A commercial cannabis room buys grow light fixtures by the photon, but it buys the photons from a company. The fixture on the quotation is a specification. The company behind it decides whether the second batch arrives with the same diodes, whether the pallets land before the flip, and whether a driver can be replaced in week three of flower instead of after the cycle. This guide is about that second decision.
It is written for a buyer running or building a commercial room: someone comparing three or four suppliers for a flower room, a veg room, or a multi-room build-out. It assumes the lighting specification is already being handled, and links to the guides that cover it. What follows is the part that quotes rarely address: how to verify a supplier, how to make batch consistency contractual, how to schedule a delivery against a crop, how to accept a shipment, and what to put in writing so that year three is not a negotiation.
One note on vantage point. We are a grow light supplier, so this checklist is also the one we expect to be held to. Every document named here is one a buyer can request from us or from any competitor, which is the point: a standard that only one company can meet is marketing, not procurement.
A lighting purchase fails in five distinct ways, and each one has a different early warning signal. Price comparison collapses them into a single number, which is why two quotes that look 8 percent apart can be a year apart in outcome.
| Risk | How it shows up | What it costs in a room | Signal to check before ordering |
|---|---|---|---|
| Quality escape | A share of units under-deliver or fail early: low PPFD in the corners, a driver that dies in week two, a bar that flickers at 30 percent dim | Warranty replaces the fixture. Nothing replaces the canopy gap or the week lost at the wrong density | Burn-in hours and criteria, field failure rate with the three main failure modes, not a generic defect rate |
| Batch inconsistency | The second order is not the first: different diode bin, different driver revision, a spectrum that reads 40 K cooler | Recipes stop reproducing, two rooms under one controller behave differently, yield drifts without an obvious cause | A written change-notification clause, a retained sample per batch, batch codes on carton and fixture |
| Lead time slip | Pallets arrive after the flip, or half the order arrives and the rest follows in three weeks | Plants flower under the wrong density, or a room sits half lit while the schedule runs | A production calendar with the week your PO is slotted, and the ship dates of the last three orders |
| After-sales failure | No reply, no part, or a return-to-factory policy on a fixture that weighs 25 kg | A 40 dollar driver idles a 1,300 W fixture for two weeks of flower, and the replacement becomes a second purchase with a second lead time | Part numbers and prices for drivers, bars and fans, plus a response commitment written into the order |
| Compliance gap | The listing covers a similar model, not the quoted one; the DLC entry is for last year's configuration | Failed inspection, a rebate denied after installation, re-testing on your schedule | The NRTL file number and the QPL identifier checked against the exact model string on the quote |
The first two are the ones buyers underestimate. A quality escape is visible within weeks and can be argued about with photographs. Batch drift is slower and quieter: the room still lights up, the plants still grow, and the recipe that worked in spring produces 8 percent less in autumn for reasons nobody can name. By the time it is suspected, the samples are gone and the cartons are in the baler.
This is not a question about which type is better. Each has a different cost structure, and each fails in a different place. The mistake is buying from one type while expecting the behaviour of another, which is how a buyer ends up with factory pricing and trading-company lead times.
| Type | What you actually get | Where it fits | Where it fails |
|---|---|---|---|
| Factory with OEM and ODM work | Configuration control, direct engineering access, lower unit cost at volume, your brand on the carton | A single model at volume, a custom spectrum or housing, an OEM line, a buyer who can specify and inspect | Thin finished-goods stock, longer lead times, little local support, and nobody to escalate to when the line is full |
| Trading company | A basket of factories, faster samples, English-speaking project management, consolidated shipments of fixtures, bars and controls | Mixed orders, small first projects, buyers who need one contact and one invoice | Less control over the line, quality that depends on which factory quoted that month, and a warranty chain with one more link |
| Brand with stock | Listed products, local stock, published warranty process, technical support in your time zone | Schedule-critical retrofits, rebate projects, multi-site rollouts where downtime costs more than the premium | Price premium of 20 to 40 percent, limited configuration freedom, and model changes you do not control |
The practical question is not "are you a factory" but "who controls the variables I care about". A trading company that has shipped the same model for four years and holds stock may beat a factory that builds it to order. A brand that stocks the driver you need in your country may beat both, even at a premium, if the alternative is a two-week wait for a part.
