How Many Pieces Should First Article Inspection Take, and When Must It Be Redone? — Three Comparison Tables for Sample Size, Evaluation Criteria, and Trigger Conditions
In a certain stamping factory, after a mold change, the operator took the first piece and measured three external dimensions, all of which were within specification, and signed off. By the 260th piece, the patrol inspection found that the positional tolerance of the installation hole had reached the upper limit, and any further production would result in nonconformities. When tracing back, the first article record showed that all three measurements were within specification—this issue was not due to falsified records, but because the first article from the beginning lacked representativeness: the mold was a four-cavity mold, and he only took one piece from cavity 1; the critical positional tolerance was not even included in the first article record.
The failure of first article inspection is rarely due to the inspector's irresponsibility but mostly because three things were not clearly defined: when it must be done, how many pieces should be taken, and how the pieces should be evaluated. If these three things are not clearly defined, the first article inspection will degenerate into "signing off after the machine starts." The following three tables will help fix these issues.
1. First, Correct a Misnomer: First Article Is Not "the First Piece"
The first article refers to a small batch of pieces produced according to approved parameters after a change in production conditions, before batch release, not literally "the first piece."
This distinction is crucial. The first piece after machine adjustment is often in a state where the machine has not yet stabilized, the tooling has just been changed, and the mold temperature is not uniform. It could be the most deviated piece in the entire batch or, by chance, the best one. Using it to represent the entire batch is like deciding the fate of a steel furnace with a single coin toss.
Therefore, the "pieces" in the first article are plural, and its purpose is not to "prove that there is one conforming piece" but to prove that the conditions have been restored and are stable. Proving the existence of a conforming piece requires only one piece, but proving stability requires at least three pieces—three pieces to see the range, and five pieces to roughly assess the center and dispersion.
2. Table 1: When Must the First Article Be Redone?
This table should be posted on the machine's bulletin board rather than in the quality department's file folder. The criterion is: whenever an input affecting the process changes, the output must be re-verified.
| Trigger Condition | First Article Scope | Explanation |
|---|---|---|
| Mold change, model change | 1 piece per cavity, 1 piece per station | Multi-cavity molds are prone to missing cavities; cavity 1 being qualified does not mean cavity 4 is qualified |
| Tool change, grinding wheel dressing, electrode replacement | 3 consecutive pieces | The first cut with a new tool often has the largest dimensional deviation |
| Material batch change, supplier change | 2 consecutive pieces | Differences in hardness, viscosity, and thickness between batches will directly affect dimensions |
| Process parameter adjustment (temperature, pressure, speed, time) | 3 consecutive pieces | The first few pieces are still in a transitional state after parameter changes |
| Shift change, new employee working independently | 1 piece, double confirmation | Changes in human factors, focus on confirming consistent understanding |
| Machine downtime exceeding 2 hours, post-holiday startup | 3 consecutive pieces | Cold machine startup and mold temperature recovery can cause systematic deviations |
| Equipment maintenance, servicing, software upgrade | 3 consecutive pieces + stricter key characteristics | Maintenance may alter zero points, gaps, and compensation parameters |
| Design change, process change | Full size, 5 pieces | The impact of changes is unknown, treat as new pieces |
| Rework or repair of nonconforming products | Piece by piece | The status of reworked pieces is the most unpredictable |
| Customer or regulatory requirement | As per customer specifications | When customer requirements exceed internal standards, follow the customer's requirements |
The value of this table lies in transforming "judging based on experience whether to do it" into a "checklist." Experienced team leaders know to perform a first article after a mold change, but they often forget "software upgrades" and "post-holiday startups"—both of which have caused batch rejections more than once.
3. Table 2: How Many Pieces to Take Each Time
The sample size for the first article is not the same as that for sampling inspection. Sampling inspection aims to estimate the defect rate, and the sample size is determined by AQL and batch size; the first article aims to detect anomalies, and the sample size is determined by the structure of the process—how many cavities, stations, and key characteristics there are, and how many must be covered.
