What a 3% Rejection Rate Really Costs an Indian CNC Shop (₹)
In a mid-size Indian CNC shop supplying auto Tier-1s, a 3% rejection rate can easily cost about twice what the scrap bin suggests. Beyond the metal, it is machine-hours spent cutting parts that were never sold, setters and inspectors re-doing work, and the debit notes and PPM marks that come back from the customer. In our worked example a 3% rate costs about ₹2.1 lakh a month, close to 2% of sales.
By the Mama Editorial Team. For owners of turning and VMC shops in Manesar, Faridabad, Pune-Chakan, Hosur and Rajkot who track rejection as a percentage: we put it into rupees, show where on the floor it starts, and give a weekly routine. For the general case for cameras in quality control, see Catching Quality Problems on the Line with Cameras; this piece is about the money.
Key takeaways
- The scrap bin is less than half the bill. In the worked example below, material written off is about ₹1.03 lakh a month gross. The full cost of 3% rejection is about ₹2.13 lakh a month, or roughly ₹25 lakh a year.
- Machine-hours are the biggest hidden line. About 108 machine-hours a month go into parts that are scrapped, reworked or re-set. That is time you paid for at your machine hour rate and could not sell.
- Customer-side costs are contractual. Tenneco's published India supplier manual lets it charge sorting, rework and administration costs to the supplier, and counts a whole returned batch against your PPM unless you respond fast.
- Most rejection starts at a few predictable moments: the first piece after a setup, a cold machine at shift start, a worn or wrong insert, chips and coolant, and an in-process check that slips when the line is behind.
- Cameras show time, not dimensions. Video can show how long a machine waited for first-off approval or stood during a re-setup. It cannot tell you whether a bore is in tolerance.
The full cost stack of one rejected part
Most shops cost a rejection as "material plus a bit". The full stack has seven lines: material net of scrap value; machine-hours already spent before the defect was caught; rework machine and re-inspection time; re-setup when the cause was the set-up; in-house sorting of suspect WIP and stock; customer debits and premium freight; and rating damage, which costs nothing this month and can cost you the next programme. The first five happen inside your gate. The last two are governed by your customer's supplier quality manual.
What your customer's manual says about debits and PPM
Manuals differ, so read your own. Two published ones show the pattern.
Tenneco's India supplier requirements (Appendix C, last updated 30 June 2014) state a "0 PPM" philosophy and say Tenneco "will charge costs related to the nonconformance (sorting, rework, administration etc.) to the supplier". The counting rule matters more than the debit: if the plant returns a defective batch, 100% of the returned batch counts towards your PPM, adjustable to the real defective count if the supplier responds with good results in reasonable time (Tenneco India regional manual). A lot of 5,000 with 18 bad pieces can count as 5,000 rejects unless you respond fast. Speed is priced in: if you replace the whole defective batch within 24 hours, 1% of the replaced batch counts; after 24 hours, 10% of the returned quantity. Missing expectations "may lead to new-business hold".
Ramkrishna Forgings' Supplier Quality Manual (2018), from a Jamshedpur forging company with its own CNC machine shop, shows how fast the clock runs and what a poor rating leads to. On a nonconformance the supplier must contain immediately, reply in writing within 24 hours and report at the plant with an interim containment plan within 48 hours. Product made while a process was out of control must be sorted, scrapped, reworked or otherwise dispositioned. Suppliers are graded A to D on a composite of quality, delivery and price, and a supplier in group D for three consecutive quarters "should be de-registered" from the approved vendor list (RKFL SQM). That is the rating line in the cost stack.
Behind both sits IATF 16949. Its customer-satisfaction clause (9.1.2.1), as publicly summarised, asks certified suppliers to monitor indicators such as delivered part quality performance, customer disruptions, field returns and delivery schedule performance (16949store summary). Your customer tracks you the same way, and that PPM figure is what their buyer sees at the next sourcing meeting.
