One grader. Every shift. No more grade drift.

Against a fixed reference scorecard, Orion achieves ±0.3-grade variance on a six-grade scale across iPhone, Samsung Galaxy and major Android lines, and ±0.5 on laptops and Macs. Human graders typically run ±2 grades depending on shift, fatigue and SKU familiarity. The number that matters: customer dispute rate drops 60–75% after operators move to AI grading.

Yes. Laptop and Mac grading uses a different capture flow (lid, palm rest, base, ports) and a different defect taxonomy (hinge wear, palm-rest pitting, port damage, screen dead pixels visible without disassembly), but the underlying model architecture and replayability are identical.

Yes. Orion's universal grade is a six-point scale mapped at sale-time to whichever channel grade ladder the listing targets. The mapping is per-channel and per-SKU, so a Galaxy A53 sold on Swappa maps differently than the same device sold on BackMarket. You can also publish your own internal ladder.

Override is a one-tap action. The override, reason and operator are recorded; the photos are flagged for the next training cycle. Over time the model learns your inventory's edge cases — fitness for purpose improves, not just abstract accuracy.

Phonecheck, Piceasoft, NSYS, FutureDial and similar tools rely on a human grader following a checklist. The result is fast but inconsistent and impossible to audit after the fact. Orion's grading is consistent across operators and across shifts, and every grade is replayable from the original capture.