Quality metrics that matter are the small set of measures that connect defects to money and trigger action: first-pass yield, defective parts per million, escape rate, cost of poor quality, and on-time CAPA closure. Track those five. Ignore the vanity counts.
Most quality dashboards fail the same way: too many numbers, none of them tied to a decision. A screen with forty gauges tells an operator nothing, because nothing on it says "do something." A good quality scorecard is short. Every metric on it answers a question a manager actually asks, and every one has a line that, when crossed, starts a specific action. This post covers the five metrics that clear that bar, how they fit together, and how to build a scorecard people use instead of ignore.
What makes a quality metric worth tracking?
A quality metric is worth tracking when it changes a decision. That is the whole test. If the number goes red and nobody does anything different, the metric is decoration, and decoration costs you attention you cannot spare. Three properties separate the metrics that matter from the ones that fill space:
- It ties to money or the customer. Scrap pounds, rework hours, complaints, returns. If you cannot draw a line from the metric to a dollar or a customer, it will lose every fight for attention against metrics that can.
- It has a threshold and an owner. A number with no target is trivia. Every metric needs a line and a name: who reacts when it crosses.
- It is honest. The measurement has to be consistent shift to shift. A metric people can game (or that quietly depends on who is counting) rots the moment it starts driving decisions.
Notice what is missing: "it's easy to collect" and "the software already reports it." Those are reasons metrics end up on dashboards, not reasons they belong there. Start from the decision and work backward to the number.
Which five quality metrics belong on the scorecard?
These five cover the ground: how good the output is, how good the process is, whether defects are getting out, what it all costs, and whether you are actually fixing anything.
First-pass yield (FPY) is the percentage of units that pass through a defined process step correctly the first time, with no rework, repair, or scrap. A unit that fails and gets fixed still counts as a miss. FPY is the honest yield number, and it is a leading indicator: it moves before your defect rate and cost numbers do. When you chain several steps, rolled throughput yield multiplies the step yields together and exposes the "hidden factory" of rework that a final-inspection number hides. Our guide to first-pass yield walks the calculation.
Defective parts per million (DPPM) expresses your defect rate on a scale big enough to be useful. A 0.3% defect rate sounds fine until you write it as 3,000 DPPM and a customer who buys a million parts a year expects three hundred. DPPM is the language automotive and electronics customers use in scorecards, and it lets a small, good process and a large, sloppy one be compared on the same axis. See defective parts per million for the formula and the six-sigma reference points.
Escape rate is the share of defects that got past your inspections and reached the customer. It is the metric that grades your quality system rather than your process. You can have a mediocre defect rate and a great escape rate (you catch almost everything) or a decent defect rate and a terrible escape rate (your screens leak). Escape rate is usually tracked as customer-found defects divided by total defects, and it is the number that predicts complaints and returns.
Cost of poor quality (COPQ) converts all of the above into dollars: scrap, rework, sorting, warranty, returns, the lot. It is the metric that gets a finance director to care. Because COPQ in manufacturing commonly lands between 5% and 35% of sales, it is almost always a bigger number than anyone expects, and that shock is what funds prevention. Read cost of quality for the prevention-appraisal-failure model behind it.
On-time CAPA closure is the percentage of corrective and preventive actions closed by their committed date. It is the only one of the five that measures your reaction rather than your product, and it is the leading indicator for everything else. A plant whose CAPA closure is sliding is a plant whose DPPM and COPQ are about to climb, because the problems are being logged and not fixed. Pair it with CAPA discipline so the closures are real, not rubber-stamped.
| Metric | What it tells you | Type | Good direction |
|---|---|---|---|
| First-pass yield | How often the process gets it right the first time | Leading | Up toward 100% |
| DPPM | Defect rate on a per-million scale | Lagging | Down |
| Escape rate | Share of defects reaching the customer | Bridge | Down toward zero |
| Cost of poor quality | Total dollar cost of getting it wrong | Lagging | Down |
| On-time CAPA closure | Whether root causes are being fixed on schedule | Leading | Up toward 100% |
What is the difference between leading and lagging quality metrics?
Lagging metrics tell you what already happened; leading metrics tell you what is about to. DPPM and COPQ are lagging: by the time they move, the bad parts are made and the money is spent. FPY and on-time CAPA closure are leading: they move first, which gives you a window to act before the lagging numbers catch up. Escape rate sits between them, grading whether your containment held.