A worked ₹ example: a 60,000-piece machined flange
Every figure below is an assumption for illustration. Replace each one with your own numbers.
The shop and the part. A Haryana shop with about 30 CNC turning centres and VMCs on two shifts. One part is a forged steel flange for a Tier-1: turning, then drilling on a VMC, 4.5 machine-minutes per piece. Output is 60,000 pieces a month at ₹180, or ₹1.08 crore of sales. The forging costs ₹95; a scrapped piece fetches about ₹35 as steel scrap (1.1 kg at ₹32/kg), so net material lost is about ₹60.
Rates. Machine hour rate: ₹900/hour (assumed; work yours out as in our machine hour rate guide). Inspector: assumed ₹18,500 a month basic (illustrative — each state notifies its own minimum wages by trade and skill level and revises the DA part twice a year, so use your state's current notification). Divided by 26 working days and 8 hours, that is about ₹89 an hour; we add an assumed 30% for employer contributions, bonus and leave, so ₹116 an hour. Setter: assumed ₹35,000 a month, about ₹168 an hour, well above the floor.
Rejection. 3% = 1,800 pieces a month: 60% scrap (1,080), 40% reworkable (720). Some scrap fails after turning, some after drilling, so assume 3.5 machine-minutes already spent on each.
| Cost line | Working | ₹ per month |
|---|---|---|
| Net material on scrap | 1,080 × ₹59.80 | 64,584 |
| Machine-hours burned on scrap | 1,080 × 3.5 min = 63 h × ₹900 | 56,700 |
| Rework machine time | 720 × 3 min = 36 h × ₹900 | 32,400 |
| Re-inspection of reworked parts | 720 × 2 min = 24 h × ₹116 | 2,784 |
| Re-setups traced to defects | 12 a month × 45 min = 9 h × (₹900 + ₹168) | 9,612 |
| In-house sorting after a defect is found | 4 events × 2,000 pcs × 20 s ≈ 44 h × ₹116 | 5,156 |
| Customer debit (sorting, rework, admin) | one complaint every two months, ₹60,000 each | 30,000 |
| Your own sorting visit + premium freight | ₹23,000 per complaint, every two months | 11,500 |
| Total | ≈ 2,12,736 |
That is ₹2.13 lakh a month, about ₹25.5 lakh a year, or 2.0% of sales. It averages about ₹118 per rejected piece on a part that sells for ₹180. The gross material on scrapped pieces (1,080 × ₹95 = ₹1.03 lakh) is less than half of it.
If your machines are the bottleneck, use contribution instead. The 108 machine-hours (63 + 36 + 9) are roughly 1,440 pieces of capacity; at an assumed ₹70 contribution a piece, about ₹1 lakh a month you could not ship. Use that instead of the three ₹900-an-hour lines, not on top. The rating line has no number on purpose: a downgrade costs nothing until the next RFQ.
Where rejection actually starts on a CNC floor
Rejection on a machining line clusters at a few moments, each with a cheap fix.
The first piece after a setup. A new job means new jaws, offsets and perhaps a new program revision. IATF 16949 clause 8.5.1.3 expects job set-ups to be verified, with first-off/last-off part comparison where suitable (clause summary). The usual failure is pressure, not a missing procedure: the machine waits for QC, someone runs ahead "just a few pieces", the first-off fails, and those pieces are suspect too. When one setter covers several machines, the risk rises; see One Setter, Five Machines.
A cold machine at shift start. Okuma's guidance on spindle warm-up explains that a cold spindle and its components lengthen as they heat, and that a warm-up routine before production keeps size consistent from the first part to the last (Okuma). If rejection clusters in the first hour of a shift or after a long breakdown, suspect this first.
Tool wear and the wrong insert. A worn insert drifts the size before it breaks; a wrong grade or nose radius from a badly labelled crib can spoil finish from the first piece. Sandvik Coromant's turning troubleshooting guide traces a "hairy" surface to excessive notch wear, and poor finish to too high a feed combined with too small a nose radius (Sandvik Coromant). Fixed tool-change counts per edge and a labelled insert crib do more than another inspector.
