Abstract Background Tri-axial accelerometers are frequently deployed on terrestrial quadrupedal mammals using collars, because they are easy to fit and are thought to have minimal impact on the subject. Collar-attached devices are not fixed to the body and can move independently of the body. This may result in inaccurate measures of acceleration, reducing the accuracy of measured body movement. We determined the effect of collar size and collar weight on acceleration measured by a collar-mounted accelerometer on a quadruped mammal. The aim was to suggest best practice for sizes and weights of collars on which to deploy tri-axial accelerometers. Using pygmy goats, Capra aegagrus hircus, which were trained to walk at different speeds (0.8–3.0 km/h) on a treadmill, we measured body acceleration using a collar-mounted tri-axial accelerometer, with different collar sizes (individual neck circumference + 1 cm to + 9 cm) and collar weight (0.4% to 1.2% of individual weight). Results There was a significant effect of collar size, collar weight and walking speed on measured acceleration. Measured acceleration was less accurate and more variable when collars were looser and heavier. To measure body acceleration more accurately, we found that collar size should be within 5 cm or 16% of an individual’s neck circumference when it was heavy (up to 1.2% of animal’s body weight) or within 7 cm (33%) of neck circumference if the collar was light (up to 0.6% of animal body weight). Conclusion We suggest that not only reporting collar size and weight for welfare purposes, but it is also important to consider these aspects for scientific rigour, to ensure data are collected as accurately as possible. We provide guidelines for researchers fitting collar-attached devices to ensure a higher degree of accuracy of recorded body acceleration.