Auto Parts Quality Checklist: 6 Steps from E36 Headlights to Welded Differentials

Who This Checklist Is For

If you're signing off on automotive parts—whether it's a container of aftermarket lighting, a run of sensor assemblies, or a batch of welded driveline components—this is the sequence I run before anything ships. Six steps, roughly two hours per SKU once you've done it a few times.

I'm the quality and brand compliance manager at an automotive parts company. I review every outgoing batch before it reaches a customer—around 240 SKUs a year across stamping, machining, and sourced lighting. In 2024 I rejected 19% of first articles. Not because the parts were junk. Because the paperwork, the terminology, or the inspection method didn't match what the customer actually needed.

Here's the order I do things in, and why.

Step 1 — Lock the Spec Sheet and the Vocabulary

Before anyone touches a part, the spec sheet has to be unambiguous. That sounds obvious. It isn't.

Take something as dumb as taillight or tail light. Both spellings are correct. Both show up in customer drawings, ERP descriptions, and inspection sheets. If your purchase order says "tail light" and your receiving system is keyed to "taillight," you've just created a duplicate SKU and a receiving hold that takes three people two days to unwind. I've watched that happen on a 4,000-unit order.

So the first step is a terminology pass: pick one spelling per part family, document it in the spec, and make sure the PO, the drawing, the inspection sheet, and the warehouse label all match. It's tedious. It prevents the dumbest category of rejection.

It's tempting to think specs are only about dimensions and tolerances. But half the disputes I've mediated in the last four years started with a term, not a measurement.

Step 2 — Inspect Exterior Parts Under Real Light

Exterior lighting is where cosmetic defects hide. We handle e36 headlights—specifically the e36 depo glass headlights, the glass-lens units that a lot of E36 owners swap in because the OEM plastic lenses haze over. DEPO is one of the brands we benchmark against, and the glass-lens version is a good example of why "looks fine" isn't an inspection.

What I check, in order:

• Lens clarity and mold marks. Hold it at a 45-degree angle to a window. Any flow lines, bubbles, or polishing swirls that show up in natural light will show up on the car.
• Reflector coating. Look into the bowl from the bulb seat. Uneven chrome, thin spots, or overspray on the cutoff shield are reject-worthy.
• Sealing. Run a finger around the lens-to-housing joint. Any gap in the butyl or silicone bead is a leak waiting for the first rain.
• Beam pattern. Put it on a wall at 25 feet with a known-good bulb. The cutoff should be sharp, with the correct step for left- or right-hand drive. A blurry cutoff means the reflector or the shield is off.
• Markings. ECE E-mark or DOT SAE marking, molded into the lens, legible. If the customer is in a regulated market, a missing mark is an automatic reject regardless of how the part looks.
• Connector pinout. Match the harness plug to the drawing. Pin 1 to pin 1. I've seen a batch where the ground and low-beam pins were swapped—physically it plugged in, electrically it was wrong.

The rookie mistake here—and I made it in my first year—is inspecting under warehouse fluorescents. Fluorescent light flattens everything. A lens that looks clean under a tube light can look scratched in daylight. We bought a $60 daylight inspection lamp after that lesson and never went back.

Step 3 — Test Electrical Parts as a Circuit, Not a Component

This is the step most people get wrong, and it's the one that causes the most comebacks. When you're dealing with a camshaft position sensor circuit, the sensor is only one part of the circuit. The circuit is the sensor, the harness, the connector, the shield, the ground path, and the ECU pin.

Here's what I mean. A cam sensor can bench-test perfectly—correct resistance, clean square wave on the scope—and still throw a P0340 code on the vehicle because the return circuit has 80 milliohms of resistance from a corroded ground, or the shield isn't terminated, or the connector pin has backed out halfway.

