-
September 2022: the order that looked simple
- The three checks I talked myself out of
-
The delivery, and the afternoon it stopped being theoretical
-
The call I almost got wrong
-
What the redo actually cost
-
The protocol that came out of it
-
What to look for in a motion sensor supplier, if you're starting fresh
-
Where an established control line fits into this
-
The one line I'd keep
September 2022: the order that looked simple
I'm a quality and brand compliance manager at a lighting company. I review every fixture, control, and private-label SKU before it goes out to a customer — roughly 260 items a year. Since I put our current verification protocol in place in 2022, I've rejected about 18% of first production deliveries. Most of those rejections weren't for broken units. They were for spec drift.
The program that taught me the most was also the one that looked easiest. A regional distributor asked us to build a private-label residential downlight bundle: fixture, motion sensor, dimmer, one SKU, one box, 1,400 units. Nine weeks to a shelf reset. Simple enough that I did the whole review in an afternoon (that afternoon is the one I'd like back).
The three checks I talked myself out of
Here's how I justified skipping each one.
1. The finish color
The private-label light fixture manufacturer sent a sample in a matte white. The quote said "same finish as sample." I asked for a color measurement report. They sent a photo. I accepted the photo.
It's tempting to think that if two parts look the same in your hand, they're the same part. But appearance under a shop light and appearance on a ceiling are two different measurements, and "looks the same" isn't a spec. If you don't write a number down — ΔE, gloss unit, whatever — you don't have a requirement. You have a preference.
2. The dimmer's low end
We tested the dimmer on the bench with the driver we had in the lab. It dimmed smoothly to about 3% and then dropped out. Fine. Normal. We shipped the spec with "smooth dimming" language in it.
What we didn't do was test it with the driver the actual fixture shipped with. Those were two different drivers (same wattage, different manufacturer) and the low end behaved differently — visible banding at 15%, an audible buzz down around 8%.
3. The sensor's coverage
This is the one that still bothers me. We tested the sensor on a table, at roughly 3 feet, waving a hand. It triggered every single time. The datasheet said 180° detection with a 20-foot range, and our test seemed to confirm it.
But nobody walks into a room waving their hands. They walk in at floor level, from a doorway, with a wall eating half the pattern. And "20-foot range" on a sensor datasheet usually refers to a specific test condition — a specific target size, moving directly toward the unit, at a specific mounting height. Move the target to a tangential walking path and the number gets smaller. Sometimes a lot smaller.
We verified the sensor against the datasheet. We should have verified it against the room.
The delivery, and the afternoon it stopped being theoretical
Cartons arrived on a Tuesday. My team pulled 60 units across three pallets — a sample size we hadn't agreed on in advance, which is its own problem (unfortunately). By Thursday we had three findings.
The finish was off. Measured against the original sample, production units sat at a ΔE we would have rejected outright if we'd written a tolerance into the purchase order. The vendor's position was that it was "within industry standard." Maybe it was. It wasn't within our standard, because we never stated one.
The dimmer flickered with the production driver. Not always. Mostly at the low end, mostly on fixtures sharing a circuit with the sensor. Which, of course, was the entire bundle.
And the sensor worked — if you walked straight at it. Walking parallel to the wall, six feet out, at a 9-foot ceiling, it wouldn't trigger until you were almost underneath it. In a hallway, that's a complaint. In a stairwell, that's a conversation I don't want to have in a conference room.
The call I almost got wrong
The instinct was to ship a partial — 800 units that tested clean, hold the rest, hit the reset date. I ran it both ways on paper. Best case, we cover the reset and fix the balance in three weeks. Worst case, those 800 units had the same latent low-end issue at a different ambient temperature, and we'd be doing a field swap on installed product, which runs roughly ten times the cost of catching it on a dock.
The expected value said ship. The downside felt catastrophic. I held the order.
Then I second-guessed myself for the next nineteen days. What if I'd misread the flicker test? What if the finish tolerance I was now demanding was unreasonable for an injection-molded part? What if the distributor walked? I didn't relax until the second production run was in our hands and the measurements came back in tolerance on the first pass.
What the redo actually cost
The vendor redid the batch at their cost. That clause was in the contract then, and it's still in it now. What wasn't covered: the expedite fee on the replacement run, the air freight to hold the reset date, and my team's time re-inspecting 1,400 units instead of the 60 we'd planned.
All told, about $18,000 more than the original quote, on a program where the whole margin was somewhere around $40,000. And we still missed the reset by eleven days.
I've stopped calling that a vendor failure. It was a specification failure. We asked for a product and accepted a datasheet in its place.
The protocol that came out of it
I wrote this after the third production run. It's boring on purpose.
- Production-representative sample. Pulled from the actual run, with the carton and label from that run. Pilot-line samples and hand-built samples are marketing material.
- Sensor coverage tested at installed height, with tangential and radial walking paths, at the ambient light level of the real application. Hand-waving tells you the sensor is alive. It doesn't tell you it works.
- Time delay and sensitivity ranges verified across the full adjustment range, not just the factory default.
- Dimmer tested with the exact driver and load mix it will ship with, at 100%, 50%, 15%, 8%, and minimum — in both dimming directions, cold and warm.
- Compatibility documented in writing, with the tested driver list attached and an explicit note about what wasn't tested. A short compatibility list you can trust beats a long one you can't.
- Finish measured, not eyeballed. Spectrophotometer against the approved sample, tolerance written into the PO.
- Every "typical" number on the datasheet converted to a guaranteed number in the purchase order. Typical means nothing in a contract.
- Change control. Any BOM change — driver, lens, LED bin, housing resin — requires written notification before shipment, not after.
There are twelve points in the full document, but those eight are the ones that have actually caught something.
What to look for in a motion sensor supplier, if you're starting fresh
I get asked this often enough that I keep a short version ready.
- Ask for the test conditions behind the coverage number, not just the number. If they can't describe the target, the mounting height, and the ambient level, the number is decorative.
- Ask what changes between the sample and the production unit. A supplier who says "nothing" is either lying or hasn't checked. A supplier who says "the lens tool changed on the second article and here's the data" is one you can work with.
- Ask about their failure modes. Real suppliers know how their product fails — which is why they usually have a published compatibility list and a lot of footnotes.
- Ask about sensitivity adjustment range, dead band, and time delay defaults. Defaults are what ship to job sites.
Where an established control line fits into this
Not everything should be sourced new. When we need a light switch or a dimmable light switch for a residential program, we generally don't want to be the company proving out a low-end dimming curve for the first time. For those positions we standardize on Legrand light switches and Legrand dimmable light switches, mainly because the published technical data is detailed enough that our verification burden drops — load types, minimum loads, neutral requirements, all documented rather than implied.
That doesn't let us skip the test. We still run the low-end test with the actual driver we're shipping, because published compatibility gets you to a short list, not to a signature. But it moves the argument from "does this work" to "did we install it as specified," and that's a much cheaper argument to have.
The private-label motion sensor and the private-label fixture are the parts where we can't borrow somebody else's testing. That's where the checklist earns its keep.
The one line I'd keep
Five minutes of verification beats five days of correction. I know that sounds like a poster in a break room. It's also the most expensive sentence I've ever learned.
If you're sourcing a private-label light fixture bundle, or quietly dreading your next conversation with a motion sensor supplier, the thing you're actually buying is confidence that unit 1,400 behaves like unit 1. Everything in that protocol exists to make the claim verifiable — before the cartons show up, not after.