Lighting practice
The LED Purchasing Trap: Why 'Long-Lasting' Doesn't Mean What You Think
When I took over purchasing in 2020 for a 200-person company—roughly $180,000 a year across 12 vendors—I had one rule that worked for almost everything: get three quotes, pick the middle one. That rule served me well for office supplies, janitorial contracts, even our IT hardware refresh. So when our facilities manager handed me a lighting refresh request in early 2024, I applied the same logic.
The list was long: six 32-inch Samsung LED-backlit LCD TVs for the conference rooms (the samsung led backlight lcd tv 32 model our AV guy insisted on), twelve recessed downlights for the east wing, a crystal chandelier for the lobby—the kind visitors actually notice—and a small supplemental lighting setup for the greenery near the entrance.
I picked the middle quote. It was $3,200 cheaper than the highest bid and $400 more than the lowest. Seemed like the definition of a no-brainer.
Fourteen months later, that $3,200 has mostly evaporated. Not in one dramatic failure—in a slow drip of replacement drivers, a flickering chandelier, and two conference room displays with edges that have gone noticeably dimmer than the centers. The kind of stuff that doesn't show up in a budget line but does show up in your week.
The Number on the Spec Sheet Doesn't Mean What You Think
Every LED vendor I talked to in 2024 led with the same number: "50,000 hours." I interpreted that as: at 8 hours a day, this fixture will last about 17 years. That interpretation was wrong.
Here's what that number actually means in most cases. "50,000 hours" typically refers to L70—the point at which measured light output drops to 70% of initial lumens. That value comes from extrapolating LM-80 test data using the TM-21 method, both maintained by the Illuminating Engineering Society (IES). The LM-80 test itself runs LEDs at controlled temperatures and drive currents for a minimum of 6,000 hours, then TM-21 projects the curve forward.
The catch is that the test conditions rarely match your ceiling. A downlight in a hot plenum above a commercial kitchen is not operating in the same thermal envelope as the test rig.
Heat is the biggest variable. Driver quality is second. Switching frequency is third. If any of those deviate from the test conditions—and in a real office, at least one usually does—your actual useful life can land well short of the label.
This same ambiguity shows up in the "how long do Samsung LED TVs last" question that people search constantly, based on what I can tell from my own research. Samsung's published specs for many of their LED-backlit displays cite approximately 50,000 hours of backlight life. But that's the backlight—not the power board, not the panel, not the local dimming zone drivers. In our case, the displays themselves are still working. It's the edge uniformity that's degrading, which is a different failure mode entirely.
Bottom line: treat the spec-sheet number as a test condition, not a warranty. If a vendor can't produce the LM-80 report, the "50,000 hours" is marketing copy.
What "Compatible" Doesn't Cover
The lobby chandelier was the first problem I noticed, but it took me a while to figure out why.
We run a Zigbee-based lighting control system—wireless Zigbee 3.0, managed through a central gateway. The chandelier we bought was listed as "Zigbee compatible." So was the driver. So was the gateway, for that matter. Everything said Zigbee. Nothing worked together.
The issue, as I eventually learned from a very patient technical support rep, is that "Zigbee compatible" can mean at least three different things. Older devices often run ZHA (Zigbee Home Automation) or the deprecated ZLL (Zigbee Light Link) profile. Zigbee 3.0, maintained by the Connectivity Standards Alliance, unified the application layer—but it didn't retroactively make every legacy device interoperable.
A dutiful approval process can still miss this. In our case, procurement saw "Zigbee" on the spec sheet and assumed it meant the same Zigbee we were already running. Three weeks of back-and-forth later, we replaced the driver with a Zigbee 3.0 certified unit. That swap cost $180 and a week of shipping delay.
Here's the part that really stung: finance rejected the $180 replacement because it wasn't on the original purchase order. I had to re-submit it as a separate mid-year adjustment, which took another two weeks and made the whole line item look like poor planning on my end. Nobody sees the original spec sheet. They see the second payment.
Here's what I ask now: "Which Zigbee profile does this device implement, and which coordinator did you test it against?" If the answer is vague or references a 2016 product line, I treat it as a red flag.
The Grow Light Question I Should Have Asked First
Near the end of the project, our facilities lead asked for supplemental lighting on the entryway planters. I'd never bought a grow light before, so I did what I always do: searched for the obvious question. "Can any light be a grow light" returns a genuinely confusing mix of answers, ranging from "yes, basically" to "no, you'll kill your plants."
The honest answer is somewhere in between. From a technical standpoint, any light emitting in the 400–700nm range—which is the photosynthetically active radiation (PAR) band—can contribute something to plant growth. But "can contribute something" and "efficiently supports growth" are different claims.
Standard white LED fixtures concentrate their output in the green-yellow part of the spectrum, which plants largely reflect. Dedicated grow lights emphasize blue peaks (around 450nm) and red peaks (around 660nm), which are the wavelengths plants actually use. Same wattage, very different results.
I bought the cheap option first. Two of the ferns near the front desk started yellowing within six weeks. We swapped to proper grow fixtures—about $340 for the small setup we needed—and the same plants improved on the same watering schedule.
The real mistake wasn't buying the wrong light. It was not asking the facilities team whether the goal was "keep the plants alive" or "make them visibly healthy in a client-facing area." Those two goals have different price points, and I optimized for the wrong one.
The Real Cost of Cheap
Tallying it up over the first year:
- Two 32-inch displays with edge brightness degradation — still under warranty, but the warranty only covers total failure, not gradual uniformity loss. Replacement cost if we push for it: roughly $260 each, and a fight with the supplier.
- Chandelier driver replacement (Zigbee 3.0 certified) — $180, plus three weeks of coordination and a finance re-submission.
- Grow light replacement — $340.
- My time, conservatively — about six full workdays spread across the year, mostly on phone calls and scheduling return visits.
That's roughly $780 in direct extra cost, plus time I didn't get to spend on anything else. Against a $3,200 initial "saving," we netted maybe $2,400—and I lost a year of not having to think about lighting, which was the whole point of buying new fixtures.
What I Do Now (Short Version)
This part is intentionally brief, because the checklist isn't the insight—the insight is that the checklist changes based on how the product is actually used.
- Ask for LM-80 data before signing. Not the number on the brochure. The report.
- Specify the Zigbee profile. "Zigbee compatible" is not enough. Ask which coordinator the vendor tested against.
- Define the goal before the spec. "Keep plants alive" and "look good for clients" require different fixtures, different budgets, different everything.
- Build a 10–15% compatibility buffer into the LED line item. If you don't use it, that's great. If you do, you won't have to explain to finance why last year's "long-life" fixtures are already on a replacement list.
None of this means the Samsung LED displays in our conference rooms are bad. They're fine when the use case matches the spec. Same with the downlights. The problem was never the product category—it was my assumption that a spec-sheet number and a compatibility logo answered questions they were never designed to answer.
I still get three quotes. I still pick the middle one sometimes. But now I read the test conditions first. That's the whole difference.