Recruitment Analytics Every HR Manager Should Monitor

Most recruitment dashboards report activity. Applications received, profiles screened, interviews scheduled, offers rolled. It looks like measurement, and every number on it is true, but a hiring manager reading it on Monday morning still has no idea what to do differently this week.

Recruitment analytics only earns its place when a number changes a decision. Reopen the sourcing channel or close it. Add an interviewer or take a stage out. Push the requisition approval two weeks earlier or accept that the role lands after the quarter. Everything else is reporting, and reporting is what recruitment teams already drown in.

This piece sets out the metrics worth monitoring, how each one is actually calculated, what a bad number is telling you, and how often to look at it. It also covers the ones that mislead, because the wrong metric on a dashboard does more damage than no dashboard at all.

One thing up front. Almost every recruitment metric is a ratio, and ratios need volume to mean anything. A company making 200 hires a year can read monthly conversion rates. A company making 20 cannot, and should be reading counts and rolling twelve month figures instead. We've flagged where that distinction bites, because one dashboard does not fit both. If you want the wider picture of how this data connects to the rest of the employee record, our guide to data-driven HR covers what sits downstream of hiring.

What follows is shaped by the hiring funnels we see during implementation. Engage runs recruitment for more than 200 companies across 15 states, in manufacturing, IT services, retail and hospitality and BFSI, and the same handful of measurement defects turn up in dashboard after dashboard. Where something is flagged here as a common failure, that's what we've seen in real ATS data.

What Recruitment Analytics Is Actually For

Three questions justify the entire exercise.

Will this role close in time? Where is the pipeline leaking? And are the people we hire working out?

Every metric worth tracking answers one of those. Speed metrics answer the first, funnel and source metrics answer the second, and quality metrics answer the third. If you can't place a metric under one of the three questions, it belongs in a report, not on a dashboard.

It also helps to separate three kinds of numbers, because teams routinely mix them and then wonder why the dashboard doesn't drive anything. Activity metrics count what the team did. Efficiency metrics measure how well the process converts. Outcome metrics measure whether the hire was any good. Activity is the easiest to collect and the least useful on its own, which is exactly why most dashboards are full of it.

Here is the working set. Twelve metrics, and most companies need fewer.

Notice that only one of those is a weekly metric. Weekly review is for operational triage, not for trend reading, and a conversion rate recalculated every Monday is mostly noise wearing a percentage sign.

The Data That Has to Be Right First

Every broken recruitment dashboard we've unpicked has been broken for the same reason. Not the formula, the underlying fields. Six of them matter, and if these are inconsistent nothing downstream can be trusted.

Requisition open date. Decide whether the clock starts when the business asks, when finance approves, or when the job goes live. All three are defensible. Only one can be in the system, and the gap between the first and third is often three weeks that nobody counts. Recruiters like the posting date because it flatters time to fill. The business wants the approval date because that is when it started waiting.

Stage timestamps. Every stage transition needs its own date. Without them you get one aggregate duration and no ability to say which stage is slow, which is the only version of the number you can act on.

Source of hire. Fix whether you attribute to first touch or last touch, and write it down. A candidate sourced on LinkedIn who later applies through the careers page will be counted twice or credited to the wrong channel, and the entire source comparison in section 6 depends on getting this consistent rather than getting it philosophically right.

Rejection reasons. As a mandatory picklist with fewer than ten options, not a free text box. Free text rejection reasons are unreadable at any volume, and they are the field that tells you whether you are screening out the wrong people.

Offer decline reasons. Same rule, separate field. Compensation, counter-offer, competing offer, location, role scope, process too slow. Six options and an other. This is the highest value field in your entire recruitment dataset and most teams don't capture it at all.

Joining date, separate from offer acceptance date. Two fields, not one. In India the gap between them is typically 30 to 90 days of notice period, and that gap is where offers go to die.

One practical warning about migrations. When you move from spreadsheets or an old ATS into a new system, application dates usually import as the migration date, which silently destroys a year of historical time to hire. We hit this on most migrations. The tell is a time to hire chart that is flat and implausibly short for everything before the cutover and then jumps, and it has to be caught before go-live because the original dates are rarely recoverable once the import has run. Either keep the old dates in a dedicated field or accept that your baseline starts on migration day and say so on the dashboard. If you're at that stage now, our note on replacing spreadsheets with HR software covers what else tends to break in the move.

