Recruitment Metrics That Actually Improve Your Hiring

Most talent teams in India already track recruitment metrics. There is a spreadsheet somewhere, or an ATS reporting tab, or a slide in the monthly review with four numbers on it. Almost none of it changes what anyone does on Monday morning. That is the real problem — not missing data, but data that never forces a decision.

This is a practical guide to measuring and fixing hiring performance for Indian SMBs, startups and growing companies. We will cover hiring funnel analytics from first principles: how to define stages so the numbers mean something, formulas for time to hire, cost per hire, offer acceptance rate and quality of hire, the India-specific distortions that break imported playbooks, and how to build a recruitment dashboard people actually read. Every number here is an illustrative example for a hypothetical company. Your benchmarks should come from your own history.

Why Most Recruitment Metrics Dashboards Get Ignored

Count the phones in the room during a recruitment dashboard review. The failure is rarely about the tool.

It reports activity, not outcomes. Applications received, interviews scheduled, jobs posted — these rise when the team is busy, whether or not anyone gets hired. A busy quarter and a productive one look identical.

Nobody agreed on definitions. The recruiter counts a candidate as "screened" the moment they open the CV; the hiring manager counts it when a real call happened. Two people read one number, reach opposite conclusions, and both stop trusting it.

The metrics have no owner and no threshold. A number without a target is trivia; a target without an owner is a wish. And monthly reporting arrives three weeks after the problem started, by which time the two best candidates have joined somewhere else.

Vanity, Diagnostic and Decision Metrics

Sort every metric into one of three buckets, and be ruthless about which gets airtime with leadership.

Vanity metrics rise when you work harder and prove nothing. Applications received is the classic: triple it by loosening the JD, and quality falls.

Diagnostic metrics do not justify a decision on their own but tell you where to look. Time in stage, screening pass rate, hiring manager response time.

Decision metrics trigger an action when they cross a line — a named owner and a dated fix.

A useful discipline: if a metric has sat on your dashboard for two quarters without ever changing a decision, delete it. Dashboards die of overcrowding, not of missing charts.

Define Your Hiring Funnel Before You Measure Anything

Definitional discipline matters more than tooling. A perfectly configured ATS with fuzzy stage definitions produces confident nonsense. A spreadsheet with airtight definitions produces usable hiring funnel analytics.

Use eight stages. Fewer and you cannot localise a leak; more and recruiters stop updating them.

Three rules will save you months of arguing about numbers.

One: a candidate sits in one stage at a time, and stages only move forward. Backwards movement destroys time-in-stage maths. Log a new event instead.

Two: every exit needs a reason code. A candidate who leaves without one is a permanently lost data point.

Three: the timestamp is the event, not the data entry. If Tuesday's interviews are logged on Friday, every stage duration is wrong by three days.

Write these definitions on one page and make hiring managers read it. Most "the data is wrong" complaints are really "we defined it differently" complaints.

The Core Recruitment Metrics, One by One

Here is the reference table, then how to read each one.

Time to Fill vs Time to Hire

Time to fill starts at requisition approval. Time to hire starts when a specific candidate enters the funnel. The gap between them is your organisation's fault, not the market's.

If time to fill is 68 days and time to hire is 24, the candidate journey is fine — you lost 44 days to approvals, JD drafting and the fortnight before anyone sourced seriously. That is a leadership problem. If the two are close and both high, the process itself is slow.

Fixes: pre-approve headcount at annual planning so requisitions only need a trigger, keep a library of role scorecards, cap panels at three interviewers, and make a shortlist of five mean a decision rather than a request for five more. Report medians, not averages.

Time in Stage

The most under-used diagnostic in recruitment metrics, and the fastest to pay off. Calculate median days from entry to exit for every stage, split by role family.

You will usually find candidates waiting days for a hiring manager to open a profile, a gap between interview and feedback, and a week between "we've decided" and the letter going out because comp approval sits with a travelling founder. Each is fixable without hiring another recruiter.

Source Effectiveness

Measure sources by conversion, not volume. For each — referrals, job boards, outbound, careers page, campus, consultants, rehires — track candidates sourced, screening pass rate, offer acceptance, hires, cost and 6-month attrition.

A common pattern in Indian SMBs: referrals give few candidates with high pass rates and high acceptance, while paid boards give volume with poor pass rates. That does not mean kill the board — boards are a volume tool and referrals a conversion tool, budgeted differently. Watch for sources that convert well but show up heavily in six-month attrition; those are expensive twice.

