Background and aim Parkinsonian disorders are hallmarked by gait and balance impairments. Atypical parkinsonian disorders (APD) develop postural instability with falls and gait disorders early on. Sensor-based gait recordings provide objective data in hospital and everyday life, improving mobility assessment accuracy. However, the impact of duration and distance of instrumented assessments on construct validity remains unclear. This exploratory study aims to evaluate the construct validity of gait assessments compared with clinical, functional, and patient-reported scores. Methods The multi-centered Mobility_APP study recruited 43 PD and 49 APD patients. Among others, the Berg Balance Scale (BBS) and the postural instability and gait difficulty score (PIGD) were collected. Sensor-based gait parameters were captured during standardized 2 × 10 m and 2-min walk tests (2MWT) in the hospital and for 1 day of physical activity monitoring (PAM) at home. PAM was categorized by short (10–30 s), medium (30–60 s), and long (≥ 60 s) walking bouts (WB). Spearman correlations were applied to investigate associations between scores. Results Mean gait velocity (GV) and stride length correlated more strongly with functional, clinical, and patient-reported scores in 2MWT than in 2 × 10 m. Additionally, the GV variability in the 2MWT correlated with BBS and PIGD ( r = │0.3–0.7│), but was less prominent in 2 × 10 m ( r = │0.0–0.5│). In PAM, GV of long WB correlated more strongly with the PIGD ( r = │0.5–0.6│) than short WB ( r = │0.2–0.4│). Conclusion The 2MWT tended to show the highest construct validity. PAM offered complementary but weaker correlations, highlighting that PAM provides novel insights into daily life mobility of APD patients.