Time to Fill, Time to Hire and the Gap Between Them

These two get used interchangeably and they measure different things, which is why the same team can be excellent at one and terrible at the other.

Time to fill is a business planning metric. It runs from requisition approval to offer acceptance and it includes everything, the two weeks the job description sat unwritten, the delay finding an interview panel, the approval loop on the salary. It is the number to give a department head who asks when they will have somebody.

Time to hire is a process metric. It runs from the day a specific candidate enters the pipeline to the day they accept, and it measures how well your process handles a good candidate once you have one. It is the number that predicts whether you lose people to faster competitors.

The third number, and the one Indian employers most often leave off the dashboard, is the offer to join gap. Notice periods of 30, 60 and 90 days are standard, and the candidate is available to the market for that entire window. A team with a strong time to hire and no view of the joining gap is repeatedly surprised in month three.

Use the median, not the average. One senior role that took 180 days will drag an average of thirty hires by a week and hide the fact that most roles close in five. If you have the volume, also look at the 90th percentile, because that tail is where the business loses patience with the recruitment team.

Segment by role family before you draw any conclusion. Engineering, sales, finance and plant roles have structurally different market depths, and a single company-wide time to fill compares things that were never comparable. A blended 42 days can be a healthy sales pipeline and a broken engineering one.

For a rough external reference, white collar hiring in India is commonly cited at 28 to 45 days from requisition to acceptance, with mid-level roles at the lower end and senior or niche technical searches stretching well past 80 days. Treat those as a smell test rather than a target. If your engineering time to fill is 90 days you are in normal territory for that market, and if your customer support roles are also taking 90 days you have a process problem that no market condition explains.

When time to fill is bad, break it into stage durations before you do anything else. The delay is almost never spread evenly, and in our experience it clusters in three places: the requisition approval to posting gap, the wait for interview slots from busy panel members, and the offer approval loop. All three are internal. None of them are candidate supply, which is where teams instinctively look first.

Cost per Hire and What Belongs In It

Cost per hire is the sum of internal and external recruiting costs in a period, divided by the number of hires made in that period. The formula is not the hard part. What goes into the numerator is.

External costs are the ones everybody counts: agency fees, job board subscriptions, assessment tools, background verification, career fairs, referral bonuses. Agency fees dominate wherever they exist. The Indian market prices them as a share of annual CTC, and the convention is close to standard: 8.33%, meaning one month's salary, for entry level and bulk hiring, around 12.5% for mid-level and specialist roles, and 16.67%, meaning two months, for senior and niche positions. Executive search sits higher again. Add 18% GST on top, which is the part most cost per hire calculations quietly leave out.

Internal costs are the ones most companies quietly omit. Recruiter salaries apportioned to the period, the ATS licence, and interviewer time. That last one is the largest hidden cost in the whole function and almost nobody prices it.

Work it through for a single senior engineering hire. Say eight candidates clear the recruiter screen and sit a first technical round, four go through to a second, and two reach the hiring manager and a leadership conversation. Count an hour for each interview and half an hour of preparation and debrief around it, which is conservative. That is roughly 23 hours of interviewer time to make one hire. At the salary bands of the people conducting those interviews, senior engineers on 25 to 30 lakh, an hour of interviewer time costs somewhere around Rs 1,500, so the loop has consumed about Rs 35,000 before a single rupee is spent externally.

Those are illustrative numbers, not a benchmark. What matters is the method, so run it with your own funnel ratios and your own salary bands. Two things usually fall out of it. The cost of a redundant interview stage becomes visible and arguable, and the true cost of a low screen-to-interview conversion rate stops being an abstraction, because every unqualified candidate who reaches a panel is spending the most expensive hours in the company.

Two cautions before you put any of this on a dashboard. Cost per hire is the easiest recruitment metric to improve and one of the easiest to improve destructively. Drop the agencies, cut the assessments, shorten the process, and the number falls immediately while quality falls later and less visibly. It should never be read without a quality metric next to it.

And the aggregate figure is almost useless on its own. Cost per hire by source and by role family is where the decision lives. If agency hires cost eight times a referral and retain no better at twelve months, that is an actionable finding. A single company-wide figure of Rs 62,000 supports no decision at all.

Funnel Conversion, Stage by Stage

Stage conversion is the one place recruitment analytics diagnoses rather than describes. Each pass-through rate has a specific failure mode attached to it, so an unusual number points at a cause instead of just registering a problem.

