How to Measure Candidate Experience: The 12 Metrics That Matter and the Tools to Track Them

TL;DR

• How to measure candidate experience: track three layers together – behaviour (what candidates do), sentiment (what they say) and fairness (whether the experience differs by group) – by stage and by outcome. One blended score hides where the problems are.
• The 12 metrics: application completion rate, stage withdrawal rate, time to first response, time to outcome, candidate NPS, stage satisfaction, outcome and feedback coverage, public review sentiment, completion and pass-through by group, accessibility adjustments, offer acceptance rate and early attrition.
• Benchmarks: most of the figures quoted online are vendor-owned, years old or impossible to trace. Use them as a rough guide and benchmark against yourself, by segment, over time.
• Where the experience breaks: early. The application and screening stages hold most of your candidates and the lowest scores. In the 2025 CandE research, North American candidates who weren’t selected gave a candidate NPS of −28, and 56% of candidates rejected at screening or interview got no feedback (vendor-run research).
• Tools: most teams need three – an ATS or analytics tool for behaviour, a survey tool that reaches rejected candidates for sentiment, and visibility at the screening stage, where volume is highest.

Ask most TA teams how they measure candidate experience and you’ll hear two answers: a survey after interview, and time to hire.

That leaves big gaps. Most job applicants never reach interview, so a post-interview survey misses the majority of the people who apply. Time to hire tells you how fast you fill roles, not how candidates feel about the wait. And neither says anything about whether the experience is the same for every group of candidates.

A second problem: many of the statistics in candidate experience guides are a decade old or have no source at all.

This guide gives you a three-layer model for evaluating the candidate experience, 12 metrics with formulas and the best benchmarks we could find (or a plain statement that none exists), a six-step plan you can run in a quarter, and a quick snapshot of the tools that track each layer.

A disclosure: Sapia.ai makes one of the tools in this guide. We run structured chat interviews at the screening stage for high-volume employers and have assessed 10M+ candidates (vendor-reported). That’s the stage where most candidate experience is won or lost, so it’s the vantage point this guide is written from.

What does it mean to measure candidate experience?

Measuring candidate experience means tracking how candidates move through your hiring process, how they feel about it, and whether that differs by group – at every stage and for every outcome, including the candidates you reject.

In practice, that means three layers of data:

A fourth, lagging layer – outcomes such as offer acceptance and early attrition – shows whether a good experience turns into hires who stay.

Behaviour data is quantitative. Sentiment combines quantitative and qualitative data – scores plus candidates’ own words – and the comments are often where the useful detail sits.

The layers work together. Behaviour tells you where candidates leave. Sentiment tells you why. Fairness tells you who. A programme that tracks only survey scores, or only funnel numbers, gives you half the picture.

The 12 candidate experience metrics that matter

Here’s the full set at a glance. Each metric is explained below, with its formula, the best available benchmark and what to watch out for.

Behaviour metrics: what candidates do

1. Application completion rate

What it is: the share of candidates who start an application and finish it.

Formula: completed applications ÷ applications started × 100.

Benchmark: there’s no credible, current benchmark for start-to-complete rates. Be careful with “apply rate” figures from job-advertising reports – they measure something else. Appcast’s 2026 US Benchmark Report, for example, measures clicks on job ads that turn into applications. On 2025 data from more than 300 million clicks, it put the apply rate at 7.14% for technology roles and 3.22% for healthcare (SIA, vendor-reported). That’s useful for judging job ads. It isn’t a completion rate.

Watch out: the much-quoted “365% uplift” and “60% abandon long applications” stats come from a 2016 SHRM article – a decade old. A more recent signal: in a November 2025 LiveCareer survey of 500+ job seekers, 57% said they’d abandoned an application midway because it was too complex (HR Dive, vendor-reported).

What to do: track your own completion rate by device and by step. Many ATSs and career-site tools can show you the exact step in the application process where candidates give up. The usual pain points are forced account creation, retyping a CV the candidate has already uploaded, and technical glitches on mobile.

