Three weeks after a contractor installed 200 LED downlights in a hotel bathroom renovation, six of them started flickering. His first call to the distributor was short: "You sold me a bad batch."
The distributor forwarded it to me. We pulled batch records, tested the returned units, and compared them against the spec we had on file. The downlights were fine. The spec wasn't.
The order had been placed as "suitable for damp locations," and the fixtures were marked IP44. The contractor installed them directly above the shower in guest bathrooms, where water hits the lens face daily. IP44 handles indirect splash and steam, not direct water. That install location calls for IP65.
By week six, the contractor had a $22,000 problem: ripping out fixtures, upgrading trims, and renegotiating with the general contractor over who covered the labor.
I'm a quality manager in the lighting industry. I review roughly 200 unique products a year before they reach the market, and I handle compatibility questions from distributors and contractors. I rejected about 18% of first production samples I evaluated in the last year—mostly for documentation and spec mismatches.
The pattern I keep seeing is this: the product is doing exactly what it was designed to do. It just got specified, or installed, for a different situation than the datasheet assumed.
If you're a wholesaler or installer, this is about what's actually going wrong—and why five minutes of verification can save you five weeks of correction.
The Products Weren't the Problem
Here's the counterintuitive part, and it took me a few years to fully accept: more than half of the "bad product" cases I investigate end up as specification or verification failures. The hardware is within tolerance. The process around it isn't.
The conventional wisdom says cheap factories cut corners, and budget sourcing gets you junk. That does happen—I've rejected batches for genuine defects. But the more common failure is the gap between what a spec sheet means and what the installation actually needs. And nobody catches it until the fixtures are in the ceiling or the bulbs are already on the truck.
Where Specs Actually Break Down
Automotive bulbs: "fits" doesn't mean "works"
Take the Sylvania 7506 bulb, one of the most-searched part numbers in the Sylvania lights automotive lineup. It's a cross-reference code in the 3157/3057 family, and aftermarket suppliers list it as compatible across a broad range of vehicles. What I've learned from testing these bulbs is that socket compatibility is not optical compatibility.
When an LED bulb replaces a filament bulb, the light source sits on a different plane in the housing. In a reflector headlamp, a few millimeters of difference in the Z-position changes where the beam lands. The pattern shifts, the cutoff gets sloppy, and the bulb "works"—it lights up, it has the right connector—but it can glare into oncoming traffic. FMVSS 108 (the federal standard for automotive lighting) sets photometric requirements, and a bulb can pass the socket check yet fail the beam pattern in the actual housing.
Sylvania's automotive lighting application guides document this stuff. They get ignored because "fits your vehicle" on a package is easier than checking a PDF.
(Note to self: this is the example I always lead with because nobody thinks about light source position until they see a beam-comparison photo.)
Bathroom downlights: IP ratings are zone-specific
Bathroom downlights cause the same confusion in wet and damp locations. The IP system, defined by IEC 60529 (iec.ch), classifies ingress protection. IP44 means protected against water splashed from any direction—but it is not "waterproof," and it is not rated for direct water contact.
For a bathroom wall or a vanity area, IP44 is usually fine. Directly above a shower, where water contacts the fixture face, you want IP65. Inside a shower enclosure or bath tray, IP67. These distinctions are not obscure technicalities; they decide whether a fixture fails in month 6 or still runs in year 15.
I've inspected installations where an IP44 downlight sat inside a steam shower enclosure. It corroded, the driver died, and the ceiling board took water damage before anyone noticed. Slow, silent, and expensive.
Panel lights: the "IPS vs LED" question doesn't exist
One question that lands in my inbox with confusing regularity is "panel IPS vs LED"—as if IPS and LED are two competing choices for the same product. IPS is a display technology (in-plane switching), not a lighting technology. An "IPS" fixture label is usually just marketing on a flat LED panel.
The distinction that actually matters in a flat LED panel is edge-lit vs. direct-lit (back-lit). Edge-lit panels are thinner and cheaper. Direct-lit panels use more LEDs, produce more even, stable light, and hold up better over the panel's lifetime. If a supplier's comparison chart says "IPS vs LED," that's a red flag on their spec sheet before you even talk numbers.