Screening Pass Rate

Shortlisted ÷ screened. A very low rate means the sourcing brief is wrong or the JD is aspirational. A very high rate is not automatically good — it can mean the recruiter is forwarding everything and shifting the filtering burden onto expensive interview time.

The fix is calibration, not software. At the start of each requisition the recruiter shares five profiles — two strong, two borderline, one deliberately weak — and the hiring manager rates them. Twenty minutes there saves three weeks.

Interview-to-Offer Ratio

If you interview twelve people per offer, either screening is failing or the panel cannot agree on what "good" means.

Read it with screening pass rate. High pass rate plus high ratio means weak profiles are reaching interviews. Low pass rate plus high ratio means the role definition is unrealistic for the budget. Break it down by interviewer too — a single panel member who rejects nearly everyone is quietly setting your hiring bar alone.

Offer Acceptance Rate

This is the market's verdict on your compensation, role, brand and process at once. Always read it with decline reasons attached: "compensation", "counter-offer retained" and "process took too long" demand completely different responses.

The most common self-inflicted cause in fast-growing companies is offer lag. Measure decision-to-letter-release time separately — every day in that window is a day for a competing offer to land, and it is usually the cheapest thing to fix.

Offer-to-Join Drop-off

This is where Indian hiring diverges hardest from imported playbooks. The candidate accepts, you close the requisition, and thirty to ninety days later they do not show up. Notice periods leave a window long enough for counter-offers to arrive and other processes to conclude, and some candidates accept two or three offers as insurance.

Track it by joining month, not offer month, and record the reason: counter-offer retained, joined elsewhere, personal or family, relocation, background verification failure, unreachable.

Cost per Hire

Total recruiting cost ÷ hires in the period — see the dedicated section below for inclusions. Read it as a portfolio, never as one number to minimise: segment by role family and level, and compare against your own trailing four quarters.

Quality of Hire

The metric everyone wants and few define. Worth doing properly and worth refusing to fake, because a score that really means "the manager was cheerful in month two" gets used to justify decisions.

90-Day and 6-Month Early Attrition

Track by joining cohort: of the people who joined in April, how many were gone by July, and by October?

Ninety-day attrition is almost always a promise problem — the role sold was not the role that existed, the salary structure was explained loosely, or onboarding signalled disorganisation. Six-month attrition is more often fit or manager quality. Break both by source, recruiter, hiring manager and location. One manager with repeat six-month exits is a management issue wearing a hiring metric's clothes.

Hiring Manager Responsiveness and Interview SLA

Measure median hours from shortlist shared to profiles reviewed, and from interview completed to feedback submitted. Set an SLA — 24 hours for feedback is a reasonable start — and report adherence by hiring manager.

This is politically uncomfortable and enormously effective. Publish recruiter SLAs alongside it so it reads as shared accountability, not a complaint from HR.

Candidate Experience and NPS

Survey everyone who reaches interview stage, including rejected candidates. Three questions is enough: would you recommend applying here, was the process clear, was your time respected.

The free-text comments matter more than the score. They surface the rude interviewer, the four-hour assessment and the six-week silence your funnel data cannot see.

Pipeline Diversity

Measure representation at every stage, not just at hire, because the insight is where the drop happens. If women are 30% of sourced, 28% of screened, 27% of interviewed and 9% of offers, more sourcing will not fix it.

Keep the data aggregated and handled lawfully, and be honest about small numbers — with eleven hires in a quarter, percentage swings are noise.

Recruiter Load and Requisition Ageing

Recruiter load is open requisitions ÷ FTE recruiters. Crude, but it answers a question leadership gets wrong constantly: is hiring slow because the process is bad, or because one person is running nineteen searches? Weight requisitions by difficulty — 1 for repeat volume roles, 2 for standard professional, 3 for niche or leadership.

Requisition ageing is days open per role. Sort descending and review the top five weekly. Anything past your escalation threshold needs one of three verdicts: change the requirements, change the compensation, or close it. Letting a role age quietly for 150 days is a decision too, just an unmanaged one.

India-Specific Realities That Distort Recruitment Metrics

Generic advice assumes two-week notice periods and one offer per candidate. Neither holds here.

