Background Weight-loss programmes often achieve short-term success though subsequent weight regain is common. The ability to identify predictive factors of regain early in the weight maintenance phase is crucial. Objective To investigate the associations between short-term weight variability and long-term weight outcomes in individuals engaged in a weight-loss maintenance intervention. Methods The study was a secondary analysis from The NoHoW trial, an 18-month weight maintenance intervention in individuals who recently lost >= 5% body weight. Eligible participants (n = 715, 64% women, BMI = 29.2 (SD 5.0) kg/m(2), age = 45.8 (SD 11.5) years) provided body-weight data by smart scale (Fitbit Aria 2) over 18 months. Variability in body weight was calculated by linear and non-linear methods over the first 6, 9 and 12 weeks. These estimates were used to predict percentage weight change at 6, 12, and 18 months using both crude and adjusted multiple linear regression models. Results Greater non-linear weight variability over the first 6, 9 and 12 weeks was associated with increased subsequent weight in all comparisons; as was greater linear weight variability measured over 12 weeks (up to AdjR(2) = 4.7%). Following adjustment, 6-week weight variability did not predict weight change in any model, though greater 9-week weight variability by non-linear methods was associated with increased body-weight change at 12 ( increment AdjR(2) = 1.2%) and 18 months ( increment AdjR(2) = 1.3%) and by linear methods at 18 months ( increment AdjR(2) = 1.1%). Greater non-linear weight variability measured over 12 weeks was associated with increased weight at 12 ( increment AdjR(2) = 1.4%) and 18 ( increment AdjR(2) = 2.2%) months; and 12-week linear variability was associated with increased weight at 12 ( increment AdjR(2) = 2.1%) and 18 ( increment AdjR(2) = 3.6%) months. Conclusion Body-weight variability over the first 9 and 12 weeks of a weight-loss maintenance intervention weakly predicted increased weight at 12 and 18 months. These results suggest a potentially important role in continuously measuring body weight and estimating weight variability.