I Nearly Approved a Bad Batch: A Quality Inspector's TCO Lesson with Starrett Tools

I'm a quality manager at a precision machining company. I review first-article lots before they reach customers—roughly 200 unique items a year. In Q1 2024, I nearly gave a bad batch the green light. That week changed how I buy tools and how I think about supplier quotes.

The parts were 4,800 aluminum sensor housings from a new supplier. They looked perfect. The finish was clean, the edges were sharp, and the first four dimensions I checked were right on the print. The supplier's CMM report showed every point in tolerance. If I had relied only on that report, I would have signed off.

The setup

We had been buying similar housings from another shop for about two years, with occasional headaches. This new quote was 18% lower. The purchasing manager was pushing for a quick approval. Our contract demanded first-article inspection before any production order, and that's my job.

Over 4 years of reviewing first articles, I've learned to begin with the most likely failure mode. For this part, that was wall thickness and an internal radius. I did not start with the easiest dimension. I started with the one that would hurt the most if it was wrong.

Starrett 5 6 Micrometer

The critical wall is 5.400 inches plus two tenths. My Starrett 5 6 micrometer (the 5–6 inch range model) is the tool I trust for that measurement. It isn't fancy. The frame is stiff enough that if you squeeze too hard, you can feel the frame flex, which is a good reminder to use consistent force. I calibrated it against a known standard before the lot, and I checked it again after the first ten parts.

Starrett Optical Comparator

The internal slot radius was harder. The drawing called for a radius with a center location held to 0.002 inches. A CMM can hit one point, but a radius can be shifted and still show a point or two inside tolerance. That's why I used the Starrett optical comparator. It projects the whole profile at 10x onto a graduated screen. I matched the part against a calibrated overlay. The top of the radius lined up. The bottom did not. The radius center was displaced by almost 0.004 inches—twice the allowed tolerance.

At first I thought I had placed the part crooked on the comparator stage. I re-set it, re-focused, and got the same result. The numbers on the CMM report said accept. My gut said something was off. The comparator showed that my gut was right, but it took a tool that shows the full form, not just a list of coordinates, to prove it.

How to Use a Starrett Angle Finder

The chamfer next to that radius read exactly 45 degrees, so the angle wasn't the problem. In case you're wondering how to use a Starrett angle finder, it's simple: put the base on a clean surface, swing the blade until it touches the angle you want, lock the dial, and read the measurement. On the Starrett dial, you get both acute and obtuse readings directly, so there's no mental subtraction. The common mistake is zeroing on a dirty surface or trusting a worn blade. I wiped the reference surface first and checked against a known 45-degree block.

The 117 Multimeter and the Centrifuge 5804

I also ruled out electrical and material issues. Our maintenance tech ran a bonding check on a sample with the 117 multimeter we keep on the bench. The 117 multimeter reviews I'd read before we bought it were mostly positive, and it did its job—continuity and insulation were fine. I also checked the log for our Eppendorf Centrifuge 5804, which we use to spin down coolant samples when qualifying a new supplier's process. The 5804 was calibrated on schedule, so the material prep wasn't suspect. That left the radius center as the only remaining red flag.

The rejection

I rejected the first article. The supplier called the next day and said, 'It's within industry standard.' That phrase is a red flag to me. The drawing is the standard. 'Industry standard' is what people say when the numbers are on the edge and they hope you won't look at the whole form.

The purchasing manager pushed back. The low price was attractive, and the schedule was tight. If I sent the lot back, the supplier would have to redo the batch, and we would lose two weeks. I understood the math. But I had seen a similar quality issue cost us a $22,000 redo and a delayed launch a few years earlier. I didn't want a repeat.

I asked the supplier to send their CMM program and raw data. They didn't have a point pattern that covered the full radius. I'm not 100% sure whether that was intentional or just lazy, but it was incomplete. We held the line.

The supplier redid the parts at their cost. The second first-article passed, and the production order shipped about two weeks late. The total damage included extra expedited freight, overtime inspection, and a change fee from our customer. In the end, the low-cost supplier was not low-cost at all.

The TCO lesson

It took me years and hundreds of inspection decisions to understand that the purchase price of a measuring tool is the least interesting number on the spec sheet. The cheap calipers that were sitting in the tool crib would have saved the department maybe $400. But they wouldn't have caught a shifted radius. The Starrett optical comparator cost more, but it showed the whole truth in one screen. That is total cost of ownership.

Now I calculate TCO before comparing any vendor quote. Unit price is only the first line. I also include shipping, calibration, operator training, rework rate, and the risk cost if a bad part reaches a customer's line. Even the small stuff matters. According to USPS pricing effective January 2025, a First-Class Mail letter costs $0.73. That doesn't sound like a lot, but when you're mailing calibration certificates, sealed samples, and revision documentation, those line items add up across hundreds of parts.

I didn't invent this idea. I just learned it the hard way. In my early years, I watched a $500 quote turn into $800 after shipping, setup, and revision fees. A slightly more expensive all-inclusive quote would have been cheaper. I see that pattern over and over with supplier selection and with tool selection.

Per FTC guidelines, a claim like 'guaranteed 100 percent accurate' would need substantiation, and no honest manufacturer makes that promise without calibration caveats. That's why I trust spec sheets, calibration standards, and a repeatable process. The tool is only part of the system; the system is only as good as the person using it.

If you ask me, the purpose of precision measurement is not to find bad parts. It's to understand why a part could go bad before it costs you real money. The Starrett 5 6 micrometer, the optical comparator, the angle finder, the 117 multimeter, and even the centrifuge 5804 in our lab each have a place in that system. They're not luxury items. They're cheap insurance against the real cost of being wrong.