QMS and CAPA KPIs: Benchmarks for Cycle Time, Closure, Calibration and Audits

Use these benchmarks as reference points for spotting a gap in your quality system, not as targets to hit. Your own trend, measured the same way every week, says more about your system than any published range.

Every figure shows its basis. A superscript number links to a public source in the Sources list; here those sources are registrar and customer deadlines for corrective action, not surveys of plant performance. A figure marked no cited source is kept exactly as originally published because it has not been matched to a public source that states it, so treat it as illustrative.

At a glance

CAPA cycle time

Corrective and preventive action (CAPA) cycle time is the headline KPI, and the benchmark spread is wide. Typical manufacturers run a mean closure of 45 to 90 days, and competent quality systems land at 30 to 45 days no cited source. World-class operations close a median CAPA in 20 to 30 days, with a hard cap of 60 for complex investigations no cited source.

Track the median alongside the mean, because a healthy system shows the two within 10 days of each other no cited source. A wide gap means a few stalled CAPAs are dragging the average.

For example, if your mean is 55 days and your median is 25, you have a tail of stalled CAPAs.

The lever is a staged deadline: 3 days to containment, 15 to root cause, and 45 to verified effectiveness no cited source.

Registrars set an outer bound. Under TUV Rheinland of North America's AS9100 certification conditions, a certified organization has up to 60 days from the audit date to submit the root cause and proposed corrective action for each nonconformity, major or minor¹. Other registrars and schemes set their own deadlines.

Customers set stages too. AM General's 2023 supplier quality manual asks for a written containment response within one business day², a final 8D response by the due date its supplier quality engineer sets, potentially as soon as 14 days³, and closure of non-priority corrective action requests within 120 days⁴. That is one manufacturer's requirement, not an industry average.

How to measure

• Count calendar days from CAPA opening to verified effectiveness.
• Report the median and the mean together every month.
• Split complex investigations from routine ones so one does not hide the other.

What moves it

• Staged deadlines for containment, root cause and verified effectiveness.
• Fast containment, so the investigation is not racing a live escape.
• Triage of low-severity issues into a simpler correction path.

Model how long each stage takes with the Corrective Action Cycle Time calculator.

On-time CAPA closure

On-time closure rate is the discipline metric behind cycle time. Typical shops close 60 to 75 percent of CAPAs by their committed date, and strong systems hit 85 to 90 percent no cited source. World-class holds 95 percent or better without extending deadlines to game the number no cited source.

The improvement lever is aging visibility. A weekly review of CAPAs past 75 percent of their allowed time, before they breach, cuts overdue counts faster than any policy memo no cited source.

Watch for closure padding, where teams pre-set 90 day deadlines to guarantee the metric. A padded deadline makes the rate look healthy while cycle time stays long.

How to measure

• Divide CAPAs closed on or before their due date by CAPAs due in the period.
• Count against the date committed at opening, not an extended one.
• Read it next to cycle time so padding shows up.

What moves it

• A weekly aging review of CAPAs nearing their due date.
• Realistic due dates set when the CAPA opens.
• A named owner for every open CAPA.

Check how many open CAPAs your team can carry with the CAPA Workload calculator.

CAPA reopen rate

CAPA volume and reopen rate reveal whether closures are real. World-class systems keep reopen rate under 5 percent, while typical runs 10 to 20 percent no cited source. At the top of that typical range, one in five root causes was wrong.

A rising reopen rate is the clearest signal that speed is being bought with shallow investigation. Balance it against workload: a single quality engineer sustainably owns 15 to 25 open CAPAs at once, and beyond 30, cycle time and reopen rate both climb no cited source.

The lever is not more deadlines; it is capacity. That means more headcount, or ruthless triage of low-severity issues into simpler correction paths.

How to measure

• Divide CAPAs reopened in the period by CAPAs closed in the period.
• Count a CAPA as reopened when the same failure recurs after closure.
• Track open CAPAs per quality engineer alongside it.

What moves it

• Investigation depth, confirmed by an effectiveness check before closure.
• Engineer workload held inside a sustainable range.
• Simpler correction paths for low-severity issues.

Run the open CAPA load per quality engineer with your own inputs. CAPA Workload

Calibration compliance

Calibration compliance is a compliance-floor KPI where the target is near perfect. World-class holds 99 to 100 percent of the gauge population in date, and typical sits at 92 to 97 percent no cited source. Anything under 90 percent is an audit finding waiting to happen no cited source.

