Recruiting analytics has a volume problem. Every applicant tracking system can produce forty charts, and most of them get looked at once. A metric is only worth tracking if someone will change what they do because of it. By that standard, most teams need about five numbers, reviewed regularly, with a clear owner for each. This post lays out the five that earn their place, the ten that usually do not, and how to keep the first group honest.
The five that change decisions
1. Time-to-fill, by role
Days from the job being opened to an accepted offer. Track it per role or role family, not as a company average, because a 12-day average that blends 6-day warehouse hires with 45-day engineering hires tells you nothing. Use it to set expectations with managers and to spot roles that are drifting. If a role that usually fills in 20 days is at 35, something changed: the market, the posting, the interview loop, or the offer.
2. Stage drop-off
Of the candidates who enter each stage, what share move forward, and how long do they wait? This is the diagnostic metric. It points at the exact place the process leaks. Illustratively, if 70 percent of screened candidates never reach interview, you have a scheduling or engagement problem. If offer acceptance is at 55 percent, the problem is upstream: pay, timing, or expectations set in the first conversation. Fix the leaking stage and everything downstream improves.
3. Source effectiveness
For each source, how many applicants, how many reached interview, how many were hired, and what it cost. The point is to spend money and attention where hires come from, not where applications come from. The two are often different boards. Review it quarterly and shift budget accordingly.
4. First-response time
Median time from an application or inbound candidate message to the first human reply. This metric predicts ghosting better than anything else you can measure. Get it under an hour and most engagement problems shrink. It is also one of the few metrics a recruiter can move this week, on their own, without a budget.
5. Hiring volume against plan
Hires made versus hires needed, by month and by team. This is the number leadership actually cares about, and it frames the other four. If you are behind plan, time-to-fill and drop-off tell you why. If you are ahead, source effectiveness tells you where to ease off spend.
The ten you can mostly ignore
These are not useless in every context, but for most teams they consume attention without changing behaviour.
- 01Total applications. Volume is easy to buy and says nothing about quality.
- 02Applications per posting. Same problem, divided by a number.
- 03Job views and click-through rate. Useful to a marketer optimising a posting, rarely to a recruiter deciding what to do next.
- 04Resumes screened per recruiter. Encourages fast screening, not good screening.
- 05Interviews per hire as a target. Fewer is not better if it means rushing, and more is not thoroughness if it means indecision.
- 06Company-wide average time-to-hire. Averages across different roles hide everything interesting.
- 07Cost per application. Cost per hire, by source, is the version that matters.
- 08Offer count. Offers made without acceptance rate is half a number.
- 09Candidate satisfaction scores collected only from hires. The people with a complaint are the ones you rejected, and you did not ask them.
- 10Time spent in the ATS. Activity is not output.
Keeping the five honest
Metrics rot when definitions drift or data quality slips. A few habits prevent it.
- Define the clock. Time-to-fill starts when the role is approved and opened in the system, not when someone first mentioned it in a meeting. Write the definition down and do not change it mid-year.
- Make stage changes mandatory and easy. Drop-off data is only as good as the discipline of moving candidates. If managers update stages a week late, the numbers describe the update habit, not the process.
- Tag every source at intake. Multi-posting with tracking links does this automatically. Manual referrals and agency submissions need a rule for who tags them.
- Segment before you compare. Role family, location, and level each change what normal looks like. Compare like with like.
- Look at the distribution, not just the median. A median time-to-fill of 22 days can hide four roles that have been open for 90.
What to do with the numbers
The purpose of all this is a short list of actions. Time-to-fill drifting up on one role: check the posting and the interview loop for that role. Drop-off spiking at one stage: find the owner of that stage and ask what changed. A source producing applicants but no interviews: stop paying for it. First-response time over a day: move messaging into the candidate record and set a template for the first touch. Hires behind plan: decide whether the constraint is sourcing, process, or offers, and the other metrics will tell you which.
Five metrics, reviewed weekly, with an owner and an action for each, will do more for a hiring team than forty charts nobody opens. Start there. Add a sixth only when you can name the decision it will inform.