Quality metrics: 12 manufacturing quality metrics with formulas, worked numbers and targets

What are quality metrics?

A quality metric is a number, worked out the same way every time, that tells you whether a process makes good product and whether the quality system is doing its job. First pass yield is a quality metric: of the units that went into a step, the share that came out right without rework. So is customer PPM: rejected parts for every million shipped.

Quality measures, quality indicators and quality KPIs are other names for the same idea. The difference is use, not maths. Any count you track is a metric. It becomes a KPI when the plant picks it as one of the few numbers it manages by, gives it a target and an owner, and reviews it on a fixed rhythm. A plant might track 20 quality metrics and run the business on four of them.

Types of quality metrics

Sort them by where in the flow they look. A short set with one or two from each group covers a part from the supplier's dock to the customer's line.

One worked month: a valve body line

Every metric below is worked on the same illustrative month, not data from a real plant. A cell machines aluminium castings into valve bodies, washes and leak tests them, then assembles and tests them before they ship. Each body has five checked characteristics: the bore, the thread, face flatness, the leak test and the surface. One part number, one customer, one casting supplier.

Other counts for the month, used further down: 750 defects were logged on the 696 units that failed a check (some had two). The customer rejected 4 bodies and raised 2 complaints. The foundry delivered 10,200 castings, of which 51 were rejected for porosity. The plant sells each body for $50. Six CAPAs closed during the month, and the area had its internal audit.

Scrap rate and rework rate

Scrap rate is the share of units made that are thrown away. Rework rate is the share that needed repair, re-machining or retest before they were good. Count them apart: scrap loses the material and all the work already in the unit, rework loses time and capacity, and they usually have different causes.

• Scrap rate = units scrapped ÷ units made. Month: 200 ÷ 10,000 = 2.00%.
• Rework rate = units reworked ÷ units made. Month: 496 ÷ 10,000 = 4.96%.
• Scrap cost = units scrapped × what each had cost by the step where it was found. Month: 100 × $12 at machining + 50 × $16 at leak test + 50 × $24 at final test = $3,200, an average of $16 a unit.

Check the two rates against the line table: 2.00% + 4.96% = 6.96% of units started failed a check, and 200 + 496 = 696 units, which is 400 + 99 + 197, the failures at each step.

Setting the target

Set the scrap target in money as well as percent, because a scrapped body at final test costs twice one scrapped at machining. In the scrap rate calculator, 10,000 units, 200 scrapped and $16 a unit give 2.00% and $3,200 a month; a 1.0% target saves 100 units, $1,600 a month or $19,200 a year. Track rework in hours as well as units, since a two-minute retest and a two-hour strip-down count the same in the rate. Rank both by reason code on a Pareto chart; the scrap and rework guide works one month of reason codes in detail.

Customer PPM and complaint rate

Customer PPM counts the defective parts that got out: parts the customer rejected for every million you shipped. It is the number most customers judge a supplier on, and many write it into their supplier quality manual. The complaint rate counts complaints, which are events: one complaint can cover four rejected parts, and a wrong label is a complaint with no defective part at all.

• Customer PPM = parts rejected by customers ÷ parts shipped × 1,000,000. Month: 4 ÷ 9,800 × 1,000,000 = 408.
• Complaint rate = complaints ÷ units shipped × 10,000. Month: 2 ÷ 9,800 × 10,000 = 2.0 per 10,000 units.

Check PPM the other way: 408 parts per million × 9,800 parts ÷ 1,000,000 = 4.0 parts. Normalise the complaint count when volume swings, or a busy month looks worse than a quiet one with the same performance.

Setting the target

Start from the customer's requirement, then set your internal target tighter so a bad week does not breach the contract. With small counts, one reject moves PPM a lot: at 9,800 parts a month, each reject is about 102 PPM. Show a rolling three or twelve month figure beside the month so the team reacts to the trend, and treat every single customer reject as an event that gets containment the same day. Every complaint should open a non-conformance report; repeat or serious ones open a CAPA.

