Grading a Company as a Talent Source for a Sourcing List

You have a role to fill and a growing list of companies to source from, and no principled way to decide which ones belong on it or in what order. This guide is for in-house recruiters, sourcers, talent leaders, and founders doing their own hiring. It gives you a per-company scoring rubric across four signals so you can grade any company as Tier-1, Tier-2, or exclude for a specific search, and turn "list ten competitors" into a defensible, tiered target universe.

Most talent-map advice tells you to list 20 to 50 companies and gestures at tiering by "client fit." None of it scores the single company you are about to add. That is the gap here. The four dimensions below - skill adjacency, title density, reachable pool, and timing/poachability - make the judgement call gradeable and repeatable.

What "grading a company as a talent source" means

Grading a company as a talent source means scoring one organization, for one specific search, on how much reachable, right-skilled, poachable talent it actually holds - then deciding include, tier, or exclude. It is a per-company judgement, not a market map. A target company list is a carefully constructed group of organizations where the talent you want is likely to reside, focused on passive candidates: people not actively looking but open to the right move.

The distinction matters because two companies can look identical on a competitor list and grade completely differently. One has deep title density and a fresh acquisition making its recruiters restless; the other has the same headline but a frozen hiring plan and no reachable pool. A blanket "list the top ten competitors" cannot tell them apart. A four-signal score can.

The four dimensions each answer a different question:

• Skill adjacency asks whether the company's people have your must-have skills, whether they are a direct competitor or an adjacent-industry source.
• Title density asks how concentrated the right titles are inside the company.
• Reachable pool asks how many of those people you can realistically get a reply from.
• Timing and poachability asks whether people there are movable right now.

Why score per company instead of mapping the market

Score per company because a market map tells you the shape of the whole segment, but it does not tell you which door to knock on first. The map is the input; the grade is the decision. Talent maps target 20 to 50 companies and aim for 80 to 90 percent coverage of the defined segment, and firms tag tiers A, B, and C by fit. But those tiers are usually assigned by feel. A scored grade replaces feel with named public signals.

The payoff is a ranked queue your reps or your future self can work top-down, and a rationale that survives a review. When a hiring manager asks why a company is Tier-1, "it competes with us and just got acquired, so its recruiters are at 26% attrition and 44.9% of its ML engineers carry a senior title we can reach" is a defensible answer. "It felt right" is not.

Signal one and two: skill adjacency and title density

Skill adjacency scores whether the company's people carry your must-have skills; title density scores how concentrated the right titles are inside it. Together they answer "are the right people here, and how many." Direct competitors usually score high on both. Adjacent industries widen the universe but trade skill-match for headcount, so they need a harder look.

Looking beyond direct competitors is the hallmark of a good search. Identifying adjacent industries is critical for hard-to-fill roles where the ideal hire sits in a sector you had not considered. But adjacency has a failure mode: adding a sector for its headcount when its people lack your core skill. Require an explicit adjacent-skill cluster - a named overlap in tools, domain, or problem - before you let an adjacent company in.

Title density is where Refolk's index sharpens the grade. In the US Machine Learning Engineer pool, 44.9% carry a "Senior" title. If you are hiring senior ICs, that subset is your real universe, and you can treat the non-senior remainder as noise. That halves the effective list at a company without losing target coverage.

Pool counts come from Refolk's index of professional profiles; shares are derived. The lesson generalizes: before scoring title density, isolate the exact title band your search needs, because the raw company headcount overstates the pool that matters.

Signal three: sizing the reachable pool, not the headcount

The reachable pool is the number of people who will actually reply, not the number who exist. This is the single most abused dimension, and getting it right is what separates a defensible grade from a headcount fantasy. A raw pool of 5,464 looks enormous. At a 6.4% InMail reply rate it is roughly 350 responders. Score the 350.

The discount comes from outreach benchmarks. InMail averages a 6.4% reply rate. Recruiting and staffing lead all industries at 18 to 25%, with top performers hitting 30 to 40% through personalized multi-touch sequences. So the reachable pool is not one number - it is a band set by how good your outreach is. Outreach quality is a scoring input, not a footnote.

Pool counts come from Refolk's index; responder columns are derived from published reply-rate benchmarks. Notice the geographic swing survives the discount: the US pool is 10.8x Germany's before and after applying reply rates. A small dense market can out-deliver a large one only if reply rates rise there, which again points back to outreach quality.

Remember who you are reaching. 70% of the global workforce are passive candidates, and for senior or retained mandates that number is closer to 90%. Your reachable-pool math is a passive-candidate math problem, so the responder band should sit at the lower, realistic end unless your sequences are genuinely strong.

Sizing the reachable pool by hand means pulling title counts market by market and multiplying each by a reply band, which is where a plain-English index earns its keep. With Refolk you ask for the exact title, seniority, and geography and get the count back directly, so you can size responders per company in minutes instead of stitching together filtered searches.

Signal four: scoring timing and poachability

Timing and poachability score whether people at the company are movable right now, and it is the dimension that decays fastest. The strongest public trigger is an acquisition, but its effect is function-specific and time-bound. Treat it as a decaying, per-function signal, never a blanket company-wide "hot."

Attrition after an acquisition is uneven by function. Recruiters leave at a 26% rate. Marketing, sales, and admin roles have the highest departure rates. Machine operators, technicians, and medical reps barely move. By seniority the pattern is U-shaped: middle managers are retained, but senior management at the acquired firm has the highest attrition of any group, and around 18% of entry-level staff are gone 18 months out.