2. Stage withdrawal rate

What it is: the share of candidates who pull out of your process at each stage, as opposed to being rejected. You’ll also see it called drop-off rate.

Formula: candidate withdrawals at a stage ÷ candidates entering that stage × 100. Ask for, or record, a reason every time.

Benchmark: none published by stage. For context, the most common reason candidates withdrew in the 2025 CandE research was that their time was disrespected during the interview process – interviews and appointments – cited by 32% of those who withdrew (vendor-run). And in Greenhouse’s 2026 survey of 2,950 job seekers, 38% of US candidates said they had withdrawn from a hiring process because it included an AI interview (Greenhouse, vendor-reported).

Watch out: a falling withdrawal rate can mean a better experience, or just a tougher job market where candidates have fewer options. Read it alongside time in stage. And look at who withdraws, not just how many candidates: if your most qualified candidates pull out at a higher rate than others, you’re losing top candidates to faster competitors.

3. Time to first response

What it is: how long candidates wait for a meaningful first contact after applying – often tracked as recruiter response time. Decide what “meaningful” means – an automated “we’ve received your application” email probably doesn’t count.

Formula: median hours (or days) from the moment a candidate applies to first meaningful contact.

Benchmark: none published. For context, a 2024 Robert Half survey of 1,000 Australian office workers found 57% lose interest in a role if they don’t hear back within 10 days (SIA, vendor-reported). Timely communication matters most for the candidates you most want to hire, because they usually have other options.

4. Time to outcome and time in stage

What it is: how long candidates wait to find out whether they’re through or out, and how long they spend at each stage.

Formula: median days from application (or from entering a stage) to the candidate being told the result. Also report the share of candidates still waiting after 30 and 60 days.

Benchmark: Ashby’s 2026 benchmarks, based on 54 million applications from January 2021 to March 2026, found non-interviewed candidates are archived in a median of about six days, and interviewed candidates in an average of 21 days (Ashby, vendor data). Speed shows up in sentiment: non-interviewed candidates archived within a day gave a median score of 9 out of 10 when asked whether they’d recommend applying. By 50 days or more, the median had fallen to 3.

In the 2025 CandE research, 60% of the top 10 North American award winners gave candidates an outcome within three to five days – yet 31% of North American candidates overall hadn’t heard back one to two months or more after applying (vendor-run).

Break it down by owner: split time in stage by who holds the next action – recruiter, hiring manager or the wider hiring team. Delays often sit with whoever has to review applications or return interview feedback, and that’s where a reminder or a deadline helps most.

Watch out: don’t confuse time to outcome with time to fill. SHRM’s 2025 benchmarking puts median time to fill at around 45 days (SHRM, member survey), but that measures how long a role stays open, not how long a candidate waits.

Sentiment metrics: what candidates say

5. Candidate NPS, by outcome and stage

What it is: candidate net promoter score (cNPS) asks candidates how likely they are, from 0 to 10, to recommend applying to your organisation.

Formula: % promoters (9–10) − % detractors (0–6), giving a score from −100 to +100.

Benchmark: there’s no independent cNPS benchmark, and vendor figures vary widely depending on who they survey. The 2025 CandE research put North America at −4 overall and −28 for candidates who weren’t selected (Survale, vendor-run). Ashby’s customers averaged +37 on 2022–23 data (Ashby, vendor data).

Watch out: always report cNPS separately for hired, rejected and withdrawn candidates, and by stage. A single blended score can look healthy while hiding a deeply negative score among the candidates you reject early – who are usually the biggest group. Our guide to candidate experience NPS [link once live] covers the formula, benchmarks and survey design in depth.

6. Stage satisfaction and driver scores

What it is: a short candidate satisfaction score at the end of a step (“How would you rate this step?”, out of 10) plus two to four statements candidates agree or disagree with, such as:

• “I was kept informed about my application status.”
• “The process felt fair.”
• “The questions were relevant to the job.”
• “The role matched the job description.”
• “It was easy to complete on my phone.”

