-
The checklist, in order
-
Step 1: Check the ceiling before you check the catalog
-
Step 2: Read the spec sheet for the exact SKU, not the collection name
-
Step 3: Decide the light source before you fall in love with the shape
-
Step 4: Run the lumen math per room, not per fixture
-
Step 5: Match the driver and dimming protocol to the control system
-
Step 6: For area lighting, demand the photometric file
-
Step 7: Buy one, install one, then release the full PO
-
Step 1: Check the ceiling before you check the catalog
-
What this checklist still won't catch
If you're the one who signs the purchase order on a lighting package — hotel lobby, restaurant build-out, 40,000 sq ft office refresh, a PIP renovation with a hard opening date — this list is for you. Designers, GCs, procurement leads, owners. Same seven steps.
In this order. Skipping the first one is how I lost six days and $1,180 in March 2021.
I've been handling commercial lighting specification and purchasing for nine years. I've personally made — and documented — 23 significant mistakes, totaling roughly $41,000 in wasted budget. Now I maintain our team's pre-order checklist so nobody else has to repeat them.
The checklist, in order
Step 1: Check the ceiling before you check the catalog
Not the finished elevation. The actual ceiling assembly. I don't care how good the render looked.
Six things to confirm on site, with a tape measure and a flashlight:
- Finished ceiling height, measured — not scaled off a drawing
- Junction box type and weight rating (a nail-on box and a 40 lb crystal piece is not a combination)
- Canopy diameter versus any medallion or ceiling reveal
- Slope — many decorative fixtures accept an adapter, some don't allow one at all
- Joist or beam position for linear and recessed runs
- Ceiling substrate: drywall, T-bar, concrete, plaster
March 2021. Fourteen lantern pendants, 9'-4" finished ceiling. The fixtures arrived with 6 ft of chain but a fixed 12-inch stem we couldn't shorten without cutting into the canopy. We paid $1,180 to have all fourteen re-stemmed and shipped back, and lost six days. One number on the spec sheet would have caught it: minimum overall height.
Step 2: Read the spec sheet for the exact SKU, not the collection name
The Visual Comfort Reed chandelier and the Visual Comfort Grace chandelier come from the same house. They are not interchangeable. Different diameter, different overall height range, different lamping, different canopy.
A collection name is a design language. It is not a specification. When I'm comparing, I pull the actual cut sheet for each SKU and check the same six lines every time: overall height minimum and maximum, canopy dimensions, lamp type and max wattage, fixture weight, and location rating. Dry, damp, or wet — damp matters in covered porches, coastal lobbies, and anything within a certain distance of a shower.
And check whether the listing covers the configuration you're installing. A fixture that's UL listed as shipped can still be installed in a way that voids that listing — an unlisted retrofit kit, a remote driver zip-tied to a joist inside a fire-rated assembly. I've seen both done on projects I inherited, and both had to come back out.
Step 3: Decide the light source before you fall in love with the shape
This is where a lot of decorative specs go sideways, and it's usually because nobody in the room agreed on what the source should be.
What is COB LED lighting? Chip-on-Board: multiple LED dies bonded directly to a single substrate, then covered with one continuous phosphor layer, so the whole board becomes one large, even emitting surface instead of a grid of visible dots. Practically, that means high lumen density in a very small aperture, a smoother beam without the "polka dot" pattern you get from discrete emitters, and a shorter thermal path from die to heatsink. The trade-offs are just as real: the board is a single component, so a failure means replacing the module rather than a $6 lamp, glare control depends entirely on the optic in front of it, and you're locked into whatever color point the driver and board produce.
Where it's the right call: downlights, track heads, linear and architectural runs, and most area lights. Where it's the wrong call: anything whose design intent depends on a compact, bright point source. Crystal refraction is one of those. What most people don't realize is that swapping a lamp-based source for an integrated COB module behind crystal changes the entire effect — you get a soft, even glow instead of sparkle, because the emitting surface is now flat and diffuse. If the fixture was designed around lamps, don't "upgrade" it to COB without seeing it lit in person. Photo doesn't count.
Step 4: Run the lumen math per room, not per fixture
Fixture count is an output, not an input. Start from the target illuminance for the room type, then work backwards.
