Last March, a client called me about their new 2x4 LED troffers. They'd done everything right—or so they thought. Cree fixtures, DLC Premium listed, 4,000K, 4,000 lumens, 0-10V dimming. On paper, it was a textbook commercial retrofit. In the actual space, though, the office looked cold, flat, and somehow dimmer than the fluorescent fixtures they'd just replaced. Three weeks of install, and the facilities manager was already fielding complaints about eyestrain.
I've been on both sides of that call. I handle commercial and retail lighting procurement for about eight years now, and I've personally made (and documented) 11 significant specification mistakes totaling somewhere near $47,000 in wasted budget. The one above—the spec sheet trap—is the most expensive mistake I keep seeing people make, mostly because the numbers all look correct.
The Problem You Think You Have
The instinct is to blame the fixture. Or the brand. Or "LEDs aren't as good as they said." Or you start hunting for the Cree lighting phone number to escalate, because obviously something's defective.
It isn't defective. It's the spec sheet doing what spec sheets do—describing the fixture under test conditions, not in your building.
Here's the thing nobody tells you at the counter: the published lumen output isn't the light hitting your floor. It's what the fixture produces at the photometric test point, measured per IES LM-79-19 in a controlled lab. Between the fixture and the task surface, you lose light to beam angle, ceiling height, mounting orientation, wall and floor reflectance, and driver dimming behavior. None of that shows up on the box.
The Actual Problem: Four Variables That Stack
After enough failed retrofits, I started tracking which variables actually predict whether a job looks right. There are four, and they gang up on you.
1. CCT and CRI are a package, not a pick-list
Everyone grabs 4,000K because it's "neutral." But 4,000K at CRI 80 (the usual default on budget troffers) reads cold and slightly green in a space with warm wood flooring. 4,000K at CRI 90 reads clean and natural. Same Kelvin number, completely different room. The Cree 2x4 LED troffer line offers both CRI options—but if your spec sheet only says "4,000K," you might get either one.
I watched a retail spotlight store redo an entire 1,800 sq ft sales floor because the aisle lighting came in at CRI 80 and the denim wall looked tan instead of indigo. That redo was $8,400 in new fixtures plus four days of lost sales. No one had asked the CRI question at spec time.
2. Mounting height changes everything (the grow light lesson, applied everywhere)
The question "how far above seedlings should a grow light be" gets asked because people intuitively understand that hanging height changes intensity. What they don't internalize is that this applies to every LED installation, at every scale.
Inverse square law: intensity at the surface is proportional to 1/distance². Double the height, quarter the light at the floor. A fixture that's beautifully lit at 8 ft becomes anemic at 12 ft—even though it's the exact same fixture doing the exact same thing.
Warehouse aisle retrofit gone wrong in 2022: the client went from 10 ft fluorescent to 14 ft LED high bays because they added racking. Same lumens per fixture, but useful light at shelf height dropped by roughly half. They ended up buying 40% more fixtures to compensate. That's the kind of math you can't see on a cut sheet.
3. Driver quality doesn't show up in lumens
This is the sneaky one. Two fixtures can have identical LM-79 outputs and behave completely differently under dimming or with controls, because the driver is the difference. Cheap drivers flicker at low dim levels, buzz at certain loads, and fail faster.
IES TM-21-19 projects lumen maintenance over time, but it says nothing about driver lifespan. In practice, driver failures outpace LED failures by a wide margin. If you're specifying to a lower-cost driver to hit budget, you're essentially deferring cost into your maintenance schedule—usually at 2-3x the original delta.
4. Retail lighting is marketing, not infrastructure
Spotlight marketing—literally using light to highlight merchandise—only works when beam angle, color temperature, and CRI are all tuned to the product. A tungsten-style spot at 3,000K and CRI 95 on leather goods says "luxury." The same fixture at 4,000K and CRI 82 on the same shelf says "discount."
Your lighting designer and your marketing team are solving the same problem and almost never talk to each other. When they do, merchandising results are genuinely different—but that only happens if someone in the middle knows to ask both questions at spec time.
What This Actually Costs You
The fixture redo is the visible number. It's not the whole number.
- Labor to uninstall and reinstall: often 40-60% of the original install cost
- Rebate forfeiture: many utility programs require the as-installed fixture to match what was approved. Wrong CRI or wrong wattage goes in, the rebate evaporates—often after the check was promised
- Productivity loss in offices: those eyestrain complaints from March 2024 were real; the tenant rented a co-working space for two weeks while we re-speced
- Lost sales in retail: a dim or green-tinted aisle where customers can't read labels directly costs revenue, and you rarely get attribution back to the lighting
The gut-vs-data lesson here: I once argued for a slightly cheaper fixture because the LM-79 numbers matched the premium option within 3%. My gut said the cheaper one felt "off" on the sample I saw. Turned out the cheaper driver had a known (but not published) issue with our specific 0-10V control system. That mistake on a 2,400-unit order would have been $19,000 in replacement drivers alone.
What to Actually Do (Short Version)
If you're specifying commercial LED work this quarter, three things matter more than the cut sheet:
- Specify CRI and CCT together, and demand CRI 90 for any customer-facing space. CRI 80 is fine for warehouses and back halls. It's a marketing liability anywhere a customer stands.
- Model the mounting height before you order. Ask your rep for the photometric file (the .ies file), not just the LM-79 summary. If they won't share it, that's your answer about the level of support you'll get later.
- Verify the driver separately. Ask for the driver manufacturer and model number, not just the "driver type." Then check it against your controls package.
"The value of guaranteed performance isn't the fixture—it's the certainty that the fixture will perform the way you modeled it. The cheapest path to that certainty is doing the math before the PO, not after the install."
One last thing. Before you dial the Cree lighting phone number or any vendor number to complain about an install that looks wrong, do yourself a favor: photograph the space at night with the lights on, pull your spec sheet, and check those four variables first. About 70% of the "bad fixture" calls I've seen end at one of them. The other 30% are real—and you'll get a much faster resolution when you can prove you've already ruled out the obvious.
That's the checklist. It took me 11 mistakes and $47,000 to build. Use it for free.