What's the Difference Between "Nonconforming Product" and "Defect"? —— Why np Charts and c Charts Cannot Replace Each Other

1. The Trouble with a Batch: 1 Piece or 8 Defects?

In the inspection records of a coating workshop, there was such a line:

Sampling inspection of 10 pieces, 3 pieces have scratches, totaling 14 scratches and 9 pinholes, judged as 3 nonconforming pieces.

The quality manager was puzzled when reviewing the records: Should this batch be counted as "3 nonconformities" or "23 defects"? The difference between the two numbers is nearly 8 times, and the control charts would look completely different. Even more troublesome, the inspector recorded "nonconforming count" last month and "defect count" this month, breaking the trend chart.

The root of the problem is not in the recording habits, but in a widely accepted yet incorrect statement: "Nonconformity is a defect."

They count fundamentally different things: one is a "piece," the other is an "item." This is why both np charts (nonconforming count charts) and c charts (defect count charts) must coexist in SPC. This article will clarify this distinction and provide criteria for selecting the appropriate chart.

2. Counting Objects: Pieces and Items

In short: Nonconforming products are counted as "pieces," defects are counted as "items." A nonconforming product must contain at least one defect, but a product with defects may not necessarily be a nonconforming product.

3. Authoritative Definitions: Two Ways to Count on a Wafer

The NIST Engineering Statistics Handbook, when discussing attribute control charts, specifically clarifies this pair of concepts and emphasizes that "it may sound like nitpicking, but it is worth clarifying":

To distinguish between a defect and a defective, it is equivalent to distinguishing between a nonconforming item and a nonconforming product. Take a wafer, for example: the wafer is "a unit of the product," and each chip on the wafer is "a specific point." If a wafer does not meet at least one specification, it is judged as a nonconforming product; each chip that does not meet the specification becomes a defect. Therefore, a nonconforming product must contain at least one defect. However, it is important to note: a wafer may still be judged as conforming even if it contains multiple defects—for example, if the defects are in non-critical positions. Conversely, if the number of these "unimportant" defects becomes alarmingly high, it is necessary to investigate the production of this batch of wafers.

The most valuable part of this explanation is the last half: "A product with multiple defects but still judged as conforming" is a real state. This indicates:

• Focusing only on nonconforming count can completely hide a whole category of process degradation signals.
• Defect count is a leading indicator compared to nonconforming rate.

This point will be expanded in the next section.

4. Why Both Charts Are Necessary, Not One or the Other

① They Answer Different Questions

• np Chart / p Chart: How many pieces are problematic—directly corresponding to delivery quality, customer complaint rate, and batch rejection risk.
• c Chart / u Chart: How many items on average are problematic per piece—directly corresponding to rework time, internal losses, and repair costs.

The average rework time is determined by the average of 7 defects per nonconforming product, not by the number of nonconforming pieces.

② Defect Charts Are Leading Indicators, Nonconforming Rate Is a Lagging Indicator

When a process starts to deteriorate, it is usually the defects that increase first, before nonconforming products appear—defects initially fall in non-critical positions or are on the edge of acceptable standards, so the entire piece is still judged as conforming. By the time the nonconforming rate starts to rise, the problems have already accumulated to a significant level.

Therefore, a common scenario on the production line is: the c chart alarms first, while the np chart remains unchanged. This time difference is the most cost-effective improvement window.

③ Different Judgment Costs

• The np chart requires a full criteria judgment for each piece (size, appearance, and function must all be judged to determine "conforming or nonconforming"), which increases the inspection workload and amplifies the subjectivity of the judgment criteria.
• The c chart only requires "counting how many items," making it easier to execute for appearance-related defects (scratches, pitting, pores, pinholes, stains, poor solder joints) and providing better repeatability.

④ Different Distribution Models Lead to Different Control Limit Formulas

np Chart (Binomial distribution)   UCL = np̄ + 3√(np̄(1 − p̄))
c Chart (Poisson distribution)    UCL = c̄ + 3√c̄

5. Why They Look Like "the Same thing"

This is the systemic reason why this knowledge point has been muddled for a long time: When quality levels are good, the two numbers are indeed almost the same.