Read the funnel in candidates, not percentages, when volumes are small. Ten applications to one hire is a ratio you can act on. A "10% application to hire conversion" calculated from ten applications is a number pretending to be a rate.

The single most useful derived figure here is the number of candidates you need at the top of the funnel to produce one hire, worked backwards through your own conversion rates. Once you have it for each role family, capacity planning stops being a guess. If a senior backend role historically needs 60 sourced profiles to produce one accepted offer, and you have 15 in the pipeline in week three, you already know the answer without waiting for the outcome.

Source Quality, Not Source Volume

Source reporting goes wrong when it counts applicants. Channels that produce enormous application volume and almost no hires look good on a volume chart and consume the screening capacity of the entire team.

Judge every channel on hires, and then on what happened to those hires.

The comparison that changes budgets is cost per hire against twelve month retention, by channel. Cheap channels that produce hires who leave in eight months are not cheap, and expensive channels that produce three year employees are usually justified. Neither claim can be made without joining recruitment data to exit data, which is the single most valuable integration in the whole stack and the one most often missing because the ATS and the HR system are separate products. Where hiring, onboarding and exit sit in one employee lifecycle record, this analysis takes minutes rather than a quarter.

Offer Acceptance, Declines and the Candidates Who Never Join

Offer acceptance rate is offers accepted divided by offers extended. For a reference point, Ashby's analysis of 230,000 applications that reached the offer stage put the average at 78%, with technical roles at 73% and business roles at 84%. So an engineering acceptance rate in the low 70s is unremarkable. Below 70% across the board, something upstream is broken and the candidates are not the problem.

What that something is comes from the decline reasons, which is why the picklist matters. The three common patterns look identical in the acceptance rate and need opposite responses.

If declines cluster on compensation, the problem is benchmarking, and it is usually structural rather than one stingy offer. If declines cluster on competing offers, the problem is speed, and you fix it by moving faster, not by paying more. If declines cluster on role scope or reporting line, the problem is that the job the candidate interviewed for is not the job on the offer letter, which is a hiring manager conversation.

Then there is the metric Indian employers need and international frameworks mostly ignore, because the pattern is far less common in markets with two week notice periods.

Joining ratio. Of the candidates who accepted an offer, how many actually turned up. With 60 or 90 day notice periods, an accepted offer is not a closed role, it is a claim on a candidate who remains fully available to the market and increasingly likely to receive a counter-offer from an employer who now knows they are leaving. A team with a 92% acceptance rate and a 75% joining ratio is filling three roles for every four it believes it has filled, and will find out about it eight weeks after everyone stopped worrying about those requisitions.

The scale of this is not marginal. Gartner's survey of nearly 3,000 candidates found 35% had backed out after accepting an offer in the first quarter of 2025, down from 48% a year earlier as the market softened. That is a global figure, in markets where two weeks of notice is normal. Indian notice periods of 60 and 90 days extend the window in which all of that can happen by a factor of four or five, which is why a metric most international frameworks treat as a footnote deserves its own line on an Indian dashboard.

Track the two separately, always, and track the drop-off by week of the notice period. If the losses cluster in the last two weeks, the counter-offer is winning. If they cluster in the first two, the candidate was negotiating with your offer in hand the whole time. The response to a poor joining ratio is not a better offer, it is structured contact through the notice period: documentation, team introductions, a manager call, onboarding access before day one. It is unglamorous, it takes recruiter hours, and it is the highest return activity in the entire funnel because the cost of a renege is the whole requisition, restarted.

Quality of Hire, Without Waiting for a Perfect Definition

Quality of hire is the metric everybody agrees matters and almost nobody tracks, because the definitional argument consumes the effort that measurement would have taken. There is no standard formula and there does not need to be one. What you need is a number that is consistent year to year, not a number that is philosophically correct.

Start with three inputs, all of which you can collect without new tooling.

Hiring manager satisfaction at 90 days. One question, one to five, sent automatically. "Knowing what you know now, would you hire this person again?" works better than a rating scale because it is a decision rather than an opinion. Response rates hold up because it takes eleven seconds.

First year retention of new hires. Unambiguous, already in your system, and the closest thing to a ground truth the function has.

Time to productivity. Days until the hiring manager considers the person independently effective. It is subjective, which is fine, because you are comparing it across sources and recruiters, not auditing it.

Average those into a single index if you want one number, but keep the components visible, because the index moves for reasons the components explain.

