This study aims at evaluating the error on the horizontal components of the mean-velocity and of the Reynolds stress measurements due to a misorientation of an Acoustic Doppler Velocimeter (ADV) in the horizontal plane. The misorientation of the ADV is represented by a misalignment angle, which is the angle between the ADV coordinate frame and the reference frame. The error due to the misorientation is first analytically determined. Thereafter, using experiments in a compound channel, the error is evaluated based on an angle of 2.5° representing the maximum possible angle of misalignment by human eyes adjustment. While the maximal error does not exceed 5% for the longitudinal mean velocity and 14% for the normal Reynolds stresses, it can be larger than 100% for the lateral mean velocity and largely greater than 100% for the Reynolds shear stress. Such a degree of error appears to be similar to those worked out for classical flows from the literature with a mean velocity gradient mainly occurring in the horizontal plane (mixing layers, plane jets) or in the vertical plane (boundary layers). This error is finally compared with the uncertainty due to the various sources of noise that pertubate the turbulence measurement. This study reveals that they both are comparable, therefore indicating that the error due to the probe misorientation should not be neglected, as it is too often the case.