Waiting for First-Piece Approval: The Inspector Queue That Idles Your CNCs (₹)

After every setup in an Indian machine shop, the CNC stands while the first piece walks to QC, waits its turn, gets measured and comes back signed. The check itself might take 10 to 15 minutes; the waiting around it can take three or four times that, because one inspector is serving many machines at once. The fix is rarely "work faster": it is to cut the inspector's load per setup, move the check closer to the machine, and stop setups from arriving in a bunch.

By the Mama Editorial Team. Written for owners of CNC turning and VMC shops supplying auto-component customers, in Chakan, Hosur, Manesar, Rajkot or Coimbatore, who see machines "under setup" for an hour when the setter finished 40 minutes ago.

Key takeaways

• First-piece approval (FPA) is required work; the queue in front of it is not. As an IATF consultant's summary of clause 8.5.1.3 puts it (the standard itself is a paid document), IATF 16949 asks you to verify job set-ups and retain records of approval after set-up. It does not ask the machine to wait 45 minutes for an inspector.
• Waiting grows much faster than inspector load. Queueing maths says the wait scales with utilisation divided by spare capacity, so an inspector who is 75% busy with first-offs makes each machine wait about three inspection-times, before walking and paperwork (Kingman's approximation, MIT course note).
• In our worked example (assumptions, not benchmarks), a 24-CNC shop doing 12 setups a shift loses about 728 machine-hours a month to FPA, worth roughly ₹5.8 lakh at ₹800 contribution per hour. Cutting the inspector's work per first-off and moving the check to the line brings it to about ₹1.3 lakh.
• Cheapest levers: a setup sheet with first-off check points so the setter pre-checks non-critical dimensions, a QC desk on the floor, staggered job ends, and on-machine probing where a machine has a probe.
• Cameras see the idle machine and the part on the table; only the QC record says whether it passed. You need both timestamps.

Why first-piece approval turns into a queue

On most Indian job-shop floors the sequence after a changeover looks the same. The setter (see One Setter, Five Machines) changes jaws, loads the program, sets offsets and runs one part. That part is carried to the quality room. The inspector, who is also doing patrol inspection, incoming material and customer paperwork, measures it with a height gauge, bore gauge, thread gauges and perhaps the CMM, fills the first-off report, signs, and the operator is told to start. Until then the spindle stands.

The measuring takes 10 to 15 minutes for a typical turned or milled auto part (an assumption; time yours). The rest of the time goes on four things:

1. Queueing. Three machines finish setups within 20 minutes of each other, usually just after shift start or lunch. The third part waits for the first two.
2. Walking and finding. The part travels to a room at the end of the bay, sometimes with the setter, who then waits too.
3. Fail and resubmit. If one dimension is out, the setter adjusts an offset, runs another piece and joins the back of the queue.
4. Thermal soak. A part fresh off the machine is warm. Industrial lengths have been defined at 20 °C since the CIPM adopted that reference temperature in 1931 (Doiron, NIST Journal of Research). Steel grows about 11.5 micrometres per metre per °C (the value specified for steel gauge blocks, per NIST), so a 100 mm steel part 15 °C above reference reads about 17 micrometres long. On a tight bore or length, the inspector has to let the part cool before the CMM check means anything. That cooling time is real, and it lands on the waiting machine.

The queue arithmetic owners never see

Queueing theory has a standard rule of thumb for a single server, Kingman's approximation (MIT course note): average waiting time ≈ (utilisation ÷ (1 − utilisation)) × (variability factor) × average service time. Treat the result as an estimate.

In plain terms, if first-offs arrive at random and inspection times vary about as much as their average (the variability factor is then about 1):

Illustrative, using Kingman's approximation with random arrivals and service. Bunched setups after shift start make real waits worse, not better.

An inspector who looks "only 75% busy" is already making machines wait three times as long as the check. So the strongest lever is inspector load per setup: take a few minutes off each first-off and the wait falls by far more than those minutes.

A worked ₹ example: 24 CNCs, two shifts

All inputs below are assumptions for illustration. Replace them with your own.

• 24 CNC lathes and VMCs making auto components, two 8-hour shifts, 26 working days.
• 12 setups per shift across the shop, so 624 first-offs a month.
• One inspector per shift with about 240 minutes a shift available for first-offs; the rest goes on patrol checks, incoming and records.
• Inspection takes 15 minutes per first-off; walking and paperwork add 10 minutes.
• Each machine-hour is worth ₹800 in contribution (sales value less material and consumables). Use your own costing.

At 12 first-offs × 15 minutes against 240 minutes, the inspector is 75% busy, so the queue adds about 45 minutes per first-off.

Queue minutes from Kingman's approximation with a variability factor of 1; rounded. Machine-hours = 624 × idle minutes ÷ 60.

Moving QC closer saves little on its own. Adding inspector capacity roughly halves the loss. Letting the setter pre-check non-critical dimensions, leaving QC only the critical and special characteristics, does the most, because it lowers both the inspection time and the load that drives the queue.

One honest caveat. Rupees from idle hours are real only if you could have sold those hours. If your machines are waiting for orders anyway, the cost of FPA waiting shows up instead as overtime on the night shift, late dispatches and a setter who can't move to the next machine. See Machine Hour Rate in India: Count the Hours You Never Sold for how to think about that.