Four questions separate the types without a factory visit. First, does the business licence scope include manufacturing, and does the address on it match the ship-from port? Second, which line will run your model this week, and what else is on it? Third, who designs or selects the driver, and can they name the brand and the part number? Fourth, can they walk from the office to the line on a live video call, without cutting away, at a time you choose rather than a time they choose? A factory answers these with specifics that are slightly boring. A reseller answers with photographs.
A sample proves that one unit can be built. It does not prove that the batch will match it, because samples are often hand-built and hand-tested. The documents below are the ones that describe the batch rather than the demo unit, and each can be checked independently of what the salesperson says.
| Document | What it proves | How to check it yourself |
|---|---|---|
| NRTL listing record for the exact model | That the safety scope covers this wattage, voltage, mounting and place of manufacture | Search the UL or ETL directory by file number and compare the model string character by character with the quote and the label |
| DLC Horticultural QPL identifier | That the model meets the current efficacy threshold and is eligible for utility programmes | Search the QPL by manufacturer and model, confirm the version and that the entry has not been delisted |
| Batch test report with date, serial range and sphere identifier | What the units that ship actually measured, not what the prototype measured | Compare the report date with the production date, and the serial range with the cartons you will receive |
| IEC 62471 photobiological report | The eye and skin risk group at the operating distance you will use | Check the risk group against the warning label on the unit, and against your working height |
| Capacity calendar | Whether your delivery date exists on the floor or only in the email | Ask which week your PO is slotted into, and the actual ship dates of the last three orders of this model |
| Spare parts list with part numbers and prices | Whether year three is serviceable or disposable | Ask for a photograph of the shelf and the price of one driver, one bar and one fan kit |
Two checks catch most of the theatre. Ask whether the test report describes the sample or the batch, and ask what the serial range on the report is. A supplier with a real batch process answers in one sentence. A supplier without one explains why the report has no serial numbers.
Model numbers are not specifications. A model can stay the same while six things underneath it change: the diode bin, the driver model or its firmware revision, the optic, the aluminium extrusion, the thermal interface material, and the connector. Any one of those changes the spectrum, the output or the failure rate, and none of them changes the model number on the carton.
What to ask for is a change process, not a promise. Three clauses cover it. A product change notification obligation: the supplier notifies you in writing at least 30 days before a change that affects form, fit, function or measured output, and you have the right to reject the affected batch. A batch code on both the carton and the fixture, so that a failure in month nine can be traced to a production week. A retained sample per batch, held for 24 months, so that a dispute in year two can be settled by measurement rather than memory.
The test is simple and most buyers never run it. Buy two batches six months apart, then measure the same nine points in the same room with the same meter at the same height. A drift of more than 5 percent in average PPFD, or a visible shift in spectrum, means the BOM moved. That is not necessarily a defect, but it should have been a conversation.
There is a commercial version of the same question. Ask what happens if a diode is discontinued mid-project. A supplier who has planned a substitution, qualified it and can show the test comparison is managing a supply chain. A supplier who says it will not happen has not checked.
Lead time is usually quoted as a single number, which hides the only part you can act on: the sequence. A 30-day lead time that includes 20 days of waiting for a driver is a different risk from 30 days of production with materials in stock, even though both look identical on a quotation.
| Stage | Planning days | What can compress it | What cannot |
|---|---|---|---|
| Order to material release | 3 to 7 | A clear PO, deposit cleared early, a locked configuration | Deposit clearing, credit approval, a specification that keeps changing |
| Production | 25 to 35 | A reserved line slot, drivers already in stock, partial shipments | Driver lead time if it is built to order, and the queue ahead of you |
| Sea freight | 25 to 35 | Port pair, a direct sailing, a premium service | The sailing schedule, and a rollover at the terminal |
| Customs and inland | 7 to 10 | A broker briefed before arrival, complete documents | An inspection hold, which is not on anyone's calendar |
| Install and commission | 3 to 5 | Pre-assembly, a lift plan, an electrician booked in advance | Electrician availability, and a room that is not ready |
Add the stages and a door-to-first-light window of 60 to 85 days is realistic for a full container to North America or Europe, before any buffer. A buyer who orders 30 days before the flip has ordered a delay and is about to discover what it costs.
