Local workload arithmetic is a synthetic benchmark
Read the explanation
The default radiology labeled preset assigns one hundred forty two items per minute, a twelve minute hybrid backlog and three hundred forty minute manual backlog. No patient cases, routing tasks or model requests are processed. A generic throughput expression at the same surge and confidence would be one hundred fifty times one point zero one two, or one hundred fifty one point eight, rounded one hundred fifty two. The fixture override lowers that by ten. At two pixels per item rate the displayed bar is two hundred eighty four, generic three hundred four and difference twenty. Assigned medical labels and accuracy are not clinical evidence or usable triage guidance. The exported default comparison divides assigned AI rate one hundred forty two by an assumed manual rate twenty five, rounding to five point seven times. Dividing the assigned manual delay three hundred forty by hybrid delay twelve gives twenty eight point three, rounded twenty eight times. At ten pixels per ratio the gain bar measures fifty seven and backlog ratio two hundred eighty, while one times baseline measures ten. Neither ratio comes from a measured benchmark. Batch size is copied into configuration but is absent from the metric formulas. Animation packets are browser decorations, not actual completed items, and the missing D three graph prevents the initial interval from starting. The fatigue preset sets surge three point two, confidence point ninety four and a twelve hour shift. Assigned manual delay rounds one hundred thirty six times three point two times one point five, or six hundred fifty two point eight, to six hundred fifty three. Hybrid delay rounds twelve times one point two eight times point six two four to ten. Their rounded ratio is sixty five. At one half pixel per minute manual delay is three hundred twenty six point five, hybrid delay five and difference six hundred forty three becomes three hundred twenty one point five. These formulas do not prove fraud detection, labor safety or actual operational savings. Native presets, pause, reset and actual JSON exports remain available before graph errors; original behavior and limitations are preserved.