© 2019, Springer-Verlag GmbH Germany, part of Springer Nature. Purpose: With accelerometry, the utility to detect changes in physical activity are predicated on the assumption that walking energetics and gait mechanics do not change. The present work examined associations between changes (∆) in walking energetics, exercise self-efficacy, and several accelerometer-derived metrics. Methods: Secondary analyses were performed among a sub-sample (n = 29) of breast cancer survivors participating in a larger randomized trial. During 4 min of treadmill walking (0.89 m s−1, 0% grade), indirect calorimetry quantified steady-state energy expenditure (EE), wherein, participants were fitted with a heart rate monitor and hip-worn triaxial accelerometer. Exercise self-efficacy was measured using a 9-item questionnaire, while vector magnitude (VM) and individual planes (e.g., mediolateral, vertical, and anteroposterior) of the movement were extracted for data analyses. Evaluations were made at baseline and after 3 months. Results: From baseline to 3 months, the energetic cost of walking (kcals min−1) significantly decreased by an average of − 5.1% (p = 0.001; d = 0.46). Conversely, VM significantly increased (p = 0.007; d = 0.53), exclusively due to greater vertical accelerations (acc) (+ 5.7 ± 7.8 acc; p = 0.001; d = 0.69). Changes in vertical accelerations were inversely and positively associated with ∆walking EE (r = − 0.37; p = 0.047) and ∆exercise self-efficacy (r = 0.39; p = 0.034), respectively. Conclusion: Hip-worn accelerometers do not appear well-suited to correctly detect changes in ease of walking as evidenced by reduced energetic cost. Further research should determine if a divergence between measured EE and vertical accelerations could contribute to erroneous inferences in free-living physical activity.