What are the three types of in-process inspection?

First-piece inspection is a full check of the first parts after any setup change, to confirm the setup before the run is trusted. Patrol inspection is a roving check at set intervals during the run to catch drift like tool wear. Last-piece inspection checks the final parts when a lot ends to reveal accumulated tooling wear and bracket the lot with evidence.

What is the difference between in-process and final inspection?

Final inspection checks finished parts at the end of the line and sorts good from bad. In-process inspection checks parts during the run and steers the process before it goes out of tolerance. The difference is timing: a defect found in process exists in one or a few parts, while the same defect found at final has already been built into the whole lot.

How often should you do patrol inspection?

Set the interval from risk, not habit. Tie it to how fast the process can drift out of spec and how many parts you are willing to expose between checks. Fast-drifting operations may check every 25 to 50 parts or every 30 minutes; stable ones may check hourly. When the characteristic is on a control chart, let the chart's signals set the rhythm.

What is first-piece inspection?

First-piece inspection is a full check of the first part, or first few parts, produced after any change to the setup: shift start, changeover, new material lot, tool change, or program edit. It confirms the setup is correct before the run is trusted. Because a bad setup makes every part bad, it is the highest-leverage check in an in-process program.

Who performs in-process inspection?

Usually the operator running the machine performs first-piece and patrol checks at the station, because they can adjust the process immediately. Larger or higher-risk operations add a dedicated IPQC or roving inspector who patrols several lines, verifies operator checks, and handles first-piece sign-off. Either way, the reaction plan should not depend on who is on shift.

In-Process Inspection: Catching Defects Early

Q: What is in-process inspection?

In-process inspection, also called IPQC, is quality verification performed during production rather than before or after it. An operator or roving inspector checks parts against the spec at defined points in the run, records the result, and adjusts the process while it can still be corrected. Its goal is to keep the process from making defects, not just to find them at the end.

In-process inspection is quality checking done at the source, while a job is still running, so a process making bad parts is caught in minutes instead of at final inspection hours or days later. It has three standard moments: first-piece inspection right after setup, patrol inspection at intervals during the run, and last-piece inspection when the lot ends.

Final inspection is a net at the end of the line. In-process inspection is a hand on the process while it runs. The difference is not thoroughness; it is timing. A defect found at final has already been machined into a full lot, packed, and counted as good on the schedule. The same defect found by a first-piece check exists in exactly one part, and the fix is a tool offset, not a scrap ticket for the batch.

Plants that only inspect at the end are always inspecting the past. Plants that inspect in process are steering the present. This guide covers the three checkpoints, how to decide what to check and how often, and how in-process inspection fits with statistical process control and first article inspection.

What is in-process inspection?

In-process inspection (often written IPQC, in-process quality control) is verification performed during production rather than before it starts or after it finishes. An operator or a roving inspector checks parts against the spec at defined points in the run, records the result, and acts on it while the process can still be adjusted. The goal is not to find bad parts; it is to keep the process from making them.

That distinction matters because it changes what you do with a reading. Final inspection sorts good from bad. In-process inspection asks a different question: is the process still centered and stable, or is it drifting toward the edge of tolerance? A dimension measuring 6.37 mm against a 6.35 ±0.05 spec passes at final. In process, the same reading is a signal: the process has moved most of the way to the upper limit, and the next thermal cycle or tool-wear step may push it over. You correct now, not after the next hundred parts.

Final inspection tells you the lot is bad. In-process inspection tells you the process is drifting while only a handful of parts are affected.

Why check during production instead of only at the end?

Because the cost of a defect climbs every minute it goes undetected. A tool that wears past its offset does not make one bad part and stop; it makes bad parts continuously until someone measures one. If your only measurement is at final, the gap between "process went bad" and "we found out" equals the length of the run. On a long automated run that can be a full shift of scrap.

In-process inspection shrinks that gap to the patrol interval. It also moves the decision to where the fix lives. The operator standing at the machine can adjust an offset, change an insert, or re-zero a fixture in the time it takes a final inspector to fill out a rejection tag. And because the checks are recorded, they feed the same data stream that drives control charts and defect tracking turning routine measurements into an early-warning system instead of a filing exercise.

There is a maintenance angle too. Many in-process signals are really equipment signals: a dimension creeping in one direction is tool wear, a sudden step is a fixture that shifted, oscillation is a loose spindle. Catching these on the part often catches them on the machine before they become a nonconformance or a breakdown.

What are the three in-process inspection checkpoints?

Most in-process programs are built on three standard checks. Each answers a different question, and each catches a different failure mode.

First-piece inspection

First-piece inspection (sometimes first-off or first-article-off) is a full check of the first part, or the first few parts, produced after any change to the setup: shift start, changeover, new material lot, tool change, program edit, or a new operator. It verifies the setup is correct before the run is trusted. This is the single highest-leverage check in the whole program, because a bad setup makes every part bad, and first-piece is the only checkpoint positioned to catch it before the run begins. It is a cousin of first article inspection but lighter: first-piece confirms today's setup, while a formal FAI validates a new or changed process end to end.

