Building the Source-Company Universe for a Live Search

At search kickoff you have a brief and nothing else, and the temptation is to jump straight to a Boolean string. Resist it. Before you search for a single person, you need to decide which companies to pull them from, and that decision deserves its own ordered procedure. This guide is for in-house recruiters, sourcers, talent leaders, and founders running their own hiring, and it delivers one thing: a method that converts a role brief into a ranked, pool-sized, clearance-flagged list of source companies that feeds candidate search without rework. It stops exactly where the talent-map and candidate-search playbooks pick up.

What a source-company universe is, and why it is not a longlist

A source-company universe is the finite, scoped set of employers you will pull candidates from, built and sized before any names are gathered. It is a different object from the people longlist, and conflating the two is the first way searches get mis-scoped.

The sizes diverge sharply. Company lists cluster near 40 organizations. People longlists run much larger and scale with seniority. If you build one list in your head and treat it as both, you will either research 300 companies you cannot cover or size the search from 40 names when the role needs 200.

The math that connects them is the step most practitioners skip. A C-suite or VP map of 100 to 300 names can come from only 20 to 50 companies, which means each company must yield roughly 5 to 6 names. That ratio is exactly what the pool-sizing step later in this guide makes visible. Without it, you have a company list and a name target that do not reconcile.

The universe is not a brainstorm and it is not a wish list. A useful target list begins with clear criteria, not company names, and every row ends with a pool estimate and a clearance status. Those two disciplines, criteria-first and sized-before-committed, are what separate a universe from a list of logos.

How large the universe should be

A managed company list sits near 40 organizations, and talent-mapping guidance targets the 20 to 50 companies most likely to employ each high-impact capability today. Those two ranges are the spine of your sizing decision. Everything else adjusts around them.

One practitioner floor and ceiling is worth memorizing: fewer than 10 companies leaves you too exposed if a few stop hiring, while more than 30 makes it hard to research each one deeply. That is a research-capacity ceiling, not a quality ceiling. If you can genuinely cover 50, size up; most solo recruiters cannot.

The companion numbers are for the people side, so you know what the universe must feed.

Read this table as two blocks. The top two rows are what you are building here. The bottom three are what the universe hands off to. When you pick a company count, check it against the longlist the role actually needs, divided by a realistic per-company yield of 5 to 6 names.

The criteria that define scope at kickoff

Scope is set by a fixed set of dimensions, written down before any company name appears. The documented retained-search dimension set is industry including both direct and adjacent, company size matched to the role, growth stage, geography, culture, and reputation. Lock these first.

These dimensions flow directly from the intake conversation, where the recruiter and hiring manager align on the business goal of the hire, the technical skills and domain expertise, leadership style and cultural fit, and career stage. Get the intake wrong and every downstream filter inherits the error.

The single dimension that moves the universe most before any other lever is geography. The same role returns 472 profiles in the US and 159 in Germany in Refolk's index, a 2.97 to 1 ratio. That gap is member density and market size, not candidate quality. A recruiter who skips the geography decision over-weights whichever market they happen to sit in.

The build procedure, in order

Run these eight steps in sequence. Each has an owner, a rough duration, and a done condition. The procedure front-loads criteria and defers sizing and clearance, which is the ordering that keeps you from committing companies you cannot reach.

One honest caveat on ordering. Public sources disagree about where sizing and clearance belong. Retained methodology front-loads client conflict input at intake, while list-build guides treat sizing and vetting as a later qualification pass. I place both after expansion because you cannot size or clear companies you have not yet named, but if your client relationship surfaces conflicts naturally at intake, capture them early and confirm at step seven.

Expanding beyond direct competitors

Direct competitors are the seed, never the finished list, because a list of pure competitors ignores transferable and feeder sources. Four documented methods widen the universe, and a complete expansion uses all of them rather than just one.

• CAVAC hub-and-spoke. Put a seed company at the center and radiate out to its clients, competitors, and vendors. Then take one of those outer companies, drop it into the center of a fresh template, and iterate. This is the most systematic of the four.
• Adjacent-industry layering. List your direct competitors, then add adjacent industries where the needed experience transfers easily. A common talent-mapping ratio is ten direct competitors and five adjacent industries.
• Feeder and destination tracing. Use talent-flow to find where your top performers came from, which are feeder organizations, and where they go when they leave, which are destinations. Both tiers are legitimate sources.
• Alumni-network tracing. Follow the diaspora from a known strong employer to the companies its ex-staff now cluster in.

Why expand at all, beyond covering your bases? Because rare skills force it. Skills-based matching rather than title matching expands the reachable pool roughly 6 times, and about 8.2 times for AI positions. The rarer the specialization, the wider the company universe has to be to contain enough people.

Expansion is the stage where Refolk removes the most friction, because tracing feeders and destinations by hand across public profiles is slow and error-prone. Asking for the role plus the adjacent startups its alumni move to collapses steps three and four of the procedure into one query, and it does so across public LinkedIn and GitHub records rather than a single platform's view.

Sizing each company's reachable pool

Every company row must carry a pool-size estimate before it is committed. A list of 40 companies with no sizing column is a wish list, not a universe. The countable signal is a filtered pool count per employer, run in a talent-analytics tool.