| Production Form | First Article Quantity | Evaluation Points |
|---|---|---|
| Single cavity, single station, process capability Cpk ≥ 1.33 | 3 consecutive pieces | Look at the range and trend, not just individual values |
| Multi-cavity mold, multi-station, multi-axis equipment | At least 1 piece per cavity and station, total no less than 5 pieces | Record each cavity separately, do not combine into one row |
| Safety characteristics, key characteristics (SC/CC) | 5 consecutive pieces | Measure key characteristics for the first and last pieces |
| New mold, new production line, new process first production | 5-10 consecutive pieces, full size report | This is essentially a FAI, not a regular first article |
| Manual operations, dependent on personnel skills | 5 consecutive pieces, 1 piece by a different person | Consistency among people is the hardest to ensure |
| Sensitive to temperature, pressure, viscosity | 5 consecutive pieces, 2 more pieces after 30 minutes | The last two pieces assess the steady state, not the startup state |
For an injection-molded part with an eight-cavity mold, the correct approach is to take 1 piece from each cavity, totaling 8 pieces, and record the critical dimensions for each cavity, rather than "taking 3 pieces all from near cavity 1." The differences between cavities are due to mold runner balance issues, and failing to sample from each cavity is equivalent to assuming there are no problems.
4. Table 3: How to Evaluate the Pieces Taken
The most common mistake in first article evaluation is "if all dimensions are within tolerance, release it." Tolerance is for parts, not for the process; a first article that just meets the tolerance limit is itself a risk signal.
Evaluation is divided into three layers, and all three layers must pass before release:
- First Layer: Individual Conformity. 100% of key characteristics must be measured, and general characteristics should be measured according to the work instruction. It is not allowed to "measure only a few main ones this time."
- Second Layer: Consistency Within the Group. For data from 3 or more pieces, there should be no monotonic increase or decrease (indicating tool wear or mold temperature rise has not stabilized); the range should not exceed 1.5 times the historical range for that characteristic.
- Third Layer: Reasonable Position. The first article values for key characteristics should fall within the middle 60% of the tolerance band or within the control limits of the control chart. Pieces that fall within the 20% edge bands of the tolerance limits should be evaluated as "conditionally released," with stricter inspection of the first 20 pieces (or the first 30 minutes) during the conditional release period. If any piece falls within the edge bands again during this period, the line should be stopped.
There are only three possible conclusions, not "basically okay":
- Release: All three layers pass, sign "first article qualified," and start the batch.
- Conditional Release: Individual characteristics are qualified but are on the edge or show a trend within the group; perform stricter inspection of the first 20 pieces (or the first 30 minutes), and stop the line if any piece falls on the edge during this period.
- Reject: Any characteristic exceeds the tolerance. At this point, the action is not just to "readjust the machine and redo the first article," but also to isolate and trace all products produced after the last qualified first article. This step is the easiest to skip and the most likely to turn into customer complaints.
5. Five Steps to Implementation
- List Trigger Conditions. Adjust the list in Table 1 according to the factory's processes, print and post it on each machine, and include a copy in the mold change work instruction.
- Specify the Number of Pieces in the Inspection Work Instruction. Do not write "appropriate" or "several pieces," but specify "1 piece per cavity, total 8 pieces, record each cavity separately."
- Define Measurement Projects and Instruments. The first article must measure key characteristics; the measuring instruments must meet the resolution and GR&R requirements. Do not use a caliper with a large range and 0.05 mm resolution to evaluate a 0.02 mm tolerance.
- Double Signature for Release. The operator performs a self-inspection and signs, and the inspector or team leader reviews and signs. Both signatures are required for release.
- Track First Article Values. Each time, plot the key characteristic values of the first article on the control chart or trend table. A single first article is just one point, but connecting dozens of first article points can reveal chronic issues such as mold change deviations, seasonal drifts, and cavity wear.
Returning to the stamping factory mentioned at the beginning: after taking pieces from each of the four cavities and including positional tolerance in the mandatory first article inspection items, the same mold change deviation was caught during the first article stage, reducing the number of rejects from 260 pieces to 4 pieces. The cost of the first article inspection is just a few minutes and a few sample pieces, but it prevents the entire batch from being scrapped.
The key to first article inspection is not in the "inspection" but in the "definition": defining which changes require a new first article, defining how many pieces to take from each cavity and station, and defining the three-layer evaluation and three-tier conclusions—by clearly defining these three things, the first article truly guards the starting point of batch production.
Knowledge code: 11.1.1
Version: v20261018
Author: QTank QTank is dedicated to providing systematic professional knowledge, methodologies, and practical tools for quality management practitioners, helping companies continuously improve their quality capabilities.