Notice periods of 30 to 90 days. A role filled on 15 May with a 90-day-notice candidate is not staffed until mid-August. Track time to onboard — requisition approval to actual joining — or business planning stays permanently optimistic. If you fund notice buyouts, that is a cost-per-hire line item.

Multiple offers in play. Assume every strong candidate is in three or four processes. Being two weeks faster is often worth more than being ₹50,000 higher.

Offer shopping and backup offers. Some candidates use your letter as leverage elsewhere. You cannot eliminate it, but you can flag it: log a post-offer risk rating at offer stage, then check later how predictive it was.

Counter-offers. Retention counter-offers are common, especially in IT services and in-demand engineering skills. If that is your leading decline reason, the answer is usually not more money — it is building enough conviction about the role that money alone does not flip the decision.

Ghosting on the joining date. The candidate stops responding a week before joining. This is why "requisition closed" must mean joined, not accepted.

Background verification lead time. Employment, education and address checks take one to three weeks, longer with gaps or small former employers. Started at offer stage, a failure is a delay; started after joining, it becomes a termination. Track BGV turnaround and failure rate separately.

Tier-2 and remote hiring. Hiring into Indore, Coimbatore, Kochi or Bhubaneswar changes source mix, salary bands and acceptance rates. A single blended acceptance rate across metro and tier-2 hides both a problem and an opportunity.

Campus vs lateral. Two funnels with different physics — campus has huge volume, near-total acceptance at offer, then heavy drop-off across the long gap to joining. Blend them and you will conclude your acceptance rate is excellent and your drop-off catastrophic, both artefacts of the mixing.

Referral-heavy pipelines. Referrals convert well and cost little, and they narrow the pool toward people who resemble your existing team. If referrals exceed roughly half your hires, look hard at what your funnel no longer sees.

Conversion Maths: Working Backwards From the Hiring Plan

This takes twenty minutes and turns "we need 12 engineers" into "we need this much sourcing capacity", which is what determines whether the plan is achievable. Take your own trailing conversion rates and divide backwards from target joiners.

The numbers below are invented for a hypothetical 140-person B2B SaaS company in Pune hiring backend engineers. They demonstrate arithmetic, not benchmarks.

Three questions fall straight out of it.

Is that volume physically possible? Seven hundred and fifty sourced candidates and 300 genuine screens for one role family in a quarter is a serious operation. If one recruiter is also running six other requisitions, the plan is not a plan.

Which conversion rate is cheapest to improve? In this example, lifting acceptance from 70% to 85% cuts required top-of-funnel by roughly 18% for the same output. Improving conversion almost always beats adding volume.

Where is the leverage? Stages near the bottom, because everything above scales with them. Fixing offer-to-join drop-off is worth more than another job board subscription. Rebuild this table quarterly, per role family.

Cost per Hire: What to Include and Exclude

Cost per hire fails in two directions. Some teams count only agency fees, producing a flattering number. Others load in every tangential cost until it is un-comparable.

Include, external: agency and consultant fees, job board subscriptions, sourcing tool licences, assessment platforms, background verification charges, campus drive costs, referral bonuses paid, candidate travel reimbursement, ATS subscription apportioned to the period.

Include, internal: recruiter salaries apportioned to the period, TA leadership time, and an estimate of interviewer time. Interviewer time is the line most people skip and frequently the largest hidden cost.

Exclude: the new hire's own compensation, relocation and onboarding costs (track those separately as joining costs), training, laptops and assets, and the cost of failed hires — that belongs in quality-of-hire analysis.

The figures below are invented to show structure only.

Now segment it. Strip out the two agency-sourced niche roles and the remaining 15 hires cost roughly ₹1,50,000 each — that is the number to trend quarter over quarter, with the agency roles treated as a separate question about whether the search was worth the fee.

Always report cost per hire beside quality of hire and time to hire, because it is trivially easy to lower cost by hiring worse people more slowly. And calculate per joined hire — offers that never join are pure sunk cost, and burying that hides the value of fixing drop-off.

Making Quality of Hire Measurable Without Pseudoscience

Quality of hire fails when teams either refuse to measure it or over-engineer it into a fourteen-input index nobody trusts. The middle path is four components, measured at fixed intervals, reported as a distribution rather than one company-wide average.

Hiring manager satisfaction at 90 days. One question on a 1-5 scale with a mandatory one-line comment: knowing what you know now, would you hire this person again? Ask it identically every time.