Zigbee: compatibility is a spectrum, not a checkbox
Smart lighting is where the same failure pattern goes digital. A spec sheet says "Zigbee 3.0 compatible," the buyer assumes it will work with their gateway, and then sensors drop off the network or group commands fail. Both vendors' support teams point at each other.
I've tested "Zigbee compatible" devices that implement different subsets of the standard. They might support basic on/off but not touchlink. They might bind fine to one hub and poorly to another. Certification by the Zigbee Alliance (now part of the Connectivity Standards Alliance) is meaningful, but it doesn't predict behavior with every gateway in the field.
The fix is boring and effective: ask the manufacturer for the Zigbee PDF—the device profile that lists clusters, endpoints, and binding modes. Then check your gateway firmware against it. In every interoperability case I've dug into, the answer was in those documents. Nobody had opened them before installation.
What a Failed Spec Actually Costs
Let me give you three numbers from my audit files.
500 bulbs, three days of labor. A distributor ordered Sylvania 7506 bulbs based on a cross-reference table that matched the connector base but not the light source position for the target vehicle. The customer returned 200 units, and the distributor had to hand-sort the whole batch to separate correct from questionable. That's a relationship hit that doesn't show up on an invoice.
1,200 downlights, $22,000 redo. The hotel bathroom from the top. The products were fine; the IP spec didn't match the installation zone. New fixtures, labor, drywall repair, and a general contractor who stopped trusting the subcontractor's specs.
50 Zigbee sensors, one lost week. The controls retrofit passed acceptance in the facility manager's office, then 12 sensors dropped off the network in the first week on the actual floor. Root cause: gateway firmware was two versions behind what the manufacturer's Zigbee PDF specified. A 20-minute firmware update fixed it, after a week of troubleshooting.
In all three cases, the hardware was within tolerance. The failure was a process gap. And the cost of that gap was orders of magnitude higher than the time it would have taken to verify the spec upfront.
That's the principle I've come to believe after years of reviewing batches: five minutes of verification beats five weeks of correction. It sounds too simple to be a headline. The difficulty is making it a habit before something is already in the ceiling.
The Five-Minute Verification Protocol
This is the protocol I use when I'm on the distributor or installer side of the table. About five minutes per product line.
- Demand the engineering datasheet, not the product page. Product pages are marketing. Datasheets have the numbers: light source position, IP rating, operating temperature, cluster support. If a vendor can't produce one, that's your answer.
- Verify cross-references against the application. For automotive, check the actual vehicle, not just the connector base. For fixtures, check mounting and environment.
- Match the IP rating to the installation zone. A bathroom isn't one zone. Define where the fixture goes before you order.
- Ask for the Zigbee cluster list, not just the word "compatible." Compare it against the gateway firmware you're actually running.
- Test a sample in the real environment. Not on a bench—in the actual ceiling, with the actual driver or gateway, for 48 hours. This single step catches more than a month of spec-sheet reading.
Even after signing off on a sample batch, I second-guess myself. I remember one order where I approved the sample and immediately thought: what if the production run uses a different LED bin? The three weeks until the container arrived were stressful. But we had verified the datasheet and documented the photometric test. When the product arrived, it matched the sample. The doubt was wasted energy—but the verification wasn't.
Where My Experience Stops
I'll be honest about limitations. I review products within a portfolio that mostly has its documentation in order—Sylvania and related lines. If you're sourcing from vendors that don't publish datasheets or application guides, you'll hit worse problems than the ones described here.
My numbers come from my own audits over the past four years, not from a formal industry study. I don't have peer-reviewed statistics. What I have is a pattern that repeats often enough that I'm confident writing it down.
If you're in a different lighting segment—horticultural, entertainment, emergency—your spec traps are different. I can't speak to those with authority.
The Bottom Line
Next time a lighting product fails on site, before blaming the manufacturer or distributor, pull the spec sheet. Check what the seller claimed, what the datasheet says, and what the installation actually required.
You might find the products were fine and the specification was the weak point.
That's the most fixable problem there is.