My rough working method — and yes, verify against a real photometric calculation for anything client-facing:
- Pick the target maintained illuminance for the space from the applicable IES recommended practice
- Multiply by square footage
- Divide by a maintenance factor (I use 0.80 as a default for dirty-ceiling commercial interiors; adjust for atrium and industrial)
- Divide by delivered lumens per fixture
A 300 sq ft conference room at 35 fc maintained works out to roughly 10,500 delivered lumens, about 13,000 raw. Six downlights at 1,800 lm and a 0.8 maintenance factor lands you close. Then check the color story: I want to see TM-30 data, not just a CRI number, because two fixtures can both say 90 CRI and still render warm wood completely differently. And check that the fixtures in a single sightline came from the same production batch — color shift between batches is real and it photographs badly.
Step 5: Match the driver and dimming protocol to the control system
This is the step most people skip, and it's the one that generates the most expensive surprises. Protocol mismatch doesn't show up until the fixtures are hung and the electrician hits the switch.
September 2022. Sixty recessed fixtures across a two-floor restaurant. Our submittal said 0-10V. The lighting control package on site was ELV. Nobody compared the two documents until the fixtures flickered at 15% on the first walkthrough. $2,300 in replacement drivers, three days of delay, one very unhappy GC.
Before you release the order, get four things in writing from the manufacturer:
- The dimming protocol and the minimum load
- Driver location — integrated or remote, and the maximum remote distance
- Whether your specified in-wall dimmer is on their compatible list (not "compatible with standard dimmers")
- The dimming range and what happens below 10%
If the answer to any of those is "depends," you have a conversation to have before the PO, not after.
Step 6: For area lighting, demand the photometric file
Area lights — pole-mounted, wall-pack, broad-coverage fixtures for parking, yards, facades, and loading zones — get specified off cut sheets far too often. The cut sheet tells you lumens. It doesn't tell you where they land.
Ask for the IES or LDT photometric file, and run your own calculation with the actual mounting height and pole spacing. Then check the BUG rating (backlight, uplight, glare) against local code, confirm cutoff classification if your jurisdiction requires it, and check DLC listing if you want the utility rebate. Barrier: a fixture that isn't listed doesn't qualify, and the rebate is usually already in the budget by the time anyone notices.
2020, a small retail parking lot. Fixtures were fine on paper. Neighbors across the street filed a complaint about light spilling into their bedrooms within two weeks. We added external shields to every head — $640 plus labor — because the uplight and backlight numbers were never checked at the design stage.
Step 7: Buy one, install one, then release the full PO
I went back and forth on this one for years. A pilot unit costs money and adds a week to the schedule, and on smaller jobs it genuinely feels like overhead. But after the third rejection in Q1 2024 — a linear fixture that measured correctly and still looked wrong at the actual mounting height — it became non-negotiable on our team.
Order one. Install it. Photograph it in the real space, at the real mounting height, with the real dimmer, at night. Look at it from the entrance, because that's where the client will look first. Then release the rest.
While you're at it, build spares into the order. Our rule: 10% attic stock on anything with glass or crystal, 5% on plaster and alabaster, and at least two extra drivers per driver type. Replacement lead times on made-to-order decorative pieces are not short, and the alternative is a dark socket in a lobby for six weeks.
What this checklist still won't catch
Five things that live outside the document:
Lead-time language. What most people don't realize is that published "standard lead time" on made-to-order decorative fixtures often includes buffer the vendor uses to batch production runs. It's a queue number, not a build time. Ask what the actual production window is for your order, and add your own buffer on the front end.
Lookalikes. I spent two weeks in 2022 going back and forth between an original fixture and a substantially cheaper alternative that looked identical in photos. The risk was never the shape. It was the listing, the warranty, the availability of replacement glass in year three, and finish consistency across a 40-piece order. Cheap fixtures that all match in a photograph often don't match each other in a room.
Freight damage. Inspect crates before you sign the delivery receipt. Once it's signed, the damage conversation changes completely.
Batch variation. Plaster, alabaster, and unlacquered brass vary between production runs by design. Order the whole job at once, or accept that the second shipment will read slightly different from the first.
The thing the client actually remembers. Nobody ever reviewed our line-item budget. They walked into the lobby at 7 pm on opening night and decided whether we were good at our jobs. The $50-per-fixture difference between adequate and right disappeared from the conversation about four seconds after the doors opened — and the opposite is also true. Cheap out on the visible layer and you'll be explaining that decision for years.
Print this. Tape it in the binder. Then go measure the ceiling.