The sharpest single indicator in the set is early attrition, meaning exits within six months. It is worth a separate line on the dashboard because it usually is not a screening failure at all. People who leave inside six months typically left because the role was described differently in the interview than it turned out to be, or because of the manager, and both causes are fixable in a way that "we hired the wrong person" is not. Cut early attrition by hiring manager and the pattern is often uncomfortable and immediately obvious.

One structural point. Quality of hire is a lagging metric by definition, so it cannot manage current hiring. Its job is to validate or invalidate everything else on the dashboard. When cost per hire and time to fill are both improving while quality of hire drifts down, you are not getting better at hiring, you are getting faster at lowering the bar. That combination is worth watching for specifically, because both of the improving metrics get reported upward and the drifting one usually doesn't.

Reading Diversity Data by Stage

Reporting the share of women in this year's hires tells you the outcome and nothing about the cause. The useful version is the same funnel from section 5, run separately for each group, compared at every stage.

That comparison localises the problem. If the application pool is 15% women and offers are 15% women, the process is neutral and the issue is sourcing and job description language. If applications are 30% and offers are 12%, the loss is inside your process and you can see exactly which stage it happens at. Those two situations produce the same headline number and need completely different work.

A rough diagnostic that borrows from selection ratio analysis: if any group's pass-through rate at a stage falls below about four fifths of the highest group's rate, that stage is worth examining. It is not a legal threshold in India, and it is not proof of anything, but as a screen for where to look it is more useful than eyeballing percentages.

Two cautions. Targets set at the offer stage tend to get met by adjusting the top of the funnel instead of the process, which produces a better number and no change. And be deliberate about what you collect. Gender is standard in Indian HR data. Caste and religion are not, and collecting them for analytics without a clear, documented and lawful purpose creates risk that outweighs the insight. Measure what you can act on and can justify holding. Our note on recruitment software and inclusive hiring covers the process side of this.

Building a Dashboard People Actually Open

The failure mode is a dashboard with 30 tiles that nobody reads twice. Three views beat one page, split by what the reader can do about it.

Weekly, for the recruitment team. Open requisitions by age, candidates stuck in a stage more than seven days, interviews awaiting scheduling, offers pending decision. All counts, all named, all actionable on Monday. No percentages, no trends.

Monthly, for HR and hiring managers. Median time to fill and time to hire by role family, funnel conversion by stage, offer acceptance and joining ratio, source mix by hires. This is the review where you change how the process runs.

Quarterly, for leadership. Cost per hire by source, quality of hire, early attrition, hires against plan, recruiter capacity. This is the review where budget moves.

Three habits separate a dashboard from a wallchart.

Every metric gets a target or a comparison. A number with nothing to be measured against cannot be read as good or bad, and readers stop trying after the second look.

Report medians with a count next to them. "38 days, n=6" is honest. "38 days" from six hires implies a precision that isn't there.

If you make fewer than about 50 hires a year, run everything on a rolling twelve months and report counts rather than rates. Monthly percentages on low volume swing wildly, and a team that gets burned twice by explaining a swing that turned out to be one delayed offer will stop looking at the dashboard entirely.

Where Recruitment Analytics Goes Wrong

These are in rough order of how often they turn up, not in order of severity. The last one is the most damaging and the least common.

Measuring activity and calling it performance. Profiles screened and interviews scheduled measure how busy the team was. Neither correlates with hiring well, and both rise when the process is inefficient. This is close to universal, because activity data is the easiest thing an ATS exports.

One company-wide number across role families. A blended time to fill compares an engineering search with a customer support hire. The blend is always fine while one half of it is on fire.

Not capturing decline reasons. The acceptance rate tells you that you are losing candidates. Only the decline reason tells you whether to fix pay, speed or the job description, and it costs one picklist field to collect. It is the cheapest fix on this list and the one most consistently missing.

Averages hiding the tail. One 180 day senior search distorts every average it touches. The median is the honest number and the 90th percentile is the one the business actually feels.

Treating the accepted offer as a closed role. The joining ratio is a separate metric for a reason, and in a market with 60 to 90 day notice periods it is where forecasts break.

Recruitment data that never meets performance and exit data. Without that join, quality of hire and source retention are unanswerable, and the function is left arguing for budget on cost and speed alone.

