responds at 80%. A hand-built, permission-based intro to one named person you actually want is the sharpened version of the same mechanic.

What counts as a real connection

A real connection is a verifiable overlap between two people, not a shared logo on two profiles. The seminal research operationalised tie strength as contact frequency, and LinkedIn's own experiment measured the strength of network as the intensity and frequency of contact. You cannot see frequency from the outside, so you infer it from public, checkable overlaps and then confirm the human relationship with the connector directly.

Three overlap types are checkable from public data. Each proves something specific, and each has a way it lies.

• Shared employer with overlapping dates. Proves the two people were in the same building at the same time. It lies when the employer is large and the dates overlap but the teams never met: two people at a 100,000-person company are not a tie. Confirm a shared team, project, or manager.
• Shared school cohort. Proves a plausible social overlap when the years line up. It lies when the school is huge and the years match by coincidence. Confirm the same programme or a small enough cohort to matter.
• Co-contribution on code. Proves collaboration when it is real, because GitHub collaboration is explicit and logged, either through manually invited collaborators or through fork-and-pull where changes are reviewed and merged. It lies when it is incidental: two people starring or forking the same repo have never worked together. Confirm merged pull requests or shared commit history.

The connector who converts

The connector who converts is usually a moderately weak tie - a former colleague, not a best friend - in a digital industry, who carries standing with the candidate. This is the least intuitive part of the method and the part the research is clearest on.

The load-bearing finding comes from an experiment run across LinkedIn with roughly 20 million people over five years, generating about 2 billion new ties and 600,000 job changes. The result is an inverted-U: the greatest job mobility comes from moderately weak ties, sitting between the very weakest ties and ties of average strength. Weak ties increase job transmission, but only to a point, after which returns to tie weakness diminish. The strongest ties help least.

The mechanism is information environment. Your closest contacts know what you know and reach who you reach, so a best-friend connector adds no new access into the candidate's world. A moderately weak tie bridges into a network you cannot otherwise touch.

The closest friend feels safe and adds nothing; the former colleague you both half-remember is the bridge.

Two refinements matter for a technical search. First, the weak-tie advantage is strongest in digital, high-tech industries, where networks are more diffuse - which is exactly where most sourcing for senior and staff engineers happens. Second, tie strength is not the only variable. Reid Hoffman's point is that an introduction is a credit instrument, and the introducer's past delivered value is what converts. The canonical Nvidia introduction worked because the introducer had already made the investor money. So rank partly on the connector's standing with the candidate, not just on whether a tie exists.

This is where the sources genuinely disagree, and you should hold both. Granovetter and MIT favour the moderately weak tie. Hoffman stresses the introducer's credibility. In practice: a strong tie who owes you and carries weight with the candidate can beat a weak tie who barely remembers them. Use tie strength as your default sort, then override for credibility when you have a clear read on it.

The double opt-in request

A double opt-in introduction is one where the connector confirms with both sides before making the intro, and it is the only structure you should use. Popularised by Fred Wilson's 2009 essay and championed by Reid Hoffman, the rule is simple: always require double opt-in.

Its power is confirmatory, not merely courteous. As Hoffman puts it, it is not that you are getting permission from your connections, though that is the right thing to do; it is that you are confirming in advance that the intro will very likely be value-creating for both parties. The candidate has already said yes before they meet you. That is why the response rate climbs.

The flow is fixed. The requester notices a mutual connection and contacts the connector. The connector says they must check with the prospect first and contacts the candidate separately. If the candidate is open, the connector introduces both parties. Nobody is surprised, and the connector's credibility is protected because they never forwarded anyone into an ambush.

The single point where this collapses is ask size. If the opt-in ping does not make the ask crystal clear in two sentences, the candidate says no by default. Double opt-in works for a 15-minute intro chat, not a 90-minute request. Big asks belong in the second email, after the intro is made. Keep the first ask small enough to earn a fast yes.

Give the connector a ready-to-forward blurb so they write nothing. The less friction on the person doing you the favour, the faster and cleaner the yes.

Hey [Connector], quick favour - I'm hiring for a [role] on my team and [Candidate] is exactly the person I'd want. I saw you two overlapped at [shared employer/repo/school]; would you be up for checking if they'd take a 15-minute intro call with me? Totally fine if it's a no. Here's a blurb you can forward as-is if it's easier:

---
[Candidate] - a recruiter I trust, [Your name] at [Company], is building out [team] and thinks you'd be a strong fit for a [role] they've got open. They're not asking for anything but 15 minutes to say hello. Want me to connect you two? No pressure either way.
---

This is the moment the method earns its keep, and it is the moment the tooling matters most. Mapping every former colleague, school cohort, and co-contributor of one candidate against your own team's networks is slow by hand. Refolk lets you ask for those paths in plain English and get them back mapped.

Run the play

This is the end-to-end procedure for one named candidate. It runs from lock to introduction in a few working sessions plus the connector's own turnaround time. Each step names who does it and what done looks like.

Budget roughly 15 minutes to lock the target, 30 to 45 to enumerate the footprint, 30 to build the longlist, and 20 to score it. The connector's check with the candidate typically takes 1 to 3 days; the introduction itself takes five minutes. The compliance step has a hard 30-day clock attached to it.

How this goes wrong

Most failed warm intros fail in predictable, nameable ways, and every one has a check that catches it before you burn a connector or a candidate. Treat this section as the pre-flight, not the appendix.

Two of these deserve extra weight because they carry cost beyond a missed hire.

Skipping the candidate opt-in feels faster - the connector just forwards you straight to the candidate - but it burns the connector's credibility if the candidate resents the ambush. The whole point of the structure is that the connector never spends their reputation without a yes in hand. Confirm the connector asked the candidate separately before the introduction exists.

Data stockpiling is the quiet one. Simply building your talent database by adding candidate data in case you need it later is not lawful under GDPR. Your basis is legitimate interest tied to a specific open role, and that basis carries a duty: inform sourced candidates that you are processing their data within one month, and it cannot override the candidate's own rights and freedoms. If there is no live role, there is no lawful basis. This is why step one locks the role before you touch the footprint.

Keep the play current

A warm-intro program stays healthy when you track which connectors and overlap types actually convert, and when you account for how thin your warm paths are in the market you are hiring in. Both change over time, so re-check them rather than assuming.

Connector supply is not evenly distributed. In Refolk's index the internal warm-path pools shrink sharply outside the US, which means in smaller markets your own team's network will not carry the search and extended-network connectors matter far more.

The candidate side is thin outside the US too, and its shape differs by market.

The US senior and staff engineer pool is about 13 times Germany's, and the title mix leans differently - staff-leaning in the US sample, senior-leaning in the German one, though that mix comes from a 25-record sample and is directional, not a census. The practical read: in a market with roughly a tenth of the people, you will exhaust internal paths fast, so build your extended-network connector list early and lean on it.

Two habits keep the method honest over time. Log which overlap types produced the introductions that led to hires; if co-contribution ties convert and school ties do not for your searches, weight your longlisting accordingly. And re-run your compliance read whenever platform terms or your legal basis might have shifted, because the CFAA, contract, and GDPR tracks each move independently and a favourable ruling on one does not protect you on the other two.