The Lighting Upgrade That Measured Better and Felt Worse

Why this matters

The retrofit closed out in March. The survey said the sales floor went from 41 fc to 52 fc. In April the store manager wrote that the aisles feel like a stockroom and customers cannot read the shelf. Nobody is lying and nobody is imagining it. By the time this lands on your desk the old fixtures are in a skip, the crew has moved on, and the only witnesses left are two survey sheets, a submittal and a commissioning record. This is a reconstruction from those documents, and the two things it turns on are that the sheets were never on the same basis and that the quantity the shoppers were responding to appears on neither of them.

What was in the file, and what was not

Five documents survived.

• The 2019 survey. One line: "Average 41 fc, main aisles." Twelve readings, no plane, no height, no basis, timestamped 14:20 in June.
• The 2024 survey. Ninety readings on a nine-point grid per bay, horizontal at 30 in above finished floor, initial basis, taken at 22:00 with the store closed. Average 52 fc.
• The submittal. Ordering code plus an absolute photometric report of the finished luminaire, described as a narrow-distribution unit.
• The commissioning sheet. Dimming verified at 100 percent, daylight sensors not installed on this circuit.
• The complaint log. Nine entries in six weeks, seven of them naming the shelf rather than the aisle.

The commissioning sheet kills the easiest explanation immediately. A control system holding output down produces exactly this complaint and it is the first thing to eliminate; here it is documented at full output, so the light being made is the light that was bought. What the lighting controls card owns is how that setting is verified, and it was.

That leaves two survey numbers that everybody has been comparing since March, and they were never comparable.

Correction one: the old reading contained daylight

The building has clerestory glazing along the ridge. The 2019 read was taken at 14:20 in June with the store trading, so its 41 fc is electric light plus daylight. The 2024 read was taken at 22:00.

The correction is measured on a proxy day rather than estimated from a table, which is better than a guess and short of a measurement of the 2019 condition: same hour, same season, a different sky, and clerestory glazing five years older. On a clear afternoon at the same hour and season, with the sales floor circuit opened at its panel, the daylight-only contribution read 9 fc horizontal at 30 in above finished floor. Note what that subtraction assumes: the 2019 sheet records no plane and no height, so putting the two on one plane is an assumption rather than a correction. Carry it as an assumption, because a 2019 reading taken at grade rather than at 30 in would move the 41 fc figure more than either correction below. So the 2019 electric-only equivalent is 41 - 9 = 32 fc.

Note which direction that runs. It makes the old number smaller and the retrofit look better, and it is the correction a contractor defending the job would reach for first.

Correction two: the new reading was day-one

The 2024 average is initial: lenses out of the carton, zero hours. The 2019 fixtures were six years into a cleaning interval that the maintenance log shows was never performed, so 41 fc, and the 32 fc under it, is a maintained figure.

An initial figure and a maintained figure cannot be compared, and the fix is to re-base one of them, once. The light loss factor card owns that operation; here the new installation's light loss factor is 0.82 from the design report, so 52 x 0.82 = 42.6 fc maintained.

That correction runs the other way. It makes the new number smaller.

Two corrections, opposite directions, and why one alone is worse than none

Raw: 41 to 52 fc is a 27 percent gain. Corrected on both sides: 32 to 42.6 fc is a 33 percent gain. The level genuinely rose, and it rose by more than the raw sheets said, not less.

Doing only the daylight correction gives 32 to 52, a 63 percent gain, which is the number that would have gone in a letter. Doing only the basis correction gives 41 to 42.6, a 4 percent gain, which is the number that would have gone in a complaint. Both are defensible arithmetic on a half-corrected pair, and both are wrong, because a comparator only counts as corrected when every correction has been applied to both sides.

So the complaint is not a level complaint. Something else moved.

The quantity nobody wrote down

Both surveys measured horizontal illuminance at 30 in above finished floor. Seven of the nine complaints name the shelf. A shelf face is vertical, and the light landing on it is vertical illuminance, which is a different quantity measured on a different plane with the sensor turned through ninety degrees. No sheet in the file carries one.

A fresh read on the sales floor, sensor vertical at 5 ft above finished floor facing the shelf face, mid-bay between two luminaires, gives 11 fc initial against the 52 fc initial horizontal on the same visit.

There is no matching old number, because nobody took one, and that gap is the finding rather than a dead end. The comparator came from a stockroom aisle that was never retrofitted: same mounting height, same aisle width, original luminaire type, still dirty. It reads 30 fc horizontal at 30 in above finished floor and 19 fc vertical at 5 ft above finished floor, both maintained.

Why a ratio survives a basis the levels do not

Those two spaces are on different bases and their levels may not be compared. Their internal ratios may be, and the reason is worth having.

The vertical-to-horizontal ratio is two readings taken with one meter, at one visit, on one set of fixtures in one maintenance state. Any fixed multiplicative error the instrument carries - its calibration factor, its cosine response at normal incidence, its spectral correction for that source - multiplies both readings and divides straight out of the quotient. The maintenance state divides out the same way, because dirt on the lens scales both planes together. What is left is a property of the distribution, which is exactly what changed.

