Like with any technology, there are limitations. It does not know how to evaluate the canter gait, for example. And it requires purchasing software and the system. I don’t know if there are subscription fees or a one time thing, but it’s another expense for smaller practices.
I’ve used it a few times. Honestly, almost every time it said the horse was sound (WNL), albeit potentially with some asymmetries. But after finding the issue (particularly if the issue is in the axial skeleton) and treating it, running the program again, the horse became more sound with less noise on the graph. And then during a setback, the main problem leg was finally isolated to being consistently mildly off (assuming most of the compensation patterns had calmed down with the initial treatment and rehab).
So, on the one hand, it could be good at picking up subtle lameness. It could also be good to confirm responses to flexions or blocks by removing some human bias. But the algorithm does tolerate a range of asymmetry that may or may not be clinical and so is just as hard to interpret as watching the horse without the device.