For the same average of 8, the control upper limits calculated by the two charts are:

The higher the quality level (the smaller p is), the closer the two values are—because when p is very small, the Poisson distribution is an excellent approximation of the binomial distribution. Therefore, on-site, it may seem like "there is no difference between the two."

However, this is a numerical coincidence, not a conceptual equivalence. To determine if they are the same, look at the counting object, not the calculated number.

6. The Same Batch of Data, Completely Different Numbers from Two Perspectives

Let's use a specific example to illustrate the difference:

Sampling inspection of 10 pieces, a total of 23 defects found (14 scratches + 9 pinholes), 3 pieces judged as nonconforming

np Chart (counting pieces)         np = 3        ← The upper limit is 10, cannot exceed
c Chart (counting items, inspection unit = entire batch)  c = 23       ← Can be much greater than 10
u Chart (counting items, inspection unit = per piece)  u = 2.30 items/piece

And you can get a number that can only be obtained by counting defects:

Average number of defects per nonconforming product = 23 ÷ 3 ≈ 7.7 items

This metric directly points to rework time and internal losses; it cannot be calculated using the nonconforming product count.

7. Deliverables: Two Questions to Determine the Chart

Selecting an attribute control chart requires answering only two questions:

Question ① Are you counting "pieces" or "items"?
        Pieces  →  p Chart or np Chart
        Items  →  c Chart or u Chart

Question ② Is the sample size / inspection unit fixed?
        Fixed  →  np Chart (counting pieces) or c Chart (counting items)
        Variable  →  p Chart (counting pieces) or u Chart (counting items)

The division of labor among the four charts:

Select based on characteristic type:

• Prioritize np / p: Dimensional deviations, functional failures, assembly errors, material mismatches—these are "judged piece by piece," and there is no such thing as "2.5 items of deviation on a piece."
• Prioritize c / u: Appearance-related (scratches, pitting, pores, pinholes, color differences, dirt), number of defects per point (defects per 100 solder joints), surface treatment (defects per m²)—these characteristics are naturally "counted item by item."

The most practical approach is to use both: Use defect count as a leading indicator to monitor process degradation, and nonconforming product count as a lagging indicator to monitor delivery risk. The best improvement opportunity is when the c chart moves but the np chart remains unchanged.

8. ⚠️ A Common Terminology Trap

The Chinese term "defect" has completely different meanings in two systems, which can lead to issues when writing documents:

In other words: "Defect" in the ISO system is much more serious than "defect" in the SPC system.

• In 8D, customer complaints, and legal documents, "defect" refers to issues at the level of "the product cannot be used as intended."
• In SPC control charts, "defect count" refers to "how many nonconformities."

A general rule of thumb: Any nonconformity constitutes a nonconformity, but not every nonconformity qualifies as a "defect." Choose your words based on the context, and don't mix SPC counting criteria into customer communication documents.

9. Three Common Pitfalls

① Using an np Chart to Manage Scratches Equals Flattening Information. 1 piece with 8 scratches is only worth "1" on an np chart, even though the process is already degrading, the chart remains flat. Characteristics that are naturally counted item by item should use a c chart or u chart.

② A Piece with Multiple Defects Should Not Be Counted as Multiple Nonconformities. When counting nonconforming products, a piece with multiple issues is counted as only 1 piece. Recording "23 defects" as "23 nonconformities" would immediately inflate the nonconforming rate to an impossible level (over 100%), rendering all subsequent judgments invalid.

③ The Inspection Unit for a c Chart Must Be Consistent. The criteria of per m² and per piece cannot be mixed on the same chart. If the inspection unit changes, convert to a u chart (unit defect count) before comparing, otherwise, the control limits themselves have no meaning.

The nonconforming product count answers "how many pieces are problematic," while the defect count answers "how many issues are on a piece." The former manages delivery risk, the latter manages rework costs; focusing on only one is like looking at a process with one eye closed.

Knowledge code: 6.3.1

Version: v20261017

Author: QTank QTank is dedicated to providing systematic professional knowledge, methodologies, and practical tools for quality management practitioners, helping enterprises continuously improve their quality capabilities.