Here is the arithmetic on a 40-bed room. Each bed is 4 by 6 ft, or 24 square feet. At a planning figure of 45 g per square foot, a bed yields about 1,080 g, so a cycle is about 43 kg. Flower runs 63 days, which allows roughly 5.8 cycles a year, or about 250 kg. A two-week delay is 22 percent of one cycle, which is close to 9.6 kg of flower, plus the quality effect of flipping at the wrong density and harvesting off schedule. That is the number a supplier's lead time is measured against, not the freight rate.
Two terms make the schedule enforceable. Milestone dates rather than a single lead time: material release, production start, production finish, ship date, arrival date. And a partial shipment option, so that the first room is lit and producing while the second container is still on the water. A supplier who cannot split a 40-fixture order into two shipments has a batch process that only works at one size, which is its own warning.
Most buyers inspect a grow light shipment by opening one carton. That tells you whether the carton survived, not whether the batch is good. The workable alternative is a sampling plan agreed before the order, so that both sides know what happens when a defect is found.
ISO 2859-1 at general inspection level II, with an acceptance quality limit of 2.5 for major defects, is a common and defensible starting point for fixtures. Critical defects, defined as anything that creates a shock or fire risk, are handled separately: one occurrence rejects the lot.
| Lot size | Code letter | Sample size | Accept | Reject |
|---|---|---|---|---|
| 151 to 280 | G | 32 | 2 | 3 |
| 281 to 500 | H | 50 | 3 | 4 |
| 501 to 1,200 | J | 80 | 5 | 6 |
| 1,201 to 3,200 | K | 125 | 7 | 8 |
Three things belong next to the sampling plan. A first-article check on three units: 30 minutes of burn-in, a PPF reading with a quantum sensor at a stated distance, and a comparison against the batch report. A definition of a major defect that both sides accept: a unit that does not reach 95 percent of rated PPF, a driver that fails the burn, a bar with a dead channel, a label that does not match the listing. And a failure path: who pays return freight, whether the remedy is credit or replacement, and how many days the replacement takes.
Before signing the delivery receipt, count the cartons against the packing list, photograph the pallet labels and any damaged corner, and note exceptions on the receipt itself. A claim made after the truck leaves is a correspondence. A claim noted on the receipt is a fact.
Five years is a duration, not a service level. Two suppliers can both offer five years while one replaces a driver in 20 minutes and the other asks you to crate a 25 kg fixture and wait two weeks for a factory repair. The difference is not in the warranty paragraph. It is in whether the parts exist, whether they have part numbers, and whether they can be shipped.
| Term | What it often says | What to ask instead |
|---|---|---|
| Warranty length | Five years, parts and labour | Are drivers included for the full term, is labour covered, and what voids it |
| Lumen maintenance | 90 percent at 36,000 hours | L90 with the TM-21 projection, the case temperature and the drive current it was measured at |
| Remedy | Return to factory for repair | Field-replaceable driver and bar, with the part number and a 20-minute procedure |
| Response | Reply within 48 hours | Ship a part within 48 hours, and who pays the freight in each direction |
| Spares | Available at list price | Three percent of the order at order price, part numbers and a price list locked for three years |
| Advance replacement | Not mentioned | Ship the replacement against a PO or a card, credit on return, so the room is not dark while paperwork moves |
There is a version of this test that takes one email. Ask for the part number of the driver, the price, and a photograph of the shelf it sits on. A supplier who can answer in an hour has a spare parts function. A supplier who forwards the question to an engineer and comes back with a warranty clause has a warranty.