Formula: average rating per stage, and the percentage agreeing with each statement. Low agreement on the job-description statement is a sign you’re setting the wrong candidate expectations upfront.

Benchmark: none that’s comparable across employers.

Watch out: in-process ratings come only from candidates who finished the step, and they’re collected before any decision. They’ll always look better than a cNPS sent after the decision, and the two can’t be converted. Keep them as separate metrics and show response rates next to every score. In any candidate experience surveys, add one open question – “What’s the main reason for your score?” – because that qualitative data tells you what to fix. For question ideas, see our candidate experience survey sample questions and our guide to candidate satisfaction surveys.

7. Outcome and feedback coverage

What it is: two simple ratios that test your candidate communication and that you can improve quickly – the share of candidates who are told their outcome (the opposite of ghosting), and the share of unsuccessful candidates who get feedback. For most applicants, the rejection message is the lasting impression of your organisation.

Formula: candidates told their outcome ÷ all candidates; rejected candidates given feedback ÷ all rejected candidates.

Benchmark: the 2025 CandE research found (Survale, vendor-run):

• 56% of candidates got no feedback after being rejected at screening or interview
• only 6% got feedback on assessment or test results
• 33% of rejected candidates were told via a do-not-reply email address
• feedback raised candidates’ willingness to refer others by 50–75%

You don’t need a long report to provide feedback. A short, specific note on what the role needed, sent promptly, is more useful than a generic “we’ve gone with other candidates” – and timely feedback is the cheapest improvement on this list.

8. Public review sentiment

What it is: what candidates say about your interviews on sites such as Glassdoor and Indeed – the share of positive, neutral and negative reviews, tracked over time.

It’s the metric closest to your employer brand: a negative experience shared publicly is how a hiring problem becomes a poor reputation. In PeopleScout’s 2026 research, 37% of job seekers said they’d tell others not to apply after a bad candidate experience (TrueBlue, vendor-reported).

Watch out: reviews are a useful early warning, but a weak primary metric. People who choose to leave reviews tend to have had very good or very bad experiences, and research using Glassdoor data found voluntary reviews skew towards the extremes (Marinescu et al., 2021, independent). Most employers also get too few reviews to read much into, and you can’t link a review to a stage or outcome. One recent change: Glassdoor was legally merged into Indeed on 1 July 2026, though the brand and anonymous reviews continue (Reworked).

Fairness metrics: whether the experience differs by group

9. Completion and pass-through rate by group

What it is: completion and pass-through rates compared across groups at each stage of your selection process – especially automated steps such as screening questions, assessments and AI interviews – using demographic data candidates choose to share.

Formula: impact ratio = a group’s rate ÷ the rate of the group with the highest rate. Under the US federal Uniform Guidelines, a selection rate below four-fifths (80%) of the highest group’s rate is “generally” treated as evidence of adverse impact (29 CFR 1607.4(D)). It’s a rule of thumb, not a legal test or a safe harbour.

Why it’s a candidate experience metric: if one group drops out of a step far more often than another, that’s an experience problem before it’s a compliance problem. The step may be harder to access, take longer than some candidates can give, or feel less relevant to them.

Watch out: averages can hide problems. A peer-reviewed study presented at ACM FAccT 2026, covering about 4.2 million applications screened by a game-based assessment between 2018 and 2022, found racial disparities at the level of individual roles that pooled, all-roles analysis had masked (independent). The lesson applies to every tool, including ours: analyse by role or role family, and ask every vendor for role-level results.

The legal backdrop (not legal advice – check with counsel): US federal disparate-impact enforcement has been deprioritised since Executive Order 14281 (April 2025), but Title VII is unchanged and private claims remain. Illinois HB 3773 (in force January 2026) bars AI use with a discriminatory effect in hiring and requires notice; New York City’s Local Law 144 requires annual bias audits of automated hiring tools; Colorado’s SB 26-189 applies from January 2027; and the EU AI Act’s high-risk employment obligations apply from 2 December 2027.