Cost of poor quality (COPQ) and cost of quality

Cost of poor quality is what the plant spends because work was not right the first time: scrap, rework, retest, sorting, returns and warranty. Cost of quality adds what the plant spends to stop that from happening (prevention) and to check for it (appraisal). The four categories, prevention, appraisal, internal failure and external failure, come from Armand Feigenbaum's book Total Quality Control and are often called the PAF model.

Some plants count appraisal inside COPQ, because inspection exists only because defects do. That definition gives 7,200 + 6,300 = $13,500 here. Either is fine; say which one you use and keep it. The hidden costs, such as overtime to make up for scrap, expediting and the capacity taken by rework, are real but rarely in the ledger, so the number on the report is a floor.

Setting the target

Set a target for failure cost, not for total cost of quality. Spending more on prevention is how the failure costs come down, so a rising prevention line can be good news. Review COPQ monthly as a share of revenue and as a Pareto of defects by cost. Enter the four categories in the COPQ calculator, or collect them over a month on the COPQ worksheet.

Audit findings and audit compliance

Audit metrics show whether the system the plant wrote down is the one it runs. Count findings by grade, then measure how they are closed. The raw number of findings is a poor target on its own: a sharp auditor finds more, and a team told to cut findings learns to hide them.

• Findings by grade: major, minor and observations per audit. Month: 1 major, 4 minor.
• Closed on time = findings closed by their due date ÷ findings due. Month: 4 of 5 = 80%.
• Repeat rate = findings that also appeared in the last audit of the area ÷ findings. Month: 1 of 5 = 20%.
• Layered process audit compliance = checks passed ÷ checks made. Month: 120 audits × 8 questions = 960 checks, 24 not OK, so 936 ÷ 960 = 97.5%.

Setting the target

Target zero repeat findings and a high on-time closure rate, and leave the count of findings untargeted. For layered process audits, watch which questions fail most rather than the overall percentage, because a high score on easy questions can hide one control that fails every week. See the layered process audit guide and the ISO 9001 internal audit checklist.

How to choose quality metrics, set targets and review them

1. 1

   Pick one or two from each group

   In-process, defect rate, customer, supplier, cost and system. Six to eight metrics is enough for a plant review; a line needs two or three.

2. 2

   Write the definition down

   What counts as a unit, a defect, an opportunity, a reject, a complaint. Which costs go in COPQ. A metric that changes definition between months cannot be trended.

3. 3

   Take a baseline

   Three to six months of the same counts, so the target starts from what the process does, not from a hope.

4. 4

   Set the target from three inputs

   The customer's requirement, the best month the process has already run, and the size of the gap in money. Pick the one that matters most; do not average them.

5. 5

   Review at the level that can act

   The line looks at FPY and defects by shift, the area at scrap, rework and CAPA weekly, the plant at PPM, COPQ and audits monthly.

6. 6

   Turn a red number into an action

   Every metric off target gets an owner and a next step at the review, or the review is only reporting.

Leading and lagging metrics belong together. Customer PPM, complaints and COPQ are lagging: they report a loss that has already happened. Layered audit compliance, Cpk, CAPA on-time closure and prevention spend are leading: they move before the lagging numbers do. A board with only lagging metrics tells the team they lost, not what to do next.

If you work to ISO 9001:2015, clause 6.2 asks for quality objectives that can be measured, and clause 9.1.3 asks you to analyse and evaluate data, including customer satisfaction and the performance of external providers. The metrics above give you that evidence; the standard does not name which ones to use.

Tracking quality metrics on paper, in a spreadsheet or in software

Most of these metrics come from three records: a defect log at the line, the shipping and receiving counts, and the CAPA and audit logs. A defect check sheet at each step gives you FPY, rework, scrap and DPMO. The manufacturing KPI dashboard template holds the monthly view. Spreadsheets break down when the defect log, the cost data and the complaints live in different files and each number is copied in by hand.

In LeanSuite, operators log each defect at the line by part number and failure code. The QA Matrix applies the standard costs you set by part number, ranks defects by cost, and shows in-process scrap next to customer returns and warranty claims on one dashboard; the costliest defects can start a Kaizen or CAPA project. Lean Studio runs tiered meetings on live, cascaded SQDCM boards, where a red metric becomes a tracked action, and KPI dashboards show the same numbers to every level. See cost of poor quality for how the defect costs are built.