A good pool read records three things per company: pool size, meaning how many members match the criteria; a mobility signal, meaning members who changed companies in the past 12 months; and, where available, a hiring-difficulty score. That difficulty score matters more than raw headcount, because it is derived from InMail volume over a rolling 12-month window, making it a demand-adjusted figure. A company with a large pool but saturated outreach is a worse source than its raw count suggests. Rank by reachable yield, not gross pool.

Two cautions govern how you read every count. First, counts are estimates, not exact figures, so a precise-looking number is false precision and worth cross-checking against a second source. Second, filters collapse counts fast, and the collapse is non-linear.

That table, from Refolk's index, is the Q3 caution quantified. Swapping one common skill for one rare skill drops the US pool from 472 to 1. Adding a seniority filter on top of a title takes it to zero. The mechanism is that rare skills are power-law distributed, which is exactly why skills-based matching expands pools so dramatically and why specialization forces you to widen the company universe rather than tighten the filter.

Geography stacks on top of this. The same role across two countries shows how much the universe moves before any skill or seniority filter is even applied.

Ranking, tiering, and clearance

Rank by reachable yield and criteria fit, then tier, then clear each row with the client before anything is committed. These three passes turn a sized list into a handoff-ready universe.

One documented tiering mix is roughly 70 percent companies where you have a real connection or genuine expertise, and 30 percent wildcards worth monitoring. That ratio keeps the list defensible without making it timid. Tier one is where you start outreach; the wildcards are where you watch for movement.

Clearance is the pass most in-house teams under-weight because it originates in retained search. An off-limits clause is a feature of retained search where a firm under contract agrees not to approach a client's employees for a set period, conventionally at least one year after a C-level assignment, and it is not usually applied to contingent work. In the build, client input is where conflicts surface: the client flags competitors or companies with non-compete sensitivities and vets the list for conflicts of interest.

Off-limits scales with firm size in a way that matters even if you never sign a retained contract. The two largest search firms each carry about 36,000 hands-off restrictions over any given two-year period, which locks them out of exactly the competitor talent pools clients most want. An in-house team building its own universe carries none of that drag, which is a structural advantage worth naming. But the lesson transfers: apply your conflict pass before handoff, not after outreach has begun, or you burn work when competitor talent goes off-limits mid-search.

The handoff schema

The universe is finished when every row carries a fixed set of fields and a reader can run candidate search against it without asking a question. No single public schema exists for this, so the one below is built from what talent-map outputs record plus the build steps above.

Each row carries: company, tier, criteria-fit, pool estimate, mobility signal, and off-limits status. Those six fields are the minimum that lets the next stage proceed without rework.

company | tier (1/2/watch) | criteria-fit (direct/adjacent/feeder) | pool_estimate (number, source, date) | mobility_signal (% changed co. in 12mo) | off_limits (clear/flagged/blocked)
Acme Cloud | 1 | direct | 472, Refolk index, pulled Mar | 14% | clear
Beacon SaaS | 2 | adjacent | 210, Refolk index, pulled Mar | 9% | flagged (client non-compete)
Northwind Infra | watch | feeder | 61, second-source est, pulled Mar | 22% | clear

How this goes wrong

Most broken universes fail in one of seven predictable ways. Each has a tell and a check. Learn to recognize the false positive, because the dangerous failures look like facts about the market rather than mistakes in your filter.

Two of these deserve extra weight. Title-locking is seductive because a zero result reads as hard evidence. In Refolk's index, Platform Engineer plus Director seniority returns zero, not because no such leaders exist but because directors are simply not titled that way. The title-level and seniority filters are mutually exclusive. The fix is to source on skills and scope, not on the exact string in the job title.

Over-filtering produces the same lie at a larger scale. Stacking every nice-to-have shrank one documented pool from 62,921 to 8. If you read that 8 as scarcity, you will widen the company universe to chase people who were always there behind an overbuilt filter. Strip to must-haves, re-count, and only then decide whether the universe is genuinely too thin.

Keeping the universe current

A source-company universe is accurate on the day you build it and decays from there, so treat it as a living document rather than a one-time deliverable. Two things move underneath it: pool counts and clearance status.

Pool counts drift as people change jobs, which is why the mobility signal belongs in every row. A company with 22 percent annual mobility is a different source six months from now than one at 9 percent. Re-pull counts when a search runs long, and always record the pull date so a stale number is obvious on sight. Because counts are estimates anyway, a re-pull is cheap insurance against building outreach on a figure that has moved.

Clearance status changes even faster, because every new client engagement can add a roster to the no-touch list. If you work alongside colleagues running retained searches, check the off-limits roster against your seed before you rely on any shared list. When a company flips from clear to flagged mid-search, move it to tier two and redirect effort to a cleared source rather than pushing on and burning goodwill.

When the universe is sized, ranked, and cleared, you are done here. The next move is the talent-map and candidate-search work, which takes this schema as its input and turns reachable companies into named, qualified people. The discipline that makes the handoff clean is the one this guide has repeated: criteria before names, a number on every row, and clearance before commitment.