10. Accessibility adjustments

What it is: how often candidates ask for reasonable adjustments, how quickly you provide them, and whether those candidates complete and progress at the same rate as others.

Formula: adjustment requests ÷ applicants; median days to provide the adjustment; completion and pass-through rates for candidates who asked.

Benchmark: none credible. For context, a 2023 PeopleScout audit found only 30% of employers clearly said, before the application, that adjustments were available (TrueBlue, vendor-reported). In the UK, the Equality Act 2010 requires employers to make reasonable adjustments, and that duty covers recruitment.

Watch out: timed steps, and steps that need a camera or a microphone, are where adjustment requests tend to cluster. If one step generates most of your requests, look at the step.

Outcome metrics: lagging signals

11. Offer acceptance rate

What it is: the share of offers candidates accept. Record a reason for every declined job offer.

Formula: offers accepted ÷ offers made × 100.

Benchmark: Gem’s 2026 benchmarks report put offer acceptance at 82%, the highest since 2021, based on 1.2 million hires from June 2021 to May 2025 (Gem, vendor data). Experience plays a part: in Greenhouse’s 2024 candidate research, 20% of candidates said they’d turned down an offer because of a poor interview experience (Greenhouse, vendor-reported).

A related metric: some guides also track the interview-to-offer ratio (interviews held ÷ offers made). It’s mainly an efficiency measure, but a high ratio has an experience cost: more candidates sit through the full interview process only to be turned down. If it’s rising, look at whether your screening is finding the right candidates early enough.

12. Early attrition

What it is: the share of new hires who leave within their first 90 days.

Formula: new hires leaving within 90 days ÷ all new hires in that cohort. (Some guides divide by all leavers, which gives a different and misleading number.)

Benchmark: we couldn’t find a credible study that links candidate experience scores to 90-day retention. Early attrition still belongs in the set, because a process that oversells the job or the company culture, or skips the questions that matter for the role, shows up here. A bad experience can also cost you beyond hiring: in PeopleScout’s 2026 research, 36% of job seekers said they’d stop buying from a company after a bad candidate experience (TrueBlue, vendor-reported).

How to measure candidate experience: a 6-step plan

You don’t need all 12 metrics on day one. This plan gets a working measurement programme running in a quarter.

1. Start with three questions, not a dashboard

Where do we lose candidates? Why? And who? Each question maps to a layer: behaviour, sentiment and fairness. If a metric doesn’t help answer one of them, leave it out for now.

2. Pick a starter set of six metrics

1. Application completion rate
2. Time to outcome
3. Outcome coverage (share of candidates told the result)
4. Candidate NPS by outcome
5. Stage satisfaction at the screening stage
6. Pass-through rate by group at each automated step

Add the others once these are reliable.

3. Instrument the stages where volume is highest

Application and screening – the first stages of the recruitment process – hold most of your candidates and, in every dataset we’ve seen, the lowest sentiment. Make sure you can see step-by-step drop-off and completion by group at these stages. Many ATSs can’t show this on their own, which is where career-site analytics and screening-stage analytics earn their place.

4. Survey every outcome, not just interviewees

Send a short candidate feedback survey to rejected, withdrawn and hired candidates soon after each decision, and add a quick in-process rating at key steps. Keep cNPS and in-process satisfaction as separate metrics. To gather honest feedback, keep surveys short and anonymous by default, with an opt-in box so you can contact candidates who are happy to talk through their answers.

5. Segment, then set minimum sample sizes

Report by stage, outcome, role family, location and group. Agree a minimum number of responses before you report a segment – 30 to 50 survey responses is a sensible floor, and impact ratios need more. Show the sample size next to every number, and don’t rank recruiters on a handful of responses.

