Reading a sourcing dashboard: 6 metrics every agency director should track

Once an agency has sourcing automation in place, the question shifts from "how do we run more searches?" to "how do we know if any of this is working?". Most directors we work with don't track sourcing metrics in a structured way — they track placements, billings, and gross profit, and assume sourcing is healthy if those numbers are healthy.

That's a reasonable shortcut when sourcing is a black box. When you can see it, six numbers are worth watching.

How many hours pass between a new role being added and the first scored shortlist of 20+ candidates being ready for the recruiter. This is the metric the automation most directly improves. Pre-automation, it's typically three to five days. Post-automation, two hours is achievable for most desks.

Watch this number per source. If LinkedIn's contribution is healthy but Reed has fallen off, you've got a connector problem to fix, not a sourcing problem.

Total candidates surfaced and scored per role. Pre-automation, a recruiter manually reviews 30–50 candidates per role. Post-automation, the system scores 150–300, and the recruiter reviews the top 25.

A high candidates-per-role number is a leading indicator of a healthy pool. A falling number across roles in the same vertical suggests the market is tightening — useful intelligence to feed back to clients.

Of the candidates the AI flags as 80+, how many does the recruiter actually shortlist? This measures how well the scoring model is calibrated to the recruiter's taste. A conversion rate above 60% means the scoring is roughly trustworthy. Below 40% means either the model needs retuning, or the recruiter is screening for something the model doesn't see.

This is the metric that tells you whether your automation is agreeing with your recruiters or fighting them.

Of the candidates the recruiter sends a personalised email to, how many reply within seven days? This isn't just a sourcing metric — it's a feedback loop on the quality of the candidates the AI is surfacing. Good candidates respond at 15–25%. Below 10% and the system is either reaching out to badly-fit people or the personalisation isn't landing.

Once a candidate responds, how long until your recruiter has them on a discovery call? This isn't an automation metric, but it's the metric the automation makes visible for the first time. Most agencies discover, post-launch, that they're losing 30% of their interested candidates to slow follow-up — not slow sourcing.

Add up everything you spend on a source per month — LinkedIn Recruiter seats, Reed subscriptions, CV-Library credits, scraping proxies — and divide it by the number of candidates from that source who made it onto a final shortlist that month. The number that comes out is the true cost of a qualified lead from each source.

Most agencies are surprised by how this breaks down. LinkedIn Sales Navigator typically delivers the best cost-per-lead despite being the most expensive seat. Job boards vary wildly. Internal CV pools are by far the cheapest source, when they're used — and most agencies don't use them enough.

These six numbers are worth pulling into a weekly view that the director and the senior recruiters look at together. The point isn't to micromanage individual recruiters — it's to spot which roles, which sources, and which stages of the pipeline are leaking time and money, before the placement numbers start dropping a quarter later.

A sourcing dashboard isn't surveillance. It's the first time most agencies can see what's happening between "new role in" and "candidate offered".