Ramp-to-productivity. Define per role family what "fully productive" means and when it should happen — first closed deal, shipping independently to production, handling tickets unassisted at target quality. Measure actual weeks against expected weeks.

Early retention. Still employed and not on a performance plan at six months. Binary and unambiguous, and it captures the failures satisfaction surveys are too polite to record.

Performance rating at the first full cycle. Whatever your review produces at 6-12 months. If you have no formal cycle, a manager-assigned band at nine months is enough to start.

Weight the four equally — defensible and easy to explain. Then cut the score by things you can act on: source, recruiter, hiring manager, referral or not, and number of interview rounds. That is where quality of hire becomes a decision. If five-round candidates score no better than three-round candidates, you have found three weeks to give back.

Two cautions. Small numbers create false patterns, so do not judge a source on four hires. And never use quality of hire to rank recruiters punitively — it is driven heavily by role difficulty, salary band and manager quality, none of which the recruiter controls.

Diagnosing Four Common Failure Patterns

The Offer-to-Join Playbook

The window between acceptance and joining is 30 to 90 days, and by default you spend it doing nothing. That is the mistake.

Build an engagement cadence. Within 48 hours of acceptance the hiring manager calls personally — the manager, not the recruiter. In week one, send the team introduction and the 30-60-90 plan. Mid-notice, invite them to a team lunch or an all-hands. Two weeks out, confirm logistics, assets and reporting details in writing. One week out, confirm the date.

Track a joining-confidence flag at each touchpoint: candidates who become hard to reach, vague about their last working day, or suddenly curious about buyout policy are signalling something. Keep the runner-up warm too. And put a rupee figure on each drop-off once — the cadence will never again be treated as optional.

Building a Recruitment Dashboard People Actually Read

One dashboard for three audiences fails all three. Build three views.

Weekly operational view. Recruiters and TA lead, fifteen minutes, requisition-level. Every open role, days open, stage counts, the oldest candidate in each stage, pending feedback, offers outstanding, joiners due. Output is actions with names on them.

Monthly leadership view. TA lead, HR head, department heads, founders. Aggregate and comparative, always with a trailing three-month trend, because a single month at SMB volumes is mostly noise.

Quarterly strategic review. Leadership team, an hour, a proper document: quality of hire, early attrition by cohort, source effectiveness with cost, pipeline diversity, rebuilt conversion maths, and plan-versus-actual with an explanation of the gap.

One rule: nothing enters the monthly leadership view unless someone will say what they are doing differently because of it.

Data Hygiene and Instrumentation for ATS Reporting

Good recruitment metrics are 20% analysis and 80% instrumentation. Here is what your ATS or HRMS must capture.

Stage timestamps for every transition — the full history, not just current stage. Without it, time in stage is impossible and time to hire is a guess.

A rejection and withdrawal taxonomy. Free text is useless for analysis. Use a fixed list, short enough that recruiters use it honestly.

Source tagging discipline. One source per candidate per requisition, set at entry, never overwritten. The usual failure is attribution drift, where a candidate first found on a job board is later "sourced" by a consultant and both claim credit. Set a first-touch rule and document it.

Duplicate prevention. Deduplicate on email and phone at entry. One candidate appearing three times inflates your top of funnel and deflates every conversion rate, making a clean funnel look broken.

Requisition fields. Approval date, target joining date, budget band, level, location, hiring manager, recruiter, and a replacement-versus-new-headcount flag — replacement roles deserve separate analysis.

Offer and joining fields. Offer release date, acceptance date, agreed joining date, actual joining date, notice period at offer, BGV start and clear dates, post-offer risk flag.

Avoid double counting. Three usual errors: summing one candidate across two requisitions, counting a revised offer twice, and counting a joiner in the month they accepted rather than joined. Decide the rule once and audit a sample quarterly.

Smaller teams will not get all of this immediately. Prioritise stage timestamps, rejection reasons and source tagging — those three unlock most of the analysis.

Setting Realistic Targets Without Borrowed Benchmarks

You will be tempted to look up an industry average time to hire and adopt it. Resist. Published figures blend industries, company sizes, cities and definitions that almost certainly do not match yours. Baseline against yourself instead.