The economics are easy to state. On a 40-fixture order, three percent spares is about 1.2 fixtures, or a driver and two bars and a fan kit. That is a small fraction of the order value, and it is the difference between a repair and a re-purchase, which is the most expensive way to buy the same fixture twice.
A quotation is a document with a number at the bottom and a set of assumptions above it. The number is what gets compared; the assumptions are what decide the landed cost. Two quotes that differ by 10 percent per fixture can reverse once freight, duty, mounting, spares and commissioning are in the same column.
| Line item | Why it changes the comparison | What a stripped quote leaves out |
|---|---|---|
| Fixture and driver | The driver is the part most likely to fail and the part most often priced separately | A fixture price that assumes a driver sold as an accessory |
| Mounting, cable and connectors | Bars need hangers, rope ratchets or rails, and the cable length has to reach the circuit | Hangers, extension cables, plugs matched to your voltage |
| Spares | Three percent at order price is cheaper than any part bought later | Any spare at all, or spares priced at list |
| Freight and Incoterm | FOB, CIF and DDP shift who owns the risk at each stage | Which Incoterm the number assumes, and who insures the sea leg |
| Duty, tax and last mile | These are real costs that appear after the invoice, not on it | Duty rate, customs brokerage, delivery to the site rather than the port |
| Commissioning | Layout, hanging height, dimming schedule and a first PPFD map | Any installation support, or support priced per day |
| Payment structure | A deposit and a balance have a financing cost, and full prepayment has a risk cost | The cost of money, and the leverage lost with full prepayment |
Normalise the quotes before comparing them. Divide the delivered cost by the µmol/s per fixture, and by the µmol/s that actually lands on the bed after spacing and overlap. The first number tells you what you paid for photons at the dock. The second tells you what you paid for photons in the canopy, which is the only number the crop sees. The guide to µmol/J explains why the two can differ by more than the efficacy gap between fixtures.
Ask for the quote in a fixed format, and ask each supplier to state the assumptions rather than burying them. Mounting height, footprint, spacing, dimming range, ambient temperature, and whether the price includes the driver. A quote that cannot be read line by line is not a cheaper quote. It is a quote you will renegotiate after the deposit.
Most disputes are not about good faith. They are about a document that did not say what both sides assumed. Ten clauses carry the weight in a lighting order, and each one is short enough to negotiate in a single call.
A second source is not a second supplier for the same model. That is a second set of batches, a second set of reports and a second learning curve, and it doubles the inconsistency risk. The workable version is a second model, qualified on two fixtures through one full cycle, so that a supply interruption does not stop a room.
Spare inventory follows the same logic as the warranty question. The right level is not a percentage chosen for comfort, but the number of parts that keeps a room lit until a shipment arrives. For a 40-fixture room that is usually one or two spare fixtures, two drivers, four bars and a fan kit, held on site rather than at a supplier's warehouse, because the point of a spare is the week when shipping takes too long.
Score each supplier from 1 to 5 on each line, multiply by the weight, and add. Two suppliers within 10 points of each other are tied on paper, and the decision should go to the sample, the answers to the uncomfortable questions, and the part number you were given without being asked twice.
| Criterion | Weight | What earns a 5 |
|---|---|---|
| Quality evidence | 25 | Batch report with serial range and date, burn-in hours and criteria, field failure rate with the three main modes |
| Batch consistency | 15 | Written change notification, batch codes, retained samples, a qualified substitution plan |
| Lead time and capacity | 20 | A slotted week, real ship dates for recent orders, partial shipment offered without being asked |
| After-sales and spares | 20 | Part numbers, prices, shelf photograph, 48-hour part shipment, field-replaceable design |
| Compliance | 10 | Listing and QPL records that match the quoted model, photobiological report, label consistent with the file |
| Commercial terms | 10 | Acceptable payment structure, delay damages, a warranty that includes drivers |
Weight the first and fourth lines hardest for a retrofit, because a retrofit inherits a schedule and a crop that are already running. Weight lead time hardest for a new build, where the flip date is fixed by construction and genetics rather than by the lighting order. The weights are a judgement, and the point of writing them down is that they stop changing after the first impressive factory video.
