Two panel lights, both marked Ra80, switched on side by side. One reads warm white, the other goes slightly green. The buyer assumes the colour temperature was mislabelled. In fact a different parameter was left out of the specification.
Start with CRI. Ra is the average of the first eight of fifteen test colour samples. R9 is saturated red, and it sits outside the Ra calculation. Cheap LEDs protect efficacy by sacrificing red content, so Ra can read 80 while R9 sits in single digits.
Where does that show? Fresh food, meat, apparel, timber veneer. Reds and skin tones go grey and flat. For retail lighting, write the R9 floor into the specification, R9 above 50 rather than just Ra80.
Now colour tolerance. SDCM describes how far the chromaticity sits within a MacAdam ellipse: 3-step, 5-step, 7-step. Lower is tighter. If a batch mixes LEDs from different bins, a corridor run will show visibly uneven patches.
This is the usual explanation for two Ra80 panels that do not match. One uses a 3-step bin, the other a 7-step, and every printed number is identical.
Only two ways to control it. Require a single bin within the batch, and require the chromaticity test report so you can see the actual coordinate spread.
Worth adding: CRI itself is being displaced by TM-30, which reports Rf for fidelity and Rg for gamut. Two numbers describe real appearance better than a single Ra, and some European and American design specifications already cite Rf. If your customer is a design practice, raising TM-30 unprompted earns credibility.
Practical version: for retail, apparel and food service, write three lines, Ra above 90, R9 above 50, SDCM 3 or lower. That adds roughly ten percent to the quotation and removes most of the rework and complaints.
For warehouses, car parks and industrial lighting, Ra80 with SDCM 5 or lower is enough. There is no reason to pay for colour accuracy nobody will look at.