Acceptance & QC

Incoming inspection: seven checks that catch a substituted batch

What to test when luminaires arrive rather than after they are installed — sampling across the delivery, the electrical, photometric, colour, flicker, thermal and mechanical checks, and a blank worksheet to record them against.

Written by Long-Term New Media Limited · Last reviewed August 2026

The risk that incoming inspection exists to catch is not a manufacturing defect. It is substitution: the sample you approved and the product that arrived are not the same specification. It happens through honest supply chain disruption as often as through anything worse, and it is invisible until the fittings are on the ceiling — at which point removing them costs more than the order.

Retain the approved sample, sealed

Before anything else. When a sample is approved, one unit is set aside, labelled with the date and the order reference, and not used. It is the reference against which the delivery is compared, and without it every subsequent conversation is one person's recollection against another's.

Photograph the sample's label, driver label and any internal marking at the same time. Driver substitution is the most common single change, and the driver label is inside the fitting where nobody looks.

Sample across the delivery, not from it

Taking five fittings out of the top carton tests the top carton. Batch variation is what you are looking for, so the sample has to span cartons, pallet positions and — if the labels show it — production dates.

For a defined sampling plan, ISO 2859-1 gives sample sizes and acceptance numbers against lot size and inspection level. Most lighting orders do not warrant formal AQL sampling, but the principle it encodes does: sample size scales with lot size, and the units are drawn at random rather than from wherever is convenient.

The seven checks

1. Documentation and marking

Model number, batch or production date, certification marks, and the driver model inside the fitting, all against the approved sample and the order. This takes two minutes and catches more problems than every other check combined. Confirm the photometric file supplied corresponds to the model actually delivered.

2. Input power and power factor

Measure with a true-RMS power meter, not by multiplying a voltage and current reading — LED drivers draw non-sinusoidal current and the naive product overstates real power. Compare against the datasheet and against the retained sample.

A power draw materially below specification usually means fewer or lower-current LEDs, which will show up as reduced output. Above specification means a different driver setting, and a thermal design that may not have been validated for it.

3. Relative light output

Absolute photometry needs an integrating sphere. Relative comparison does not, and relative comparison is what answers the question.

Build a jig: the fitting at a fixed distance and orientation from an illuminance meter, in a dark space, with everything clamped so geometry repeats exactly. Measure the retained sample, then each delivery unit, all after the same warm-up. You are looking for consistency between the sample and the delivery, and between delivery units. A 10% spread across the batch is a finding even if you cannot state the absolute lumens.

4. Colour

The wall test in colour consistency and SDCM — adjacent, same drive conditions, thermal equilibrium, viewed from the working position. Include the retained sample in the row. If a spectrometer is available, record u′v′ coordinates and compare against the specified SDCM step; if not, the eye at a boundary is more sensitive than most people expect, and a fixed-white-balance photograph records the evidence.

5. Flicker screening

Slow-motion video against the retained sample, at full output and at the dim levels the installation will use. This is a screening test and cannot produce a compliance figure — see the flicker note for why. What it does reliably is reveal a driver substitution: if the delivery shows banding the sample did not, something changed inside.

6. Thermal

Run a unit for at least an hour in a representative mounting condition — in the ceiling type it will actually occupy, not in free air — and measure case temperature at the T_c point marked on the driver. Compare against the datasheet limit.

This is the check that predicts the failures nobody attributes to inspection, because driver life falls sharply with temperature and lumen maintenance projections assume a case temperature that a hot installation does not honour. See lifetime claims for how the interpolation depends on it.

7. Mechanical and ingress

Gland sizes against the actual cable, blanking plugs present for unused entries, gasket seated and undamaged, diffuser material marking matching the specified IK rating, breathable vent present where the specification called for one. See IP and IK ratings for what each of those protects against.

Worksheet

Blank by design — this is a template to print and fill in per delivery, not a results table. Record the retained sample in the first column so every reading has its reference alongside it.

What to do with a finding

The value of inspecting on arrival rather than after installation is entirely about leverage. Before the fittings are on the ceiling, the options are open: reject the delivery, accept with a price adjustment, accept the batch for a non-critical area and reorder for the visible one. After installation the only realistic option is to live with it.

Which means the commercial terms have to allow it. Two clauses make the difference:

• Right of rejection on incoming inspection, with a stated inspection window from delivery, so that acceptance is not deemed to have occurred on receipt.
• No substitution of driver or LED package without written approval, naming both. "Or equivalent" in a specification is what makes substitution permissible in the first place.