-
March 2024: The Tuesday Morning That Started With a Complaint
-
What Was Normal in 2019 Wasn't Normal in 2024
-
The Sandvik Coromant Official Site Detour (and the Logo I Should Have Trusted Sooner)
-
The Turn: First 50 Parts Fine, Part 180 Started Walking
-
Where the Laser Cutter Cost Question Came In
-
The Fix: Sandvik Coromant Inserts, a Second Supplier, and a Lot of Apologies
-
What I Got Wrong—and What I Changed
-
The Checklist I Use Now (and Its Limits)
March 2024: The Tuesday Morning That Started With a Complaint
I've been handling machining orders for a 12-person shop for nine years. In that time I've made four major purchasing mistakes I'd rather forget, totaling roughly $14,000 in wasted budget. This is the story of the third one—the one that finally got me to build a proper supplier checklist.
It started on a Tuesday morning in March 2024. A medical device customer sent back photos of a batch of POM (acetal) turning parts. The surface had chatter marks and a visible taper on the OD. We'd shipped 500 pieces the week before. Our QC report said everything was within tolerance. The customer's incoming inspection said otherwise.
I pulled the job file. Part tolerance was ±0.02 mm, surface finish Ra 0.4. We'd quoted it as a routine job. That was mistake number one.
What Was Normal in 2019 Wasn't Normal in 2024
Here's the thing about our shop: we do prototype and low-volume work. For years, our purchasing logic was simple—get three quotes, pick the lowest one that could hit the drawing, and use whatever carbide insert was on the shelf. In 2019, most of our plastic work ran at ±0.1 mm and Ra 1.6. Cheap inserts and low-cost turning suppliers were fine. The margins were thin but the work shipped.
Then customers started tightening tolerances. By early 2024, ±0.02 mm on plastic wasn't exotic anymore. It was the new normal. I knew that intellectually. I didn't act like it.
For this job, I shortlisted five cnc plastic turning parts suppliers. Three were domestic shops we'd used before. One was a broker we'd never tried. The fifth was a referral from a colleague who kept mentioning Makera CNC machining China as a low-cost option for simple turned parts.
I did the math. The broker quote came in 23% below our usual supplier. Makera CNC machining China was even lower, but lead time was six weeks. We needed parts in three. I went with the broker.
The Sandvik Coromant Official Site Detour (and the Logo I Should Have Trusted Sooner)
Meanwhile, I was trying to solve a separate problem: the insert grade. Our shop's tooling cabinet had a mix of brands. I spent two evenings on the Sandvik Coromant official site comparing recommended cutting data for thermoplastics. I even zoomed in on the Sandvik Coromant logo on a reseller's product page to make sure we weren't buying gray-market stock—the real logo has that subtle gradient, and the counterfeit versions usually get it wrong.
I ordered the correct Sandvik Coromant inserts for the finishing pass. They arrived in two days.
Then I made the decision that cost us $2,800: I let the broker run the job using their own tooling, because they quoted lower and said they had "equivalent" inserts. Equivalent on paper. Not equivalent in the cut.
The Turn: First 50 Parts Fine, Part 180 Started Walking
The broker shipped 500 pieces in four days. Their first-article report looked clean. Our incoming inspection passed the first 50 parts. Then our machinist noticed the OD on one part was 0.015 mm larger than the previous one. He flagged it. We pulled 30 parts and measured them.
They were all over the place. Not dramatically—0.01 to 0.025 mm of variation—but enough to fail the customer's incoming inspection. The surface finish on about 15% of the batch had chatter.
I called the broker. Their response: "We ran them on our standard insert grade. Should be fine."
It wasn't fine. They'd used a general-purpose insert with the wrong rake angle and edge hone for POM finishing. The insert wore unevenly after about 150 parts, and the taper started. The broker's QC only checked every 50th piece, so they never caught the drift.
Where the Laser Cutter Cost Question Came In
While we were sorting out the tolerance issue, the customer added a requirement: laser-etched logos on each part, 0.3 mm depth. We didn't have a laser marker in-house. So I spent a day trying to answer the question "how much does laser cutter cost?" for a 20W fiber marker.
Quotes we received in April 2024 ranged from $3,500 for a no-name import to $12,000 for a name-brand unit with a rotary axis. The salesman at one distributor tried to upsell me to a 30W unit because "you'll grow into it."
We ended up outsourcing the etching to a local shop for $1.20 per part. That was the right call for this job. But the research time cost me two days of production oversight—time I should have spent checking the broker's tooling list.
The Fix: Sandvik Coromant Inserts, a Second Supplier, and a Lot of Apologies
We re-ran 180 parts. This time we used the Sandvik Coromant inserts from our own tool crib and ran the job on a lathe we controlled. The taper disappeared. Surface finish held at Ra 0.35-0.4. We air-freighted the replacement batch to the customer on a Thursday. They received it Monday.
The final damage: $1,100 in air freight, $900 in rework labor, $800 in scrapped parts. Total $2,800. Plus a painfully honest phone call with the customer's engineering manager. We kept the account. Barely.
What I Got Wrong—and What I Changed
Everything I'd read about sourcing said to get multiple quotes and pick the lowest one that could hit the spec. In practice, for precision plastic turning, the cheapest quote turned out to be the most expensive decision. Not because the broker was dishonest—they just didn't have the tooling discipline for ±0.02 mm work on POM.
The conventional wisdom is that "equivalent" insert grades are good enough for plastics. My experience with this job suggests otherwise. The difference between a general-purpose insert and a grade designed for low-melting-point thermoplastics showed up after 150 parts, not on the first article.
I still kick myself for not asking the broker for a tooling list before placing the order. If I'd requested the insert manufacturer, grade, and recommended cutting data, I would have caught the mismatch in five minutes.
So glad I checked the Sandvik Coromant official site first. That detour gave me the baseline cutting data I needed to diagnose the problem fast. If I hadn't known what the correct parameters looked like, we'd have wasted another week blaming the machine.
The Checklist I Use Now (and Its Limits)
After the third rejection in Q1 2024, I created a pre-check list for any plastic turning job tighter than ±0.05 mm. It's not complicated:
- Ask the supplier for the exact insert grade and manufacturer before the job runs. "We use good inserts" is not an answer.
- Request the first-article inspection data alongside a 30-piece capability run. Check for drift, not just pass/fail on one piece.
- Compare the quoted cutting data against the tooling maker's public recommendations—Sandvik Coromant, Iscar, whoever. If the supplier has no data, that's a red flag.
- For secondary operations like laser marking, get a per-part outsourcing quote before considering equipment purchase. Unless you're running thousands of parts, buying a laser marker rarely pencils out.
I can only speak to our context. We're a 12-person shop doing 200 to 2,000-piece runs, mostly prototype and bridge production. If you're running 50,000+ pieces a month, you'd negotiate directly with the tooling maker and probably have an in-house laser. Your calculus will be different.
My experience is based on maybe 35 supplier evaluations over nine years. If you're working with exotic polymers, five-axis work, or medical-grade traceability, my checklist won't be enough. But the principle holds: what was best practice in 2020—get three quotes and go—is not best practice in 2025. The tolerances changed. The tooling data got better. The sourcing process has to keep up.
The $2,800 lesson wasn't about finding the cheapest supplier. It was about realizing that the cheapest supplier is only cheap if you know exactly what tooling they're using and how it behaves at part 151. I didn't know. Now I ask.