1. Pull trailing performance for the last two to four quarters of completed hires. The data will be imperfect; it is still yours.
2. Segment before setting targets — by role family and level at minimum. One time-to-hire target spanning an office assistant and a VP of Engineering is meaningless.
3. Set the target as a movement. "Cut median engineering time to hire from 42 to 34 days by end of Q3" is a target. "Time to hire: 30 days" is a slogan.
4. Name the mechanism. If you plan to remove eight days, say where — four from feedback SLA, three from offer release, one from scheduling. No mechanism, no target.
5. Recompute the baseline quarterly. A target set when you had one recruiter and four open roles will not survive the quarter you open twenty.

External benchmarks genuinely matter for compensation, because you compete on it directly. For process metrics, your own trailing performance is the honest comparator.

A 30-Day Plan to a Working Recruitment Dashboard

Week 1: Define and agree

1. Write the one-page stage definition document using the eight stages above, with entry and exit criteria.
2. Circulate it and get explicit agreement in a 30-minute meeting, not over email.
3. Build the rejection and withdrawal taxonomy — around 20 codes total.
4. Decide the source tagging rule (first touch, never overwritten) and document it.
5. Agree two SLAs: recruiter time to schedule, hiring manager time to give feedback.

Week 2: Instrument and clean

1. Configure stages and reason codes in your ATS or HRMS to match the definitions exactly.
2. Deduplicate the existing candidate database on email and phone.
3. Add the missing requisition and offer fields listed above.
4. Retro-fill last quarter's hires where reasonable — cap this at one day; partial history is fine.
5. Brief the team on same-day updating and explain why late entry corrupts data rather than merely annoying someone.

Week 3: Build the baseline

1. Calculate trailing metrics for two quarters: time to hire, time to fill, time in stage, screening pass rate, interview-to-offer ratio, offer acceptance, offer-to-join drop-off, cost per hire.
2. Build the conversion maths table backwards from your hiring plan, per role family.
3. Identify your three worst numbers and write one sentence each on the likely cause.
4. Compute cost per hire using the include and exclude list, segmented by role family.
5. Send a candidate experience survey to everyone who reached interview stage in the last 60 days.

Week 4: Ship the dashboard and the ritual

1. Build the weekly operational one-pager and run the first 15-minute standing meeting.
2. Build the monthly leadership one-pager with trailing three-month trends.
3. Set targets as movements with named mechanisms — top three problem metrics only.
4. Assign an owner to every decision metric.
5. Calendar the quarterly review, and start collecting 90-day hiring manager satisfaction responses so quality of hire has data by then.

You will not have perfect data after 30 days. You will have consistent definitions, an instrumented pipeline, a baseline and two rituals — more than most companies ten times your size manage.

Common Mistakes to Avoid

• Tracking everything. Forty metrics means zero decisions. Start with eight.
• Using averages instead of medians. One 180-day search wrecks an average. Report median and 90th percentile.
• Blending role families. Sales, engineering and finance funnels have different physics; a blended number averages things that should never be averaged.
• Blending campus and lateral. Different timelines, different failure modes. Keep them separate always.
• Closing a requisition at offer acceptance. In India the role is filled when the person joins. Closing early hides drop-off and makes an incomplete plan look complete.
• Optimising cost per hire in isolation. Cheap hires who leave in four months are the most expensive hires you will make.
• Assuming declines are always about money. Without reason codes you will keep raising salaries to fix a process-speed problem.
• Measuring recruiters on what they do not control. Acceptance is shaped by salary bands, brand and manager behaviour. Use it to diagnose the system, not to rank people.
• Letting data entry lag. Stale entries do not give conservative estimates, they give wrong ones.
• Ignoring rejected-candidate experience. Every rejected candidate is a future applicant, referrer or customer.
• Presenting metrics without a recommendation. If the slide does not end with "so we should", it does not belong in the meeting.

How CozyHR Helps You Run Recruitment Metrics Properly

Most of what breaks recruitment metrics in a growing company is not analytical. Requisitions live in one place, pipelines in another, offers in an email thread, and joiner data in a payroll spreadsheet someone retypes. Every handoff loses a timestamp.

CozyHR is built for Indian SMBs and growing teams, and the recruitment side is designed around that continuity:

• Requisition tracking with real dates — approval dates, budget bands, target joining dates and named hiring managers as structured fields, which makes time to fill and requisition ageing computable rather than estimated.
• Configurable pipeline stages matching the definitions you wrote in week one, with stage history retained so time in stage and conversion rates come out of the system rather than a manual reconstruction.
• Source tagging and rejection reasons captured consistently at entry and exit, so source effectiveness and decline analysis need no chasing months later.
• Offer and onboarding handoff in one place, with offer release, agreed joining and actual joining dates recorded — the data that makes offer-to-join drop-off visible instead of anecdotal.
• Joiner data flowing into HR and payroll records, so a candidate becomes an employee without retyping or mismatched IDs, and early attrition can be analysed by source and recruiter.
• Reports for leadership without exporting three files and building a pivot table the night before the review.