Figures come from Revelio Labs, Merger Integration, and an EY 2025 study. Independent numbers converge: MIT Sloan found acquired employees leave at nearly three times the rate of comparable hires, 33% in year one versus 12%. Acquired firms lose about 4 in 10 managers within 24 months, above 50% in hostile takeovers.

The peak-risk window is announcement to the first 90 days after close, when uncertainty peaks and competitors are actively recruiting. That makes poachability a decaying score. Date-stamp it. A merger announced this month is a different grade from one that closed two years ago.

Reading the freeze signals

WARN filings are the most concrete freeze-and-flush signal. WARN notices are typically required when companies with 100 or more employees plan a mass layoff or plant closing affecting at least 50 workers, filed 60 days ahead. State rules differ: California covers employers with 75 or more, and New York's Mini-WARN covers 50 or more with 90 days' notice. Because filings often land weeks before layoffs take effect, they give an early read on who is about to be on the market.

The arbitrage is timing. WARN precedes payroll data by six to ten weeks, so a company scored on WARN is hot before competitors sourcing on news headlines even notice. But WARN has real blind spots. Not all states require full disclosure, and voluntary programs never fire a notice: Microsoft's voluntary retirement released roughly 8,750 US employees with zero WARN filings in any state. So a clean WARN record does not mean a stable company. Cross-reference leadership departures, M&A, and budget or efficiency messaging.

One legal note on mobility: do not exclude a company because you assume non-competes block cross-competitor moves. The FTC's 2024 non-compete ban never took effect; it was struck down and appeals were dropped, so state law still governs. Confirm the relevant state before excluding anyone on mobility grounds.

The procedure: scoring a company end to end

Grade a company by running the seven steps below in order, budgeting about ten minutes per company once the profile and universe are set. The output is a composite score you can map to a tier.

Sources disagree on universe size: some practitioners cite 20 to 40 companies per role, others 20 to 50. Use the range as a guardrail, and let coverage, not the count, be the finish line.

A scoring template you can copy

No public source publishes numeric per-company weights, so the rubric below is a working default, not an industry standard. Score each dimension 0 to 3, weight reachable pool and poachability highest because they are the two that most often mislead, and read the composite against the tier cutoffs.

Company: __________   Search: __________   Date: __________

Skill adjacency        (0-3): ___  x1  = ___
Title density          (0-3): ___  x1  = ___
Reachable pool         (0-3): ___  x2  = ___   (responders, not headcount)
Timing / poachability  (0-3): ___  x2  = ___   (by function, date-stamped)

Composite (max 18): ___
Tier-1: 13-18   Tier-2: 8-12   Exclude: 0-7
Expiry date: ______   Event triggers: comp / layoffs / funding / M&A

The two-variable call that decides a tier is skill-match against poachability. High on both is Tier-1; high skill but frozen is a watchlist, not an exclude; poachable but weak on skill is a false lead.

How this goes wrong: failure modes and false positives

The grade fails in predictable ways, and most of them come from trusting a single signal. Below are the seven that recur, each with the false positive it produces and the local check that catches it.

• Title-count inflation. A raw pool of 5,464 looks huge, but at 6.4% it is ~350 responders. False positive: scoring reachable pool high on headcount. Check: always multiply by a reply band before scoring.
• WARN blind spots. No WARN filing looks like stability, but voluntary programs and sub-threshold states file nothing. Microsoft shed ~8,750 with zero notices. Check: cross-reference sentiment, leadership exits, and efficiency messaging.
• "Poachable because acquired" over-read. Attrition is function-specific; recruiters and sales spike, technicians barely move. False positive: tiering an entire acquired firm hot. Check: score poachability by function.
• Stale map. Beyond 12 to 18 months, tiers reflect a market that has moved. Check: attach an expiry date and event triggers to every list.
• Adjacent-industry mirage. Adjacency widens the pool but can drop skill-match. False positive: adding a sector for headcount that lacks the core skill. Check: require a named adjacent-skill cluster before including.
• Coverage illusion. Hitting 20 to 50 companies is not the same as 80 to 90 percent segment coverage. Check: estimate what fraction of the true population your list captures.
• Non-compete assumption. Treating cross-competitor moves as legally blocked is outdated; the FTC ban never took effect and rules are state-specific. Check: confirm the relevant state before excluding on mobility grounds.

A raw pool of 5,464 is not 5,464 people you can reach. It is about 350.

The through-line: every one of these is a single-signal error. The grade is deliberately built from four signals precisely so that no one number can carry a company into or out of a tier by itself.

Keeping the list current

A target company list is not a project; it is an ongoing process, so attach a refresh cadence the moment you finish grading. Mapped markets stay valuable for 12 to 18 months, which is the outer bound of your list's shelf life. Beyond it, re-grade from scratch rather than trusting the old tiers.

Between full refreshes, event triggers drive re-scoring. Comp shifts, layoffs, funding rounds, and M&A each change a company's poachability and can move it across a tier boundary. WARN is a leading trigger because it precedes payroll data by six to ten weeks. And the list itself teaches you: as candidate feedback arrives, add companies on fresh insight and remove those returning low-quality leads or flat disinterest. The initial list is never final.

Grade, tier, ship the queue, and let candidate feedback and event triggers pull companies in and out. The rubric holds; the market underneath it moves, and the refresh rules are what keep the two aligned.