Sourcing that runs overnight and reports in the morning.

Your database is the cheapest pipeline you own.

Replies do not sit in an inbox. They land in your ATS pipeline already ranked against the same scorecard as every inbound applicant, so a sourced candidate and an applied candidate are compared on identical evidence.

How does candidate sourcing software actually work?

Underneath the interface, sourcing software is three things joined together: a search layer, a ranking layer and a message queue. The search layer runs structured queries against candidate records — your own database first, then whatever external profiles the tool can legitimately reach. Boolean search is the classic form of that query, combining operators like AND, OR and NOT with job titles, skills and location modifiers to narrow a very large index to a relevant set. X-ray search does the same job through a general search engine when a site has no usable search of its own. Semantic matching sits alongside both, translating a job description into an approximation of the same query so nobody has to write the syntax by hand.

The ranking layer decides reading order rather than outcome. It scores the matches against the role, and the useful version of that is explainable and overridable, because a sourcing query is a hypothesis about who might fit and hypotheses need correcting. This is also where re-mining happens: the same query run against your existing records surfaces silver-medalist candidates and past applicants who were already assessed once, which is why the search starts inside before it goes outside.

The message queue is the part that turns a list into a pipeline. An outreach sequence carries the first message and a small number of follow-ups across channels, stops on a reply, and stops on an opt-out. Then the handoff: replies do not sit in a personal inbox, they land in your ATS pipeline ranked against the same scorecard as every inbound applicant. Search, rank, reach out, hand over — and the last step is the one most teams under-build.

Who needs a sourcing tool, and who genuinely does not

Plenty of teams buy sourcing tooling before they have a sourcing problem. If your job ads fill your roles, one person runs the whole search, and nobody has ever waited on a shortlist, sourcing software adds machinery to something that already works. "Not yet" is a real answer, and it stays true until a role sits open past the point where waiting for applicants stops working.

What tips the decision is rarely volume. It is repetition and memory. A company that hires the same scarce specialist every few quarters is starting the same search from nothing each time, and the people it assessed last round have gone cold in the meantime. Two people sourcing the same role hit the wall sooner than one person sourcing five, because the moment a second recruiter needs to know who has already been contacted, a personal inbox stops being a system.

The clearest signal sits in your own records. If you cannot answer "who did we interview for this last year and where did they end up", you are paying to rediscover people you have already met. If you are still weighing categories, the candidate sourcing software buyer's guide covers what the tools do, and the roundup of AI sourcing tools covers who builds them.

What changes in the first 90 days of outbound sourcing

The first fortnight is spent making your own data usable, and it is duller and more valuable than anyone expects. Records get imported and de-duplicated, roles get described precisely enough to search against, and the first re-mine runs over people you already have. Teams routinely find candidates they assessed and forgot. That is not a disappointment, it is the cheapest result available, and it is worth taking before any external search runs.

Inside the first month the visible change is discipline rather than volume. Outreach moves out of personal inboxes into sequences with a defined cadence, follow-ups stop being something a recruiter remembers on a good week, and everyone can see who has already been contacted about what. The unglamorous benefit is that two recruiters stop writing to the same person in the same fortnight.

Between days 60 and 90 you get evidence. Which channels produce people who reach an interview rather than merely reply, which messages open conversations, and which roles are genuinely scarce as opposed to badly described — that is sourcing yield and source of hire doing their job. One honest caveat: that reporting is only as good as your attribution. A referral that was first surfaced by a search, or a candidate who applied months after being contacted, will be miscounted unless someone decides in advance how to credit them.

Free 1-user plan · No credit card · Talk to a real recruiter

How to evaluate a sourcing tool without a six-month process

Feature grids are how sourcing selection turns into a quarter-long project, because every vendor ticks every box and the grid ends up measuring marketing copy. Run one real role instead — ideally the hard one you have been avoiding. Search for it, build a shortlist, send outreach, follow up, and see what comes back. A tool that is awkward for one scarce role will be unusable across ten.

While you do that, test the three things that break after purchase. Point it at your own messy database first, because re-mining quality is decided by how well it handles your worst records rather than a clean demo tenant. Check what happens when a candidate opts out or a sequence hits quiet hours, since consent controls are what protect your employer brand once volume rises. Then look at the exit before the entrance: what your candidate and outreach history looks like on the way out, in what format, and who owns it.

On price, compare the model rather than the headline. Per seat, per contact reveal, per credit bundle and per sourced hire all cost differently in month eighteen, and a credit-metered tool prices your curiosity. Ours has no contact-data credits and sits in one line item, which is on the pricing page in full. Two weeks of real searching beats sixty days of demos.