• Old bay: 19 / 30 = 0.63
• New sales floor: 11 / 52 = 0.21

A factor of 3.0 on the plane the shoppers are looking at. Put the new floor on a maintained basis and its vertical illuminance is 42.6 x 0.21 = 8.9 fc maintained, against 19 fc maintained in the surviving old bay. Less than half the light on the shelf face while the horizontal number on the sheet went up a third.

The distribution explains it without any mystery. The old lensed unit threw a wide pattern with a large share of its flux leaving at high angles, which is what lands on a vertical surface. The submittal's narrow-distribution replacement puts that flux on the floor, where the meter was.

Raising the level is the wrong lever, and the arithmetic says so. To recover 19 fc on the shelf face at the installed 0.21 ratio you need 19 / 0.21 = 90 fc maintained horizontal, which is 90 / 42.6 = 2.1 times what is installed. That buys a floor bright enough to become its own problem, and what a level increase does and does not fix at a fixed distribution is the glare card's subject.

What was changed, and how it was confirmed

The rows were re-optic'd rather than re-lamped: asymmetric aisle-lighting optics fitted to the existing housings, redirecting flux off the floor and onto the rack face. Fitting them means reaching the mounting height from a lift rated for it with the fall protection the equipment requires under 29 CFR 1910 Subpart D in general industry or 29 CFR 1926 Subpart M in construction, and opening each housing is electrical work: de-energize at the panel, lock and tag, and prove dead live-dead-live per NFPA 70E-2021, 120.5, in the edition the employer's electrical safety program adopts, under 29 CFR 1910.333(b)(2).

Confirmation was written to be repeatable rather than to be flattering: same aisles, same mid-bay positions, same meter, store closed, dimming at 100 percent, both planes read on the same visit.

• Horizontal at 30 in above finished floor: 44 fc initial, down from 52.
• Vertical at 5 ft above finished floor: 21 fc initial, up from 11.
• Ratio: 21 / 44 = 0.48, against 0.21 before and 0.63 in the old bay.
• On a maintained basis at the same 0.82: horizontal 44 x 0.82 = 36.1 fc, vertical 21 x 0.82 = 17.2 fc, against the old bay's 19 fc vertical maintained.

So the shelf face came back to 17.2 / 19 = 91 percent of the old bay, and the horizontal average is still 36.1 against the corrected 32 fc the store had in 2019. Better on both planes than the original installation, and worse on the horizontal number than the retrofit's own closeout sheet. The complaint log went quiet inside two weeks.

The store's file now carries plane, height, basis, time of day, daylight state and control setting on every reading, which is what the warranty-record card specifies and what would have made this whole reconstruction a ten-minute comparison.

Checking your own figures

• Each correction applied to the side it belongs to, and both sides then on one basis. Old: 41 - 9 = 32 fc maintained, electric only, plane and height unrecorded. New: 52 x 0.82 = 42.64, reported as 42.6 fc maintained, horizontal at 30 in above finished floor. The daylight term is zero on the new side because that read was taken at 22:00, and the loss-factor term is already inside the old side because those fixtures were six years unclean. Gain 42.6 / 32 = 1.331, reported as 33 percent.
• The half-corrected numbers are printed as errors, not results. 52 / 32 = 1.625 and 42.6 / 41 = 1.039, quoted as the 63 percent and 4 percent a one-sided correction produces.
• The ratios come from same-visit pairs. 19 / 30 = 0.633 on one visit to the stockroom; 11 / 52 = 0.2115 on one visit to the sales floor; 21 / 44 = 0.477 on the confirmation visit. No ratio mixes two visits.
• The factor is the quotient of the two ratios. 0.633 / 0.2115 = 2.99, reported as 3.0.
• The level route is priced against its own basis. 19 / 0.21 = 90.5 fc maintained, and 90.5 / 42.6 = 2.12, reported as 2.1 times, with both terms maintained.
• Every illuminance carries plane, height and basis except the two that cannot. The 2019 41 fc and the 32 fc derived from it carry basis, maintained and electric only, but no plane and no height, because none was recorded, and that gap is named in the file inventory rather than papered over. Every other figure is horizontal at 30 in above finished floor or vertical at 5 ft above finished floor, each labelled initial or maintained where it is used.
• The complaint count matches the file. Nine entries, seven naming the shelf, and the shelf face is the plane the finding lands on.

References

• IES recommended practice for the space type, in the edition your specification names, which binds only through that specification and is where a vertical-illuminance recommendation for merchandised aisles lives; an IES recommended level is a design target, not a code minimum
• 29 CFR 1926.56(a) and its Table D-3, which set enforceable single-value construction illumination floors and carry no plane, no maintained basis and no uniformity ratio
• 29 CFR 1910.333(b)(2) with NFPA 70E-2021, 120.5, for de-energized work inside the housings; 29 CFR 1910 Subpart D and 29 CFR 1926 Subpart M for the work at height
• See related: How to Take a Light Level Reading Somebody Else Can Repeat; What a Light Loss Factor Is Accounting For; Why Uniformity Matters More Than Average Level; What Glare Is, and the Two Kinds That Matter