Chips and coolant. The same Sandvik guide lists chips that break against the component and mark the finished surface as a cause of poor finish. A chip on the locating face of a fixture or chuck tilts the next part. Swarf clearing and a daily coolant check are cheap.
The in-process check that slips. The control plan says gauge one in twenty; when the line is behind, that quietly drops, and a drifting size runs for an hour instead of five minutes. It is a system problem (gauges far from the machine, too many machines per person, targets that reward count over checks), so fix layout and loading first.
What cameras can and cannot tell you
Limits first: a camera does not measure dimensions. It cannot read a bore, a thread or a surface finish, or tell which insert or program revision is loaded. Gauges, CMMs and dedicated vision rigs check parts. Video of machines and aisles shows time and flow:
| Question | Can video answer it? |
|---|---|
| How long each setup took, last good piece to first approved piece | Yes, if the machine is in view |
| How long a machine stood waiting for first-off approval from QC | Yes: idle spindle, part walked to the QC room, restart |
| How often parts went back to a machine for re-cutting | Often, if rework bins are kept in a fixed, visible place |
| How many person-hours went into sorting suspect WIP | Yes, as an aggregate count of people and hours |
| Whether a gauge check happened at a station | Sometimes, if the gauge table is in view; not reliable, and not a tool for policing individuals |
| Whether a part is in tolerance, or why it failed | No |
Keep the output per machine and per shift, never a ranking of operators, and give workers written notice; see the worker CCTV notice under DPDP. On old machines with no PLC, video is one of three ways to log stops (Tracking Old CNCs Without a PLC).
How an owner measures it weekly
One page every Monday, from rejection tags and the inspection register:
- Pieces made and rejected at each stage: in-process, final, customer.
- Scrap vs rework split for each part number.
- Top five reasons by count and rupees; every tag needs machine, shift, time and reason.
- Machine-hours lost to quality: scrap, rework and re-setup, times your machine hour rate.
- Customer PPM, their way: rejected ÷ received × 1,000,000, with their counting rules.
- Average minutes waiting for first-off approval per setup.
- One rupee total from the table above, against last week.
First hour of shift points to warm-up; after setups, to first-off discipline; spread across one machine's day, to its tooling and fixture. Keep the sheet with your IATF records, since an auditor will ask how you track quality performance (see cameras and IATF audits).
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 plain-language note on where machine-hours went yesterday, including machines waiting for first-off approval, re-setups and long restarts. It watches machines and flows, counts people only in aggregate, and never measures parts or ranks operators. Your gauges still decide what is good; Mama shows the hours rejection costs around them. Leave an email or phone in the form below, and Mama will get back to you with a view of where the hours go and a camera plan for any blind spots.
FAQ
How do I calculate the real cost of rejection in a CNC shop? Add net material, machine-hours already spent, rework, re-setup, in-house sorting, customer debits, and freight or travel to contain the problem. In our illustrative example that is about twice the scrapped material's value.
What PPM do Indian OEMs and Tier-1s expect from machining suppliers? Tenneco's published India manual states a "0 PPM" philosophy; the working target sits in your contract or scorecard. Check how your customer counts: some count a whole returned batch.
Can a camera detect dimensional defects on machined parts? No. That needs gauges, a CMM or a dedicated machine-vision station. Video shows the time around rejection: first-off waiting, re-setups, rework and sorting.
Is rework cheaper than scrap? Usually per piece, since you keep the material. But it still uses machine-hours and inspection, and on a full machine it displaces saleable output, so price it at lost contribution there.
What is a realistic first target for cutting rejection? Attack the biggest rupee reason in your weekly top five, not the overall percentage. First-off and shift-start problems are often cheapest, because the fix is a routine, not a machine.