Optimising cost and speed without a quality counterweight. Rarer than the rest, because it takes a team disciplined enough to be measuring properly in the first place, and worse than the rest for exactly that reason. Both numbers improve immediately when the bar drops, both get reported upward, and the consequence surfaces two or three quarters later against a metric fewer people are watching. Any speed or cost target needs a quality metric published beside it.

Questions People Ask

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

Time to fill runs from requisition approval to offer acceptance and measures the whole hiring cycle including internal delays. Time to hire runs from a candidate entering the pipeline to their acceptance and measures how quickly your process moves once a suitable candidate exists. Time to fill is the number to give the business for planning. Time to hire is the number that tells you whether you are losing candidates to faster competitors.

How is cost per hire calculated?

Add all internal recruiting costs, meaning recruiter salaries apportioned to the period, ATS licences and interviewer time, to all external costs, meaning agency fees, job boards, assessments, background verification and referral bonuses, then divide by the number of hires made in that period. Interviewer time is the component most often left out and frequently the largest. Calculate it by source and role family rather than as a single company figure, because the aggregate number supports no decision.

What is a good offer acceptance rate?

Ashby's analysis of 230,000 applications reaching the offer stage puts the average at 78%, with technical roles at 73% and business roles at 84%. So a rate in the mid to high 70s is unremarkable, and engineering will sit below sales on the same dashboard for structural reasons. Below 70% across all functions, something upstream is wrong, usually compensation benchmarking or a process slow enough for competing offers to arrive. The rate alone does not tell you which, which is why decline reasons need to be a structured field rather than free text.

Which recruitment metrics matter most for a small company?

Four. Time to fill by role, offer acceptance rate, joining ratio and six month attrition. Everything else needs hiring volume to be meaningful. Run all four on a rolling twelve month basis and report counts alongside any percentage, because monthly rates calculated from three or four hires are noise.

How do you measure quality of hire?

There is no standard formula, and consistency matters more than correctness. A workable composite is a 90 day hiring manager rating, first year retention and time to productivity, averaged into one index with the components kept visible. Track early attrition, meaning exits within six months, as a separate line, because it usually points to expectation setting or the manager rather than to screening.

Why do candidates accept offers and not join?

Because a 30 to 90 day notice period leaves them available to the market for the whole of it, and their current employer now knows they are leaving and can counter-offer. Gartner found 35% of candidates globally had backed out after accepting an offer in early 2025, and that is in markets where notice periods are measured in weeks rather than months. Track the joining ratio separately from the acceptance rate and look at which week of the notice period the drop-offs cluster in. Losses late in the notice period point to counter-offers, losses early on point to a competing offer that was already in play. The effective response is structured contact through the notice period rather than a higher offer.

How much data do you need before recruitment analytics is useful?

For ratios like conversion and acceptance rates, roughly 30 hires or a few hundred candidates in the comparison group before the percentages stop swinging on individual cases. Below that, counts, medians and stage durations still work, and pipeline ratios of the "we need 40 profiles to make one hire" kind remain useful because they are built from candidate volumes rather than hire volumes.

Which recruitment metric is the most commonly misused?

Cost per hire. It improves the moment you cut agencies, assessments or interview stages, and the resulting quality drop shows up two or three quarters later against a different metric that fewer people are watching. It is a genuinely useful number when it is broken down by source and read next to retention, and a misleading one as a single company figure with a downward arrow next to it.

Where This Leaves You

Recruitment analytics is not a reporting problem. It is a data hygiene problem followed by a discipline problem, and both are smaller than they look.

So work in that order. Fix the six fields in section 2 so the numbers mean the same thing every month. Pick four metrics rather than twelve, one for speed, one for the funnel, one for closing and one for quality. Split the dashboard by who reads it and what they can change. Then hold the rule that no speed or cost metric gets reported without a quality metric beside it, because that pairing is the only thing standing between an efficient recruitment function and a fast one that hires badly.

What makes this hard in practice is that the answers span systems. Time to hire lives in the ATS, retention lives in the HR record, and quality of hire needs both, so the single most valuable thing most teams can do is put hiring, onboarding and exit data in one place. Our guides on AI in recruitment and on employee retention strategy cover what to do with the pattern once the data shows you one.

If you'd rather your hiring funnel, offers and joining tracking ran out of the same system that already holds attendance, payroll and exits, that's what our recruitment software is built to do. Book a free demo and bring your last twenty hires. It's a faster conversation when you can see your own funnel and joining ratios on screen rather than a feature list.