6. Take a baseline, change one thing, then check again

Run for a quarter to get a baseline. Then pick the lowest-scoring, highest-volume segment – so you allocate resources to the fix that affects the most candidates – and change one thing – faster decisions, feedback for rejected candidates, or a shorter, easier screening step. Compare that segment’s completion rate, time to outcome and sentiment with the baseline. Report behaviour, sentiment and fairness together on one page.

How often to measure: track behaviour continuously, report sentiment monthly or quarterly depending on your volume, and review fairness by role at least once a quarter and after any change to your process or tools.

For a deeper, one-off review of the whole journey – mapping each stage, testing it yourself and prioritising fixes – see our guide to a candidate experience audit [link once live].

Candidate experience benchmarks you can trust (and the ones to stop quoting)

Most candidate experience benchmarks come from vendors. That’s a reason to label them and compare like with like, not to ignore them.

Worth using, with labels:

• CandE Benchmark Research, 2025 – 66,000+ candidate surveys, January to August 2025; run by Survale, a survey vendor.
• Ashby’s 2026 benchmarks and Talent Trends – large, clearly dated datasets; customers lean towards tech.
• Appcast’s 2026 US and UK reports – strong on job ads, but click-to-apply, not completion.
• Gem’s 2026 benchmarks – pass-through and offer acceptance.
• SHRM’s 2025 benchmarking – a member survey; time to fill.

Stop quoting as current:

• the “365% uplift” and “60% abandon” application stats, both from 2016
• time-to-hire-by-industry tables built on 2017 data
• unsourced cNPS thresholds such as “+10 to +30 is typical” or “+50 is excellent”
• offer acceptance figures with no traceable source
• the Virgin Media “£4.4m lost revenue” story as a current cost – it’s a 2014 internal estimate

The rule of thumb: for most candidate experience metrics, there’s no independent, current benchmark. Compare like with like – same stage, same outcome, same region, same year – name the source and period, and treat your own trend over time as the benchmark that matters most.

Tools to measure candidate experience: a quick snapshot

Most teams need three things: an ATS or analytics tool for behaviour, a survey tool that reaches rejected candidates for sentiment, and visibility at the screening stage, where volume is highest. Here’s what the main tools cover. (Sapia.ai is ours.)

Where Sapia.ai fits: Sapia.ai tracks the screening stage from the inside. Discover Insights reports candidate sentiment, drop-off and representation across the hiring funnel, by role or business unit. Our chat interview is untimed, text-based and works on any phone, which removes much of the friction those metrics tend to reveal. At Woolworths, 82.6% of candidates completed it and rated it 9.2/10 for satisfaction (Woolworths case study, customer-reported). LNER kept ethnic-minority representation at a consistent 30% through every hiring stage (LNER case study, customer-reported). Scoring is based on candidates’ answers, not their CVs, and BLDS independently audited our US and Canada models in late 2022 (FAIR framework). What Sapia.ai isn’t: a whole-funnel measurement tool. Pair it with a survey tool such as Survale or Starred and your ATS analytics.

Three questions to ask any vendor:

• Does it reach candidates rejected at application and screening, not just those who interviewed?
• Who is in its benchmark, over what period, and is it run by the vendor?
• If it scores candidates, what is the score built from, and is there a recent, independent, role-level audit?

For a fuller comparison of tools that improve candidate experience, see our guide to candidate experience software.

Conclusion

A positive candidate experience isn’t one score. To measure candidate experience properly, track behaviour, sentiment and fairness together, by stage and by outcome. Label your benchmarks honestly, set minimum sample sizes, and treat your own trend as the benchmark that counts.

Then put your effort where the volume and the lowest scores are: application and screening. Use an ATS or analytics tool to see what candidates do, a survey tool to hear what they say, and screening-stage analytics to check the step most candidates never get past is fair and easy to complete.

If that screening step is where you’re losing candidates, see how Sapia.ai’s chat interview and Discover Insights work, or book a demo.

FAQs

About Author

Barb Hyman

CEO & Founder