Patrol inspection

Patrol inspection is a roving check of parts at set intervals during the run, done by the operator or a dedicated IPQC inspector walking the line. It catches drift that appears after a good setup: tool wear, thermal growth, material variation, gradual fixture loosening. The interval is a risk decision, covered below. Patrol checks are usually a subset of characteristics, the ones known to move, not the full first-piece list.

Last-piece inspection

Last-piece inspection checks the last part, or last few parts, when a lot or tooling run ends. Where quality depends on a mold, die, or perishable tool, the last piece reveals wear that accumulated across the run and confirms the whole lot stayed inside spec from first to last. It also creates a clean bookend: if first-piece passed and last-piece passed, the parts in between are bracketed by evidence. If last-piece has drifted, you know the drift happened during this run and you know which lot to quarantine.

First-piece guards the setup, patrol guards the run, last-piece brackets the lot. Skip any one and you leave a failure mode uncovered.

How to set up an in-process inspection plan in 7 steps

Rank the characteristics by risk. You cannot check everything every time. List the part's characteristics and flag the ones that are critical to fit or function, that have tight tolerances relative to process spread, or that have a history of drifting. Those get in-process attention; the stable, forgiving ones ride on first-piece and final.
Assign each characteristic to a checkpoint. Setup-sensitive features go on the first-piece list. Wear-sensitive and drift-prone features go on the patrol list. Tooling-life features go on the last-piece list. Some appear on more than one.
Set the patrol frequency from risk, not habit. Tie the interval to how fast the process can go bad and how many parts you are willing to expose. A fast-drifting operation checks every 30 minutes or every 25 parts; a stable one checks hourly. When a characteristic is on a control chart, let the chart set the rhythm.
Define the method and the tool. Specify exactly what to measure, with what gauge, and what "pass" means, including the reaction limit that triggers action before the spec limit. Confirm the gauge is capable with gage R&R so you are reacting to the process, not to measurement noise.
Write the reaction plan. The check is worthless without a documented what-to-do-when-it-fails: adjust the offset, quarantine back to the last good check, notify the lead, open a nonconformance. Put it in the work instructions at the station so the answer does not depend on who is on shift.
Record every result where it can be seen. Capture the reading, the time, the operator, and the disposition. Trend the data so a slow drift becomes visible before it becomes a reject. Paper on a clipboard technically satisfies an auditor and helps nobody at 2 a.m.
Review and re-tune the plan. Every escape that reached final or the customer should force a question: which in-process check should have caught this, and why did not it? Move intervals, add characteristics, or retire checks that never find anything. The plan is a living document, not a laminated relic.

How does in-process inspection relate to SPC?

They are two halves of the same idea. In-process inspection is the act of measuring during production; statistical process control is the method that decides whether a measurement means something. Plot your patrol readings on a control chart and the chart tells you the difference between normal variation you should ignore and a real signal you should act on. Without SPC, operators tend to over-adjust, chasing every reading and adding variation. With it, they leave a stable process alone and react only when the chart flags a shift or a trend.

The practical link is the reaction limit. A control chart's limits sit inside the spec limits, so the chart flags a shift while parts are still good. That is the whole point of in-process work: act on the process before the product goes out of tolerance, not after. For single-piece or low-volume work where subgroups are not possible, the individuals and moving range chart does the same job with one reading at a time.

In-process inspection numbers worth knowing

The economics behind checking early are well documented in the quality literature. A few anchor points from primary and standards sources:

The 1-10-100 rule of quality cost, long used in quality management and tied to the appraisal-versus-failure trade-off in ASQ's cost of quality guidance, holds that a defect costing roughly $1 to prevent or catch in process costs about $10 to fix at final and about $100 once it reaches the customer. The exact numbers vary; the order-of-magnitude escalation does not.
In-process reaction limits are set inside spec limits by design. Under the standard three-sigma logic described by ASQ's control chart guidance control limits sit at plus or minus three standard deviations of the process, so a stable process signals a shift before it produces out-of-spec parts.
The cost categories that in-process inspection targets, appraisal and internal-failure cost, are formalized in the cost of quality framework, where catching defects earlier shifts spend from expensive external failure toward cheaper prevention and appraisal.

Where in-process inspection fits with the rest of quality

In-process inspection is the middle of a chain. Upstream, incoming inspection keeps bad material from entering the process in the first place, so your first-piece check is not fighting a supplier problem. Alongside it, a formal first article inspection proves a new process is capable before routine in-process checks take over. Downstream, final inspection becomes a confirmation rather than a rescue, and defect tracking closes the loop by showing which checkpoints are earning their keep.

The friction is almost always the recording, not the measuring. Operators will happily mic a part every half hour; what they resist is filling a second clipboard while the machine waits. That is the gap Harmony was built for: capturing the check at the station, trending it automatically, and surfacing a drift before it becomes scrap, working on top of your existing gauges and systems with no rip-and-replace. See how that plays out on a real floor in our CLS case study or explore the capture and search features that turn routine readings into an early-warning signal instead of a stack of paper.