Three quotes are only comparable if all three answered the same question. The fastest way to get there is to send the same five inputs to every supplier in the same email, and to ask for the answer in the same units.
Ask for the reply in a fixed format: fixture count, µmol/s per bed, a PPFD map at your mounting height, wall watts, BTU/h, lead time with milestone dates, and the part numbers for the driver and bar. A supplier who answers in that format is sizing a room. A supplier who answers with a price list is selling cartons. The buying checklist covers the technical side of the same brief, and the fixture count guide shows the geometry behind the numbers a good reply should contain.
Ask four questions and listen for specifics. Does the business licence scope include manufacturing, and does the address match the ship-from port? Which line will run this model this week, and what else is on it? Who selects the driver, and what is its part number? Can you walk from the office to the line on a live video call at a time I choose? A factory gives short, boring, specific answers. A reseller gives photographs and adjectives.
Six: an NRTL listing record for the exact model, a DLC Horticultural QPL identifier if the project needs rebates, a batch test report with a date and serial range, an IEC 62471 photobiological report, a capacity calendar with your week slotted, and a spare parts list with part numbers and prices. Each one can be checked independently, which is what makes them useful.
Plan 60 to 85 days from order to first light for a full container: 3 to 7 days to material release, 25 to 35 days of production, 25 to 35 days of sea freight, 7 to 10 days of customs and inland transport, and 3 to 5 days to install and commission. Add a buffer and order about 100 days before the flip. Production and freight can compress; the sailing schedule and a customs hold cannot.
Five years with drivers included, L90 at 36,000 hours with the case temperature and drive current stated, field-replaceable drivers and bars, and advance replacement so a room is not dark while paperwork moves. A five-year term that excludes the driver, or requires a return to factory, is a shorter warranty in practice than a three-year term with parts on a shelf.
Two to three percent of the fixture count, plus consumable parts. On a 40-fixture room that is one or two spare fixtures, two drivers, four bars and a fan kit. Hold them on site, because the purpose of a spare is the week when shipping takes longer than the repair.
AQL is the acceptance quality limit used in ISO 2859-1 sampling. At general inspection level II with an AQL of 2.5, a lot of 500 units is sampled at 50 pieces, and three defects are accepted while four reject the lot. Critical defects such as a shock or fire risk reject the lot on the first occurrence. Inspect on arrival, and note exceptions on the delivery receipt rather than in a later email.
Buy from whichever controls the variables your project depends on. A brand with local stock and a published warranty process is the lower-risk choice for a schedule-critical retrofit. A factory is the better choice when you need a configuration, an OEM brand or a lower unit cost at volume, and you have the capacity to specify, inspect and hold a supplier to a change process.
Yes, if the supplier can stagger production and ship in phases. Ask for a milestone plan across the rooms, a partial shipment schedule, and the same batch code across a room so that a fault can be traced. A supplier who can only build the order as one lot will light the last room last, which is rarely the order the plants need.
The model and configuration tested, the production date, the serial range covered, the sphere or laboratory identifier and calibration date, PPF in µmol/s, efficacy in µmol/J, the spectrum, input watts and power factor, and the test conditions. If the serial range is missing, the report describes a prototype, not the batch that will ship.
None of this replaces a lighting specification. A room still needs the right photon output per bed, the right spectrum and the right heat load, and the guides linked above cover those. What the supplier decision adds is the difference between a specification on paper and the same photons arriving in the same quantity, on the date the crop needs them, with a part number you can call in year four.
Send the room plan, mounting height, current fixtures and your target PPFD. We answer in the same format we ask buyers to use: fixture count, µmol/s per bed, a PPFD map at your height, wall watts, lead time with milestone dates, and the part numbers for the driver and the bar.