The levers are recall lead time and buffer scheduling. Set calibration due dates with a 2 to 4 week grace buffer before the true drop-dead date, and auto-recall gauges at 30 days out no cited source.

For example, shrinking the overdue list from 18 to under 5 instruments usually takes one scheduling change, not more calibration spend.

How to measure

• Divide in-date instruments by total active instruments.
• Measure it monthly across the whole gauge population.
• List overdue instruments by name, not only as a rate.

What moves it

• A grace buffer between the scheduled due date and the true drop-dead date.
• Automatic recall notices sent well before the due date.

Price the finding that out-of-date gauges invite with the Audit Finding Cost calculator.

Training record completion

Completion of required, role-based training should hold at 95 percent or higher no cited source. World-class runs 98 to 100 percent current at any moment, while typical drifts to 85 to 92 percent as new hires and revised procedures create lag no cited source.

Track time to complete new assignments separately, with a target of under 30 days from procedure release to full crew sign-off no cited source. The lever is release-linked assignment: when a controlled document is revised, training tasks should generate automatically, so completion never depends on someone remembering to assign them.

How to measure

• Divide current completions by required assignments, per role.
• Track days from procedure release to full crew sign-off as a second measure.

What moves it

• Training tasks generated automatically when a controlled document is revised.
• Onboarding plans that schedule required training for every new hire.

Plan the audit preparation that checks these records with the Audit Preparation Workload calculator.

Audit findings per audit

Findings per audit benchmark the health of the whole system. A mature QMS averages 0 to 3 minor findings and zero majors per surveillance audit no cited source. Typical shows 4 to 8 minors and an occasional major, and a struggling system produces 10-plus findings including repeats no cited source.

The most damning sub-metric is repeat findings. World-class is zero repeats, because a repeat means a prior CAPA failed no cited source.

Track finding rate and severity mix rather than a raw count, since one major outweighs ten minors. Under TUV Rheinland of North America's AS9100 conditions, a major nonconformity at a surveillance audit leads to a re-audit, and the organization has up to six months from the audit date to implement the corrective action and have the re-audit conducted⁵.

Open findings also carry consequences beyond the audit. The same registrar will not transfer an existing certificate if closing its nonconformities takes more than 90 days⁶.

The lever is internal audit coverage: shops that internally audit every process at least annually surface issues before the registrar does no cited source.

How to measure

• Count findings per audit, split into majors and minors.
• Flag repeat findings separately.
• Weight findings by severity rather than summing a raw count.

What moves it

• Internal audits that cover every process on a set cycle.
• Effective CAPAs, so a finding does not come back.

Put a cost on each finding with the Audit Finding Cost calculator.

Cost of quality

Cost of quality as a percent of revenue is the executive-level benchmark that ties the KPIs together. World-class manufacturers run total cost of quality at 5 to 8 percent of revenue, with the failure share (internal plus external) under 2 percent no cited source. Typical operations sit at 15 to 20 percent, with failure costs dominating no cited source.

The shift from typical to world-class is a mix change, not a spend cut: it moves dollars from failure and appraisal into prevention. Every 1 dollar of well-targeted prevention typically returns 4 to 10 dollars in avoided failure over 12 to 24 months no cited source.

How to measure

• Sum prevention, appraisal, internal failure and external failure costs.
• Divide the total by revenue for the same period.
• Report the failure share on its own line.

What moves it

• Moving spend from failure and appraisal into prevention.
• Choosing prevention projects by their payback.

Test the payback of a quality investment with the QMS ROI calculator.

Run the cost of a nonconformance with your own inputs. Nonconformance Cost

How to build your own baseline

Prioritize improvement by leverage, not by whichever KPI is reddest. Rank your gaps by cost-weighted impact.

For example, a calibration score at 94 percent may be less urgent than a CAPA reopen rate at 18 percent that is spawning repeat audit findings and re-scrapped product.

• Write down each KPI's definition and hold it constant so the trend is real.
• Set one target per quarter and instrument it weekly.
• Pair each KPI with its published band, its current value and one named lever.
• Give every KPI a named owner.

A scorecard built that way will move numbers faster than a dashboard of twenty metrics nobody owns.