There Isn't One Checklist
I've tried to write a single sourcing checklist for architectural lighting. It doesn't survive contact with real projects. I work in quality and brand compliance for a distributor that carries iGuzzini commercial lighting alongside other architectural lines, and roughly 200 to 250 fixture specifications cross my desk in a year. In 2023 we sent back about 30% of first deliveries—not because the fixtures failed, but because the paperwork, the optics, or the dimming behavior didn't match what we'd promised the customer.
What those returns taught me is that the questions which save a retrofit will waste your time on a new build. So instead of one checklist, I keep three. Which one applies depends on the constraint you're working against—not the budget, and not the logo on the box.
Retrofit, specifier-led, multi-site. In that order of how often people misjudge which one they're in.
Scenario 1: The Retrofit, Where the Old System Sets the Rules
The counterintuitive part: don't start with the luminaire. Start with whatever is staying.
In a retrofit, the binding constraint isn't the fixture—it's the dimming protocol, the ceiling cutout, and the driver architecture you're inheriting. If the building runs 0–10V, a DALI-2 luminaire with an internal driver is the wrong answer no matter how good its photometrics look on paper. We check three things before we read a single lumen figure: control protocol and supply voltage, mounting interface (cutout size, trim depth, plenum clearance), and whether the driver sits inside the fixture or remotely. Get those wrong and you've bought a nice object that doesn't fit the hole.
In 2023 I skipped a bench test on a 150-piece retrofit because the new fixtures came from the same manufacturer family as the ones already installed. Same brand, same protocol, should be fine. Two weeks after install, the client reported flicker at low dim levels and an audible buzz in three corridors. The drivers were a different generation. We swapped 150 drivers at our cost and pushed handover back nine days. Every retrofit order now includes a three-piece dimming test at 1%, 10%, 50%, and 100% before the main shipment leaves.
Quotes are the other place retrofits go sideways. The headline number usually excludes mounting brackets, through-wiring connectors, dimming modules, and commissioning. On a 2024 retrofit quotation, those items added 11% on top of the fixture line. Nobody was hiding anything—the line item simply said 'luminaire.' Which is why I ask the exclusion question first.
I've learned to ask what's not in the number before I ask what the number is.
Scenario 2: The Specifier-Led New Build, Where Documentation Is the Product
This is the scenario where architectural lighting sourcing turns into document review. A specifier has already run the calculations. Your job is to prove the fixture does what the calculation assumed.
The distinction that trips people up
LM-79 covers electrical and photometric measurement of a complete luminaire. LM-80 covers lumen maintenance of the LED package or module—not the fixture, and not the driver. TM-21 extrapolates from LM-80 data. When a vendor hands you an LM-80 report and calls it a luminaire lifetime claim, they're either confused or hoping you are. Ask for LM-79 on the exact configuration you're buying, and treat any lifetime figure as valid only at the drive current and thermal conditions that were actually tested.
For anything in a workspace, EN 12464-1 sets the glare limits—UGR 19 for offices and UGR 25 for general industry are the numbers people quote most often. Those come from the standard, not from a brochure. If your calculation assumes UGR 19 and the delivered optics measure 22, the room is out of spec even though nothing is technically broken.
Spotlights carry a separate set of checks. Spotlight compliance requirements usually include IEC 62471 photobiological safety (the blue-light hazard group matters for narrow-beam, high-luminance products), documented beam and shielding angles, and—for retail and museum work—UV and IR content where the application is light-sensitive. Add IEC 60529 for IP and IEC 62262 for IK wherever the fixture is exposed or within public reach. In the EU, Regulation 2019/2020 covers light sources and separate control gear. And to be clear, I'm citing frameworks, not certifying your product: compliance is per configuration and per market, and the current published version is the one that counts.
On optics, compare photometric files rather than rendered images. When we were evaluating narrow-beam options for a gallery project, we pulled the published IES files for iGuzzini architectural lighting products in the same beam class and overlaid their candela distributions against our shortlist. Two fixtures that looked identical in a catalog had meaningfully different beam edges. That comparison took an afternoon and ended a three-week argument.
Ask for the raw IES or LDT file, not a screenshot of a photometric report. A ceiling light manufacturer that publishes full LM-79 reports and IES files per configuration is telling you something about how they work. One that sends a PDF summary with no named test lab is telling you something too.
Scenario 3: The Multi-Site Rollout, Where Consistency Beats Peak Performance
Here's the one that runs against normal procurement instinct: on a multi-site project, a slightly less efficient fixture that matches perfectly across 40 buildings is usually the better buy than the best-in-class product you can't repeat.
We learned that the expensive way. I told a supplier we needed 'standard 3000K.' They heard a nominal color temperature. I meant 3-step MacAdam binning. They delivered 5-step. On one fixture, nobody would have noticed. Across a corridor of 60 downlights, the variation was obvious to anyone walking the space, and the client spotted it before we did. Our purchase orders now state the CCT, the binning step, and the reference standard—ANSI C78.377 defines the nominal bins—because 'standard 3000K' means at least three different things in this industry.
I do not mean that as a small nitpick. Color consistency is the thing end users actually experience, and it's the thing that quietly decides whether a rollout gets extended to the next building.
Here's the thing about light output: a 5 lm/W difference is a spreadsheet line. A visible mismatch between floors is a maintenance ticket you'll be answering for years. The other three items I insist on for multi-site work are a locked bill of materials with written change notification, an end-of-life notice period long enough to buy spares (12 months is my floor, and yes, I've had pushback), and retained samples from the first production batch. The samples have settled two disputes in four years, both about color. Cheap insurance.
How to Tell Which Scenario You're In
Most projects are one scenario with pieces of another. A new build with a preserved historic wing is Scenario 2 with a Scenario 1 pocket inside it. A rollout that starts in one building is Scenario 1 turning into Scenario 3 as it expands. So the useful question isn't which label fits—it's which constraint is actually binding.
- If the ceiling, wiring, and controls stay put, you're in Scenario 1. Your first deliverable is a compatibility test, not a photometric file.
- If someone else's calculation has to be defended, you're in Scenario 2. Your first deliverable is the test report and the raw IES file.
- If the same fixture will appear in more than three locations, or over more than 12 months, you're in Scenario 3. Your first deliverable is the BOM lock and the binning spec.
Where it gets messy is when two constraints are live at once—a phased renovation usually is. In that case, run the tighter checklist first. The retrofit compatibility test takes an afternoon and saves a shipment. The photometric review takes a week and saves a re-specification.
This reflects what I've been working with as of early 2025. Standards get revised and regulations get replaced, so check IEC 60529 and the EU ecodesign rules against the current published version before you write them into a contract. What hasn't changed in the six years I've been doing this is the pattern. Projects almost never fail because someone bought a bad fixture. They fail because nobody asked what wasn't in the quote, or because two people used the same words to mean different things.
If you take one thing from this, take the ordering: match your vetting to the constraint, not to the budget. The budget tells you what you can afford. The constraint tells you what will actually fail.