The old advice was "swap the sensor, clear the code, see if it comes back." That was reasonable fifteen years ago when sensors were the usual failure point and harnesses were overbuilt. Today, on a lot of platforms, the connector and the harness are the weak link, not the sensor. So the inspection has to cover the whole circuit:

• Sensor resistance and signal waveform against the OEM spec.
• Connector pin retention—does each terminal hold its position under a light pull?
• Harness continuity and resistance end to end, with a wiggle test while the meter is connected.
• Shield termination, if shielded.
• Ground path resistance from the sensor body to the battery negative.

If you only test the sensor, you're testing one-sixth of the thing that fails.

Step 4 — Verify Welded and Structural Parts at the Joint

Welding is where visual inspection gives you the least information per unit of effort. A bead can look beautiful and be cold-lapped underneath. Differential welding is a good case study. The job—how to weld differential gears and housings without inducing cracks—is highly stressed, and the consequences of a bad weld are expensive.

When I inspect a welded differential or any structural weld, I'm checking the process, not just the bead:

• Base metal identification. Differential housings and gears are often cast iron or a high-carbon alloy. You can't treat those like mild steel. Wrong filler or wrong preheat and you get cracking a week later.
• Preheat and interpass temperature. Recorded, not guessed. If the WPS calls for 400°F preheat, someone needs to have measured it.
• Filler match. The filler spec should be on the traveler, not in someone's head.
• Post-weld stress relief. For cast iron and high-carbon steels, this isn't optional if you want the part to survive thermal cycling.
• Dimensional check after welding. Welding pulls. Flange runout and bearing seat diameter need to be verified after the part cools, not before.
• Dye penetrant or magnetic particle on critical joints. Visual alone is not enough on a part that sees torque reversal.

The simplification fallacy here is thinking "good-looking weld equals good weld." It doesn't. The only reliable signal is process compliance plus non-destructive testing on the joints that matter.

Step 5 — Run a Physical Fit Test on a Small Batch

This is the step people skip when they're behind schedule.

A CMM report tells you the part is within tolerance. It doesn't tell you the part installs. I've seen parts that passed every dimensional check and still wouldn't seat because a bracket ear was 0.4mm off in a direction the drawing didn't measure, or because the plating thickness pushed a press-fit hole undersize.

So before full release, pull five pieces—not one—and install them on the actual mating assembly. Torque to spec. Check clearances. Check serviceability: can a technician get a socket on the fastener? Does the connector latch? Does the bulb seat fully?

Five pieces, because one piece tells you nothing about variation. If four out of five fit and one doesn't, you don't have a fit problem—you have a process capability problem, and it's cheaper to find that now than after 2,000 units ship.

Step 6 — Close the Loop with Total Cost

Once the part passes Steps 1 through 5, the purchasing conversation usually starts with unit price. That's the wrong number to lead with.

I calculate total cost before I sign off on any vendor comparison. Unit price plus freight plus duty plus incoming inspection labor plus the expected rework rate plus downtime cost if a batch gets held. On one lighting order, the $14 unit price came with $6 of freight and about $9 of inspection labor per piece, because the incoming defect rate was high enough that we went to 100% inspection. The $19 quote from a vendor with a clean first-article history was cheaper all-in. It wasn't close.

Time is a cost too. A three-week delay on a $40,000 production run isn't a $40,000 delay—it's whatever the line-down cost is for those three weeks, plus the expedite freight to catch up.

I'm not saying always pick the higher quote. I'm saying compare the right numbers.

Common Mistakes and Notes

• Inspecting only the sample, not the process. A good sample from a bad process is a coin flip on the next order.
• Trusting the drawing over the mating part. If the drawing and the actual vehicle/assembly disagree, the drawing wins on paper and loses in production. Flag it before you ship.
• Skipping the terminology pass because it feels like admin work. See Step 1.
• Treating certifications as a substitute for inspection. An IATF certificate tells you the system exists. It doesn't tell you this batch is right.
• Not recording the numbers. "Looked fine" is not an inspection record. Write down the measurement, the date, and who took it. As of January 2025, every customer audit I've sat through in the last two years asked for records, not opinions.

Run the six steps in order. The order matters—you can't inspect a part against a spec you haven't locked, and you can't calculate total cost until you know the defect rate.