Five ways to shorten the wait without weakening the check

1. A setup sheet with first-off check points. For every part and machine, one laminated page listing the fixture, tools, program revision and the five or six dimensions that matter, marked "setter checks" or "QC verifies". The setter then sends QC a part he has already measured, with his readings written on the tag. A Bosch Rexroth plant study found that introducing first-piece inspection charts and an inspection fixture cut inspection time by almost 50% (Bavuluri, Clemson University thesis, 2012). That was a US plant; treat it as direction, not a promise.

2. A QC desk at the line. A surface plate, height gauge, bore gauges and the part's gauge set between the machine rows; the CMM stays in the room for features that need it. It pays mainly as part of fix 1.

3. Tiered approval. In our reading of the clause summary, 8.5.1.3 asks for verification and records but does not name who must sign; the same summary lists "responsibilities for release after set-up" among the things a shop defines for itself. That is an interpretation, not an IATF ruling. Some shops let a trained, authorised setter approve repeat jobs on proven fixtures and keep QC sign-off for new parts, new revisions and special characteristics. Check your customer-specific requirements, control plan and certification body before you change who signs.

4. On-machine probing where you already have a probe. A UK subcontract machinist in a Renishaw case study sampled 1 in 10 parts on a CMM, a check that took "at least 30 minutes", and had to choose between stopping the machine and quarantining parts made meanwhile; moving the check onto the machine took two minutes per part (Renishaw). In Coimbatore, injection-moulded-parts maker Vasantha Tool Crafts reported part setting (a set-up step, not inspection) falling from one hour to three minutes after fitting a spindle probe and tool setter (Renishaw case study). Both are vendor-published, best-case figures.

5. Stagger job ends. Most of the queue comes from setups landing together. If the planner spreads batch ends across the shift, the same inspector serves the same number of first-offs with a fraction of the waiting. It costs nothing.

What not to do: let the machine "run a few pieces while QC checks". If the first-off fails, everything made since is suspect, which is exactly the trade-off the Renishaw case describes. See What a 3% Rejection Rate Really Costs an Indian CNC Shop for what that suspect stock costs.

How to measure it for two weeks

You need two timestamps per setup, and most shops record neither:

• Part off the machine (setter finished, first piece leaves the spindle).
• Approval received at the machine (operator allowed to run).

Add two columns to your first-off report: "received at QC" and "approved". With four times per setup you can split the gap into walk, queue, inspection and return. After two weeks, sort by hour of day. If most waiting sits in the first 90 minutes of a shift, the problem is bunching, not staffing (compare late start and early wash-up, where FPA queues often hide in the "late start"). For what happens after approval, see parts waiting between operations.

What cameras can and cannot show here

A camera covering the machine rows can show, without anyone filling a sheet:

• when each machine stopped cutting after a setup and when it started running again;
• a part being carried away and, if the QC desk is in frame, sitting on the inspection table;
• how many machines were standing at the same time, which is the queue itself;
• which hour of which shift the waiting piled up in.

It cannot show:

• Pass or fail. Video does not measure a bore or read a gauge. The approval and the readings live in the QC record, which IATF 8.5.1.3 already asks you to retain.
• Why it failed or which dimension. That is in the report, not the frame.
• What happened in an unwatched QC room. If the inspection room has no camera, video sees only "machine waiting".
• The reason behind every stop. A machine standing after a setup might be waiting for QC, a missing insert or the setter's next job. Pair video with the QC timestamps to tell them apart.

Keep the measurement about machines and queues, not individual inspectors or setters, and tell your workforce in writing that cameras are in use and why (worker CCTV notice under DPDP).

Where Mama fits

Mama reads the cameras a shop already has over RTSP/ONVIF, or a few added ones, and every morning sends the owner a short note on where machine-hours went the day before, including machines that stood after a setup waiting to restart and the hours when several stood together. It watches machines and flows and counts people only in aggregate; it does not measure parts or rank anyone, and your QC record still decides what is good. To see what that would look like on your floor: Leave an email or phone in the form below, and Mama will get back to you with a camera plan for any blind spots.

FAQ

How long should first-piece approval take in a CNC shop? There is no standard figure. In our assumptions the check itself takes 10 to 15 minutes for a typical turned or milled part (time yours); much beyond that, from part-off to approval, is usually queueing, walking or resubmission.

Does IATF 16949 require QC to approve the first piece? As commonly summarised, clause 8.5.1.3 requires verification of job set-ups, first-off/last-off validation where applicable, and retained records of approval after set-up. In our reading it does not name the approver; your own system and customer requirements decide that.

Should I hire another inspector or change the process? In our example a second inspector sharing first-offs roughly halved the loss, while setter pre-checks plus a floor QC desk cut it by about three-quarters. The two can be combined.

Why does the CMM check take so long after machining? Part of it is queueing, and part is temperature. Dimensions are defined at 20 °C, and a warm steel part reads long, so precise checks need the part to cool first.

Can a camera replace the first-off report? No. A camera can time the wait and show the part on the table, but the measured values, gauge used and signature must still be in the QC record.