Tuesday afternoons are when incoming inspection reports show up on my desk. On that particular Tuesday in March 2024, the report wasn't from our lab—it was a customer rejection notice.
The customer had received a lot of 5,000 enclosures for a SICK photoelectric sensor series. Functionality was fine. Every unit passed our electrical and optical tests. But the anodized aluminum housing color was visibly off. They were right. Next to the original sample, the new batch had a greenish tint that I hadn't caught because I never looked.
SICK has a reputation for reliability in automation. A visible color defect on our sensor housing felt like a stain on that reputation. The customer had been using our sensors for years, and it was their brand that was on the line when our enclosures arrived on their assembly line.
I'm the quality compliance manager at SICK. I review every custom sensor variant before it reaches customers—roughly 200 unique SKUs each year (if you count variants). In 2024 I'd rejected 6% of first deliveries for dimensional issues, tolerance drift, or missing documentation. But color? That wasn't on my checklist.
It wasn't that we didn't care about appearance. The problem was that our incoming inspection protocol focused on things we could measure quickly: switch points, output signals, cable lengths, mounting dimensions. We had a reliable pass/fail line using our own SICK photoelectric sensors for presence detection, and a dial indicator for concentricity checks. Color was treated as a cosmetic item, something the naked eye would catch if it mattered.
When the rejection came in, our supplier quietly admitted they'd changed the paint supplier for the batch. They thought it was 'close enough.' I pulled out the Pantone fan deck, matched the reference, and measured Delta E. The number was 3.8. The spectrophotometer gave us exact numbers, but the fan deck was the tiebreaker. Under a 5000K light booth, the greenish tint was obvious. I photographed both samples and sent them to the supplier without a comment. The photo did the talking.
According to the Pantone Color Matching System guidelines, industry standard tolerance for brand-critical colors is Delta E < 2. A Delta E of 2–4 is noticeable to a trained observer; above 4 is visible to most people. We were at 3.8.
That's a level that most people wouldn't notice immediately, but once you see it, you can't unsee it. The customer's brand manager saw it. So we reworked all 5,000 enclosures—stripping, re-anodizing, re-testing. The total cost, including express shipping to meet the revised delivery date, was $22,000. The delay also pushed two other projects back by a week.
But the $22,000 wasn't the whole picture. We spent 45 minutes on a kickoff call to coordinate the rework. We consumed two production slots that were originally earmarked for other orders. And we had to tell two customers that their deliveries would slide. If we include lost goodwill and overtime labor, the real cost was closer to $30,000.
The supplier's first reaction was defensive. They pointed out that the color was within their own spec. But our spec said 'match the approved sample,' not 'match the numeric value.' That phrase mattered.
That was the expensive lesson. But it forced a real change. Over the next month, I rebuilt the incoming inspection protocol. The first addition was a color verification step using Delta E measurement. The second was a set of mechanical fixtures for things we'd been eyeballing.
On the runout fixture, we installed an encoder from the SICK DFS60 series—specifically, the encoder SICK DFS60 with 0.1° resolution. That gave us exact angular position for radial runout tests. It made the dial indicator readings significantly more repeatable. Before, we'd manually rotate the part and write down the highest number. Now we had an automated scan, and the data fed straight into our quality database.
Calibrating the dial indicator itself was another step. We checked it against a gauge block every Monday morning. It sounds obvious, but before the rework incident, we didn't have a scheduled calibration routine for manual tools.
For surface defects, we brought in a Pico microscope—a compact digital microscope small enough to keep on the inspection bench. It immediately showed micro-scratches on sealing surfaces that had previously passed as 'fine.' We also used it to verify edge break quality inside threaded holes, something the naked eye genuinely can't assess.
For electrical assemblies, I wanted to add thermal imaging. The first question from management was, how much is a FLIR thermal camera? The answer: a capable entry-level unit costs around $1,999–$2,900 (depending on model and resolution). We started with one mid-range model and a small training session for the inspection team.
The procurement team did a quick comparison between a handheld imager and a fixed-mount camera. For our bench, a handheld unit made more sense. We ordered one with a 160×120 thermal sensor—that's enough for board-level hotspots.
In the first month, the FLIR camera caught a power module running 15°C above normal on an otherwise functional unit. That one could have become a field failure. A field failure would have meant a recall, a warranty claim, and an upset customer—far more than the camera cost.
The surprise wasn't that the new process found defects. It was the kind of defects it found. I expected dimensional issues. Instead, the best catches were invisible to the old process: a hairline crack on a bearing seat, a cold solder joint, an overheated regulator. None of those would have shown up in a standard functional test.
We also started photographing every inspected unit that failed, so we could track recurring patterns. That simple archive helped us spot a fixture that was slightly misaligned—it had been causing scratches all along.
By Q2 2024, we had zero customer returns related to appearance or assembly quality. The earlier rejection is still in our monthly review as a case study, not as a statistic.
It took one $22,000 rework and about 150 shipment inspections to teach me that a checklist isn't just a list—it's a way to force your eyes where they don't naturally go.
I probably should have known this earlier. There were warning signs: a slightly different gloss on the previous batch, a photo in the supplier's email that looked off. I rationalized them away. The mindset that 'we've always done it this way' is exactly what causes these gaps.
I'd skipped a full visual check once before, almost a year earlier, on a smaller order of 800 units. Nothing came of it. That reinforced the idea that color was overrated. I thought, what are the odds? Well, the odds caught up with us.
That's the thing about near-misses: they don't feel like warnings. They feel like proof that the shortcut works. Until the day it doesn't.
Now every new inspection protocol we design includes the question: 'What could be wrong that we aren't measuring?' It's a humbling question. And it's cheaper than the alternative.
Five minutes of verification can save five days of correction. I've said that to every engineer who joins our team. This time, I mean it from experience.
