My mutt has been tested with WP, Embark, and DA.
~ 50% Australian Cattle Dog
~ 25% Boxer
~ 12.5% Australian Shepherd (asterisked, indicating uncertainty as to whether the predicted % is really accurate)
~ 12.5% too mixed to call/unknown, with Sighthound and Herding groups listed as possibilities
15.8% Australian Shepherd
15.1% Border Collie
9.1% too mixed to call/unknown, with Sheltie specifically predicted to be in the mix
6.2% Border Collie
2.9% Australian Shepherd
In my dog’s case, I’m guessing(?) the main reason for the differences may be that she really is roughly 75% assorted herders (as WP and Embark found), but that without both a very large number of markers and a very robust reference panel, it becomes difficult to accurately assign all her non-Boxer DNA to one-or-another herding breed, since all the herding breeds detected in her to date are thought to be members of the extended collie/UK sheepdog family (Parker et al. 2017), and therefore pretty closely related. So I’ll be interested to see how much of her current 26.7% unknown winds up getting reassigned to already-called herding breeds. Would be just as ready to believe there’s a splash of something that hasn’t been called yet knocking around in there, though! The ACD and Boxer have been pretty obvious in her from the beginning (nip nip nip, boing boing boing) so I’m not surprised those two findings have held fairly constant across all three tests.