Why 2700K Is Not Automatically Warm

Two sources can both be 2700K and look nothing alike. Correlated colour temperature is a single number fitted to a two-dimensional chromaticity position, so it says nothing about how far the source sits above or below the blackbody locus. A 2700K package two Duv steps green will make skin and warm stone look grey and slightly ill, and no amount of dimming will rescue it. The number on the box was never the specification we actually meant.

The second half of the problem is the surface. At 2700K a cream plaster reads warm and a cool grey paint reads dead, because the source is only supplying what the pigment can reflect. On the Port Aroha portico the calculations said 3000K and the mock-up said 2700K, and the mock-up was right, because the plaster pigment needed the additional red content to look like itself. That is a two-hour test on site and it settles an argument that can otherwise run for months.

What we specify now is a triple: nominal CCT, a Duv tolerance of plus or minus 0.003, and R9 with a floor rather than a headline CRI. Consistency matters as much as the target, so we also call a three-step MacAdam binning limit and hold a small stock of the original batch for replacements. A wall of nominally identical fittings from two production runs is one of the most common and most avoidable defects on a finished job.