Ask a recruiting team which source performs best and the answer is usually the one producing the most applications. That is a volume answer to a quality question, and acting on it is how recruiting budget ends up in the wrong place.
A channel delivering four hundred applications and one hire is worse than a channel delivering twenty applications and two hires, on every dimension that matters — cost, reviewer time, and outcome.
The four numbers per channel
Compute all four. Any one alone is misleading.
1. Conversion to hire
Hires from the channel / applications from the channel. The headline quality signal. Expect an order of magnitude of variation between your best and worst channels — referrals and direct applications typically sit far above aggregator traffic.
2. Conversion at first review
Advanced past initial screen / applications. This isolates the cost side. A channel with poor conversion to hire but decent first-review conversion is producing plausible candidates who lose later — a different problem from a channel producing noise.
3. Cost per hire by channel
(Channel spend + reviewer time on that channel) / hires from it. Reviewer time is the omitted half. A free channel that consumes fifteen hours of screening per hire is not free — see the full cost-per-hire breakdown.
4. Retention at twelve months
Still employed at one year / hires from the channel. Slow, and the only real quality measure on the list. Channels that look identical on conversion often separate sharply here.
The attribution problem
Source data is unreliable in ways that systematically favour the wrong channels.
Last touch wins. A candidate hears about you from an employee, searches your company name, finds the role on an aggregator, and applies. The aggregator gets credit for a referral. This single effect distorts most source reporting.
Self-reported source is guesswork. "How did you hear about us?" produces answers candidates cannot actually remember. Useful as a signal, useless as a measurement.
Direct traffic is a bucket, not a source. It contains brand awareness, referrals, past candidates and repeat applicants, which behave completely differently.
Two practical corrections: keep the technical source separate from the self-reported one and look at both, and treat referrals as their own category even when the candidate arrived through another route — referral hires behave differently enough to justify it.
What the pattern usually looks like
Across most funnels, the shape repeats even though the exact numbers do not:
- Referrals convert best and retain best, and are volume-limited. They cannot be scaled by spending more, only by asking more.
- Direct applications convert well, and are the return on employer brand and careers page work. They lag investment by months.
- Aggregators deliver volume and convert poorly. They are worth keeping for reach and worth watching for reviewer cost.
- Outbound sourcing converts well per contact and costs the most recruiter time per hire.
- Agencies convert well and cost the most in cash, which makes them a speed purchase rather than an efficiency one.
Your own numbers will differ, and that is the point — the exercise is to find where your mix departs from this shape.
Reallocating on the result
Three moves usually follow the first proper source analysis:
- Cut the worst aggregator, not all of them. Kill the channel with the lowest conversion and highest reviewer cost, and watch total qualified applications rather than total applications. If qualified volume holds, the spend was buying noise.
- Fund the referral programme properly. It is almost always the highest-quality channel and almost always under-resourced, because it has no invoice attached to argue for it.
- Treat careers page work as a channel investment. Direct applications are a channel you build rather than buy — and most careers pages leak badly.
Reviewing it on a schedule
Quarterly is enough. Channel performance moves slowly, and monthly reviews at typical mid-market volumes produce noise rather than signal. Segment by role family before comparing — the channel mix for warehouse hiring and senior engineering hiring have nothing in common, and a blended number describes neither.
If producing these figures currently means exporting to a spreadsheet, Hireall reporting computes them per channel directly. For sourcing itself, see how candidates get into the pipeline.