71% of Placements Sit in Your CRM. Recruiters Re-Source Anyway.

Recruiterflow's benchmark of 2,100+ search firms landed a number that should have ended the sourcing debate: 71% of placements come from candidates already sitting in the firm's own database before the role opened. And yet the same recruiters who logged those candidates last year will open LinkedIn Recruiter tomorrow and start over. This is not a laziness problem. It is a trust problem, a freshness problem, and a workflow problem, in that order.

The 71% number is the current baseline, not a stretch goal

Seventy-one percent of placements at the average agency come from candidates already in the CRM before the requisition opens, per Recruiterflow's Economics of Recruiting benchmark across 2,100+ firms. The top quartile pulls 54.4% of placements from existing records; the rest pull 48.6%. The gap looks small until you notice the top quartile also averages 5.21 placements per recruiter per year against 1.38 for everyone else, roughly 4x the output.

Read that carefully. Top firms are not out-sourcing the field. They are out-converting it from a shared pool of records that already exists inside every mature ATS. The scaling axis is not intake. It is rediscovery.

The counterintuitive part: top-quartile recruiters add fewer candidates per head, not more. About 800 per recruiter at top firms versus 930 at the rest. They add less, place more, and the delta is the CRM they already paid to build.

Your CRM is the market, especially outside the US

For most common titles, one mature agency database already holds a meaningful slice of the total national supply. That is a market structure claim, not a motivational poster.

Look at the raw title counts. In Refolk's index of professional profiles, 349,785 people in the US currently hold the title "Software Engineer." In the UK, the same title returns 43,523. That is an 8.04x spread for the same job.

A UK agency that has been running technical searches for five years plausibly holds a low-to-mid single-digit percentage of the entire national supply of software engineers already, tagged, notes attached, phone numbers verified once upon a time. Going to LinkedIn first, in that market, is renting access to a pool you have already partially bought.

8.04x

The re-sourcing tax is quantifiable, and it is large

LinkedIn requires 14.15x more candidates per hire than referrals do, and rediscovery from your own ATS behaves closer to the referral end than the cold-source end. That multiplier is where the money is going.

Recruiterflow's channel math: LinkedIn burns 283 candidates per hire. Referrals burn 20. Website applicants burn 33. Layer iCIMS's ~$3,000-per-hire savings estimate for rediscovered versus externally sourced candidates, and a mid-size agency doing 200 placements a year is leaving roughly $600K on the table before you count seat licenses.

Entelo's data says the same thing from the throughput angle: rediscovered candidates move through hiring about 4x faster than newly sourced ones. Bullhorn's number is the one that stings most: up to 70% of candidates in a typical ATS have never been contacted a second time. You paid to acquire them, tagged them, wrote notes, and then never opened the record again.

The freshness problem is why nobody trusts their CRM

Recruiters do not skip the CRM because they are lazy. They skip it because the data feels stale, and LinkedIn feels alive. Fix the freshness, and behavior flips without a training program.

Recruiterflow's own commentary calls out the mechanism: "the CRM is outdated, so recruiters go where the information is fresh." The moment a five-year-old record is silently re-enriched with a new title, a new company, and a working email, it stops being a filing cabinet entry and starts being a lead. This is the entire product category behind job change alerts.

The BLS backs the math. US median employee tenure fell to 3.9 years in January 2024, the lowest reading since 2002. Every ATS record silently ages into relevance about every four years. If you are not listening for job changes, your CRM refreshes itself for free and you throw the refresh away.

This is the exact gap Refolk closes for teams that have given up on their own database. You describe the person in plain English, and Refolk resolves that across GitHub, LinkedIn, and the open web against both your own records and the broader index, so the job-change signal shows up on candidates you already know rather than as a fresh cold list.

The biggest leak is post-source, not pre-source

Only 11.3% of screened candidates ever reach client submission, per Recruiterflow. That is where hires die, and no amount of extra LinkedIn seats fixes it.

Think about what that means for a database-first workflow. If your bottleneck was raw supply, adding a fourth LinkedIn Recruiter seat would help. It is not. Your bottleneck is that the majority of people you have already screened and liked never make it to a client. Every one of them is a warm record, sitting in your ATS, waiting for a role that fits.

The database-first firms operate on that observation directly. Penbrothers, cited in Recruiterflow's case work, requires every new requisition to be matched against its 200,000+ candidate database before any external sourcing is allowed. The result they report: a 66% reduction in time-to-hire and a 92% client submission rate. Not 11.3%. Ninety-two.

The lesson is not "copy Penbrothers' number." It is that a hard workflow rule, match internally before sourcing externally, is worth more than any tool.

The scaling axis is not intake. It is whether the record you already own gets opened before the LinkedIn tab does.

AI matching removed the last honest excuse

Semantic matching over resume text and notes has quietly killed the "our tags are inconsistent, keyword search misses people" defense that recruiters used for a decade. The workflow excuse is now the only one left.

