The Poaching-Window Score: Rating a Company's Talent Exposure

You have a list of companies you would happily poach from and no way to say which one to work first, or which team inside it is already halfway out the door. This guide is for in-house recruiters, sourcers, and talent leaders who want a single weighted score that turns scattered disruption signals into a ranked source-company queue. It gives you the dimensions that matter, how to weight each one, the segment that exits first, and the window before the opening closes.

Existing playbooks cover layoffs alone, or list signals flat with no priority, or rank rivals for research. None of them hands you one comparable number per company with the exposed segment and a source-before date attached. That is the gap this closes.

What the Poaching-Window Score measures

The Poaching-Window Score rates how ready a company's best people are to leave, right now, from signals you can see without an insider. It is not a measure of whether a company is bad to work for. It is a measure of disruption: an event happened, the event has a documented attrition lift, and the clock is running.

Three things travel together in the score, and they must stay attached to each other or the number is useless:

• The score - a weighted composite of dated triggers, higher when the triggers are causally validated.
• The exposed segment - which team or role inside the company leaves first, because you source people, not companies.
• The decay window - a source-before date, because a trigger from eighteen months ago is a company that already re-stabilised.

The core insight that makes this worth doing: across RTO mandates, mergers, and contagion, the people who leave first are the senior, skilled, high-option people. The mechanism is optionality - they can leave - and optionality is the same trait that makes them worth poaching. The window opens on exactly the people you want.

The dimensions and how to weight them

Score five trigger types, and weight them by how well the causal link is proven, not by how dramatic they feel. The dossier's attrition-lift figures set the ranking: RTO and M&A are validated with formal designs and large samples, layoffs and deferred resignation are strong but observational, and down rounds are soft.

Dataset A gives you the raw evidence behind the weights.

Translate that evidence into weights. A defensible starting scheme, which you should recalibrate against your own hit rate:

• RTO mandate: weight 3. Best-validated, pre-announced, and it concentrates on senior and skilled staff. This is your anchor.
• Acquisition or merger: weight 3. One-third of acquired staff leave in year one versus 12% of traditional hires, and the Wharton/ECGI study found merger turnover jumped 45%. Managers go worst.
• Layoff (past year): weight 2. Glassdoor found layoffs imply a roughly 40% increase in voluntary attrition the following year, and post-layoff ratings take more than two years to recover. Strong, but survivors, not the whole company.
• Manager churn: weight 1. The trap dimension. Real but inconsistent. Only counts when it stacks.
• Down round: weight 1, and only as a modifier. Morale damage is asserted everywhere but no public attrition-lift figure exists.

Scarcity multiplies the score

A trigger at a company staffed with a scarce skill is worth more than the same trigger at a company staffed with a common one, because the pool it opens barely exists elsewhere. Multiply the weighted score by a scarcity factor tied to how rare your target skill is.

Refolk's index makes the size of that effect concrete.

In Refolk's index of professional profiles, US Software Engineers listing Rust are 22.4x rarer than those listing Kubernetes. When a Rust-heavy employer hits a trigger, the window is on a pool you cannot easily replace, so it deserves urgency a Kubernetes trigger would not. Use a simple multiplier: common skill 1.0x, moderately scarce 1.5x, genuinely rare 2.0x.

Which segment exits first

Name the exposed segment before you build a single search, because the score tells you which company and the segment tells you which people. Every validated trigger has a documented first-out group, and it is almost always the most portable, highest-option talent.

Dataset C maps trigger to segment.

Read it like this. After an RTO mandate, aim at senior and skilled ICs first, and note that female turnover rose nearly 3x that of men, so senior women in your target function are the sharpest cohort. After an acquisition, aim at managers and staff-level engineers who joined before the deal closed, since acquired firms lose about four of ten managers within 24 months and above half in hostile takeovers. After a single high-signal departure, aim at the departed person's immediate team, because coworkers are 9.1% more likely to quit within 135 days, rising to 14.6% on teams of six to ten.

The window opens on exactly the people you want, because the trait that lets them leave is the trait worth poaching.

The exposed segment is what turns a company score into a search. A composite score of 8 at an acquired fintech means nothing until it reads "engineering managers and staff engineers who joined before the acquisition closed." That phrase is a query.

Attach the decay window

Every trigger has a shelf life, and a score without a source-before date will send you into companies that re-stabilised months ago. Date the trigger, apply the documented lag, and put a hard expiry on the row.

The clearest timing evidence is for M&A, and it has a wrinkle that trips people up: there are two windows, not one. Turnover peaks first in the weeks after close as integration begins, then again some months later as the merged reality lands and the wait-and-see period ends. The second cohort is harder for the company to re-recruit internally, which makes it riper for you. Set two reminders per deal.