The mistake is building a scorecard entirely from lagging metrics, because they are the easy ones to pull from the ERP. A board of pure lagging numbers is a rear-view mirror; it tells you that you crashed, precisely, in dollars. Pair every lagging outcome with a leading process metric, and you can steer instead of just reporting the wreck. This is the same logic behind a balanced set of manufacturing KPIs: outcomes tell you where you are, drivers tell you where you are going.
How do you build a quality scorecard people actually use?
A scorecard succeeds or fails on discipline, not software. The five metrics are the easy part; the hard part is keeping the board short, the thresholds honest, and the review consistent. Here is the order that works.
- Start from the decisions, not the data. List the three or four questions leadership asks about quality every month. Pick the metric that answers each. If a candidate metric answers no live question, leave it off, even if it is easy to pull.
- Choose one metric per question, five or six total. Resist the urge to add "just one more." A scorecard with six metrics gets read; one with twenty gets glanced at. Depth beats breadth.
- Define each metric in writing. Numerator, denominator, process boundary, and who counts. "Yield" means five different things to five people; write down which one you mean so the number is comparable month to month.
- Set a threshold and an owner for each. A target line and a name. When FPY drops below the line, a specific person opens a specific action. A metric with no owner is a metric with no consequence.
- Split the board into leading and lagging. Put the leading metrics where the eye lands first. They are the ones you can still do something about.
- Attach one dollar figure. Translate at least one metric, usually the defect rate, into COPQ. The dollar number is what turns a quality problem into a business priority and unlocks the budget to fix it.
- Review on a fixed cadence and act. Same meeting, same day, every week or month. The review is where a red number becomes an action with an owner and a date. A scorecard nobody reviews is a spreadsheet nobody reads.
Set targets that stretch without lying. A first-pass-yield target of 100% is not a target, it is a wish, and everyone stops believing the board. Anchor targets to your own baseline and your customers' requirements, then ratchet them as you improve. Tie the whole scorecard up to your quality objectives so the numbers on the board are the same ones the quality policy commits to.
What are the data behind these metrics?
Reference points from the standards bodies
- A process operating at six sigma performance produces 3.4 defects per million opportunities, the benchmark DPPM figure the methodology is named for (ASQ, Six Sigma).
- The cost of poor quality is widely reported to run between roughly 5% and 35% of sales for many organizations, with a large share hidden in rework and lost capacity rather than obvious scrap (ASQ, Cost of Quality).
- The prevention-appraisal-failure framework behind COPQ, which shows why catching defects early is far cheaper than catching them late, traces to Juran and is maintained by ASQ as the standard model for quality costs (ASQ, Cost of Quality).
Which quality metrics should you not bother with?
Some numbers look like quality metrics and are really activity counts. "Inspections performed," "audits completed," "training hours logged" measure effort, not result, and effort metrics reward looking busy. They belong in an operations log, not on the quality scorecard. The same goes for any metric you cannot act on: knowing your defect rate to four decimal places changes nothing if you have no lever to move it.
Be equally wary of aggregate numbers that hide the problem. A plant-wide yield of 96% can be three good lines and one line quietly running 80%. Averages soothe; disaggregation acts. Track the metric where the action happens, at the line or the process, and roll up only for the executive view. When a number goes red, the fastest path from signal to fix usually runs through a red-bin analysis at the line or a documented quality audit not another slice of the dashboard.
Why do good metrics still fail to move the needle?
Usually because the data is trapped on paper. If quality checks, hold tags, and scrap reasons live on clipboards, nobody can assemble FPY, escape rate, and COPQ without days of transcription, so the scorecard is always weeks stale and half-guessed. Stale metrics do not drive decisions; they document history. The fix is not a fancier dashboard, it is capturing the quality event as structured data the moment it happens.
That capture problem is the one Harmony was built for: digitizing station-level checks and defect logging so yield, DPPM, escape rate, and scrap cost assemble themselves from the same records operators already create, no rip-and-replace. Pair the live floor data with a scrap rate the finance team trusts, and the scorecard becomes a report you pull in seconds rather than a project you dread. See how one plant runs it in the CLS case study. The metrics that matter are useless if they arrive too late to change the shift.
Pick the five. Give each a line, an owner, and a dollar figure. Review them on a fixed cadence and act on the reds. That is the entire discipline, and it beats a forty-gauge dashboard every day of the week.