Historically, sourcing from an ATS database failed at the query layer. You wrote a Boolean, it missed anyone whose resume said "Golang" instead of "Go," and you gave up. AI candidate matching, done properly, indexes the free text of a resume and the free text of a role description, then scores fit on meaning rather than exact strings. The tag hygiene problem stops being a blocker.

That is where a plain-English layer helps. Instead of maintaining a taxonomy, you describe the person the hiring manager wants and let the matcher search the notes, the resume body, the LinkedIn headline, and the GitHub bio at once. Refolk does exactly this across your records and the open web in a single query, which is what makes recruiting CRM search finally competitive with a fresh LinkedIn pull on the metric that matters: time to a submittable shortlist.

Firms that layered AI-driven job change alerts on top of this workflow report a 34% shorter time-to-first-submittal and 12% higher placement rates, per Recruiterflow's trend report. Those are not incremental numbers. Those are the numbers that separate the 5.21-placements-per-recruiter quartile from the 1.38-placements-per-recruiter quartile.

70%

A workflow rule that beats any tool

The single change worth making is a hard rule: no external sourcing on a new requisition until the internal database has been queried and the top matches have been contacted. Everything else follows from that.

Here is the workflow the top-quartile numbers are consistent with, distilled into steps:

1. Every new req triggers an internal match first, against the ATS and any adjacent lists (BD, silver medalist pool, old submissions).
2. The match runs semantically, not on tags, so free-text notes count.
3. Every returned record is re-enriched for current title, current employer, and a verified contact before a recruiter opens it.
4. Rediscovery reachouts go out before the first LinkedIn search of the day.
5. LinkedIn is used only for the residual: real gaps the CRM does not cover.
6. Every rejected-but-liked candidate from step 5 gets tagged and pushed back into the CRM with a note, so next quarter's search benefits.

Ashby's 2025 Talent Trends Report is worth reading alongside this. Hires from external job boards in 2024 were half of 2021 levels, even as applications per hire tripled. The external funnel is getting noisier and less productive at the same time. The internal funnel, if you actually work it, is going the other direction.

What "already in the CRM" really means for common titles

For any commonly held title in your primary market, "already in the CRM" is not a hopeful phrase. It is a probability, and it is higher than most recruiters emotionally accept.

Refolk's index shows 94,151 people in the US currently hold recruiting titles (Recruiter, Talent Acquisition, Sourcer). That is the readership of this article. It is also the exact shape of the problem for BD lists: if you have been in the game five years, thousands of those 94,151 are already in your outreach history somewhere. Same math, different table.

The 349,785 US software engineers number lands the same way. An agency that placed 40 engineers a year for five years is not looking at 200 relevant records. It is looking at every rejected candidate, every silver medalist, every "not now, ping me in six months," every referral that did not convert, every intake note. It is thousands of profiles per active desk.

The database is the market. The only question is whether you can actually see it.

FAQ

What is candidate rediscovery, exactly?

Candidate rediscovery is the practice of re-matching existing ATS or CRM records against a new requisition before doing any external sourcing, typically with semantic search over resume and note text rather than tag-based Boolean. The mechanism it exploits is simple: US median job tenure is 3.9 years, so any candidate record older than about three years has a strong chance of having changed employer, title, or seniority, which changes their fit. Rediscovered candidates move through hiring roughly 4x faster than freshly sourced ones (Entelo) and save around $3,000 per hire versus external sourcing (iCIMS).

Why do recruiters keep going to LinkedIn if their CRM has 71% of the answer?

Because fresh data feels safer than stale data, and until recently the CRM genuinely was stale. Recruiterflow's own commentary flags this: recruiters go to LinkedIn because the CRM is not enriched, tags are inconsistent, and keyword search misses people. Job change alerts and semantic AI candidate matching remove both defenses, but the behavior only flips when internal match becomes a workflow rule at intake, not an optional step.

How do I know if my CRM is worth searching first?

If your firm has been active for more than three years in a given specialty, the answer is almost certainly yes. The math from Refolk's index gives a floor: even in a smaller market like the UK with 43,523 "Software Engineer" title-holders, a moderately mature agency database can hold a meaningful share of the national supply of that title. If you cannot easily query your CRM in plain English, the practical fix is a layer that sits over your records and the open web at once, so recruiting CRM search stops being a Boolean exercise.

What is the fastest first move if I want to fix this next week?

Add one rule: no external sourcing on any new req until the requisition has been matched against the internal database and the top 20 records have been enriched with current title and current employer. That single change is what separates the Penbrothers-style 92% submission rate operations from the 11.3% industry average, and it costs nothing but discipline. Add job-change alerts on top and you should see the 34% shorter time-to-first-submittal that Recruiterflow's trend report attributes to the AI-alert cohort.