For contagion, the window is roughly 135 days from the teammate's exit. For layoffs, the elevated voluntary attrition runs across the following year, so the window is wide but shallow. For RTO, no single public lag figure is established, so watch enforcement instead of the memo date: badge data and promotion ties reveal whether the mandate is real, and the exits follow enforcement, not the press release.

This is the point where the manual work gets heavy: keeping dated triggers, exposed segments, and scarcity multipliers current across 30 or 50 companies is a spreadsheet that rots weekly. Turning the exposed segment into an actual candidate list is where a plain-English search removes the friction, because the segment description is already the query.

Score a company step by step

Run the same procedure on every company so the scores stay comparable. The steps below match the ranked procedure at the top of this guide; the whole loop takes a sourcer and a talent lead a few days for a fresh universe, then a weekly re-scan.

Two notes on sources for step two. WARN databases aggregate 83,000+ filings covering 8.9M+ workers and are cited by the Federal Reserve, so they are your strongest layoff signal for large cuts. For sentiment, negative-outlook workers are 2.1x more likely to job-search on Glassdoor, and ratings dip before layoffs, so a company-relative confidence drop is a cheap pre-trigger to watch before a formal event even fires.

How this goes wrong

The failure modes below are where the score lies to you, and the section is worth more than the scoring scheme itself, because a confident wrong queue wastes weeks. Each one has a tell and a check.

The contagion false positive

A top performer leaving for school, early retirement, or a move to another city does not signal a problem with the company, so it will not spread to their team. If you score that team as exposed you will mine a stable group. Check: before scoring a team as contagion-exposed, look at where the departing person went. A jump to a competitor for a clear step up counts; a life-event exit does not.

Manager-churn overweighting

Manager departure feels like a strong signal and is not. A 39-study review found the effect is "not consistent, immediate or predictable," and turnover tracks workload and pay as much as leadership. Check: never let a single manager exit carry a row alone. It earns weight only when it stacks on a validated trigger like an RTO mandate or an acquisition.

RTO announcement versus enforcement

A memo is not a badge policy. Firms announce mandates they under-enforce, and the exits follow enforcement. Check: confirm the mandate has teeth (badge tracking, promotion or location ties) before you date the window, or your source-before date will be months too early.

The M&A single-window error

Scoring only the announcement window misses the documented second exodus months later, which is the cohort the company struggles to retain and you can most easily reach. Check: set two reminders per deal, one at 90 days and one at four to twelve months.

Down-round softness

Morale damage after a down round is asserted everywhere but has no public attrition-lift figure. Weight it like RTO and you will over-rank distressed startups that merely froze hiring. Check: treat a down round as a weak modifier, and only when equity is clearly underwater and it stacks with a dated signal.

Glassdoor macro noise

Sentiment moves with the wider economy, not just company stress, so in a downturn every company looks exposed. Check: use company-relative drops against the company's own baseline, never absolute index levels.

WARN false negatives

WARN filings are location- and threshold-bound and only capture large layoffs, so rolling or small cuts never appear. A genuinely bleeding team can be invisible. Check: pair WARN with a company-relative confidence drop to catch quiet reductions the filings miss.

Turn scores into searches

The score is only worth the searches it produces, so end each cycle by converting the top companies into named candidate lists. The exposed segment you attached in step four is already the search; you just phrase it. The example below turns an acquisition row into a query that names the trigger cohort, not just the role.

Row: Acquired fintech, deal closed 4 months ago, score 9, scarce (Rust)
Exposed segment: engineering managers and staff engineers who joined before close
Search: "Engineering managers and staff engineers at [that company] who joined before the acquisition closed, based in the US."
Second-window reminder: re-run at month 8 for the delayed exodus cohort.

Before you call a scoring cycle done, run the checklist. It catches the failure modes at the point where they cost the most: right before you spend hours mining.

Keeping the queue alive

A poaching queue is a decaying asset, so treat the weekly re-scan as the real product, not the first build. Windows close, companies re-stabilise, and new triggers fire constantly, which means a queue you built once and never refreshed is worse than none: it points you at stable teams and hides the fresh ones.

Each week, do three things. Refresh triggers across the universe and add any new WARN filings, RTO memos, deals, or confidence drops. Advance the clock on existing rows and retire any company past its source-before date. Re-run the top searches so your named lists reflect who has already moved. WARN activity alone shows how much churns: trackers logged 2,250 companies filing in one recent year, so the trigger universe is never static.

The judgement this guide encodes is small and repeatable: given a company and a handful of public signals, decide whether to source from it now, which team, and by when. Score it, attach the segment, date the window, multiply by scarcity, and re-scan weekly. Do that across your universe and the question "which companies to poach talent from" stops being a hunch and becomes a ranked, dated queue you can hand to a sourcer.