The point is not the dashboard. It is that the joins between requisition, candidate, offer, joiner and employee stay intact, because that is where hiring analytics quietly die.

Frequently Asked Questions

What are the most important recruitment metrics for a small company just starting out?

Start with five: time to hire, offer acceptance rate, offer-to-join drop-off, 90-day attrition and requisition ageing. Those cover speed, competitiveness, post-offer risk, hiring accuracy and what needs attention now. Add source effectiveness and cost per hire once you have a clean quarter, and quality of hire once you have enough joiners for it to mean anything.

What is the difference between time to fill and time to hire?

Time to fill runs from requisition approval to offer acceptance and captures the whole organisation's speed, including approvals and JD delays. Time to hire runs from a candidate entering the pipeline to that candidate accepting, capturing the process they actually experienced. Track both — the gap tells you whether the problem sits upstream of recruiting or inside it.

How do I reduce offer-to-join drop-off in India?

Treat the notice period as an active engagement window. Have the hiring manager call within 48 hours of acceptance, share a 30-60-90 plan early, stay in contact at fixed intervals, confirm logistics in writing two weeks out, and keep the runner-up warm. Record a drop-off reason every single time, because counter-offers, competing offers and personal reasons need completely different fixes.

Should I benchmark my recruitment metrics against industry averages?

For process metrics, no. Published averages blend industries, sizes, cities and definitions that will not match yours. Baseline against your own trailing two to four quarters, segmented by role family and level, and set targets as movements from that baseline. Compensation is the exception, where external market data genuinely matters.

How do I measure quality of hire without it becoming guesswork?

Combine four measurable components: hiring manager satisfaction at 90 days, ramp-to-productivity against a milestone defined in advance, retention at six months, and the first full performance rating at 6-12 months. Weight them equally, report the distribution rather than one average, and cut the results by source, recruiter, hiring manager and number of rounds so the finding leads somewhere.

What should my ATS reporting capture at minimum?

Three things above all: stage transition timestamps with full history, a fixed rejection and withdrawal taxonomy, and a first-touch source tag that is never overwritten. Add requisition approval date, offer release date, agreed joining date and actual joining date as soon as you can. Without timestamps you cannot compute time in stage; without reason codes you cannot diagnose anything.

How often should I review my recruitment dashboard?

Three cadences. Weekly, a fifteen-minute requisition-level review producing named actions. Monthly, a one-page leadership view with trailing three-month trends. Quarterly, a strategic review covering quality of hire, early attrition, source effectiveness with cost, pipeline diversity and rebuilt conversion maths. Anything you cannot act on at that cadence does not belong in that view.

My hiring managers do not give interview feedback on time. Is that really a metric?

Yes, and one of the highest-return ones available. Measure median hours from interview completion to feedback, set a 24-hour SLA, and report adherence by hiring manager alongside recruiters' own SLAs. Making it visible and mutual changes behaviour faster than almost any process redesign, and feedback delay is frequently the single biggest contributor to slow hiring.

Conclusion: Measure Less, Decide More

The goal of recruitment metrics is not a beautiful dashboard. It is a shorter list of arguments. When everyone agrees what "screened" means, when every rejection carries a reason code, and when offer-to-join drop-off sits on the same page as time to hire, hiring stops being a matter of opinion and becomes a system you can improve.

Start narrow. Define your eight stages this week, instrument timestamps, reason codes and source tags next week, build the baseline in week three, ship two rituals in week four. Set targets as movements from your own history, each with a named mechanism. Then delete anything that has not changed a decision in two quarters.

If your real bottleneck is that requisitions, candidates, offers and joiner records live in four disconnected places, fix that before buying an analytics layer on top of the mess. CozyHR keeps requisition tracking, pipeline stages, offers, onboarding and the handoff into HR and payroll records in one connected system, so your hiring funnel analytics come from data captured correctly the first time.

Take a look at CozyHR and see what your hiring funnel actually looks like when the numbers are clean.