Background: Our group recently demonstrated no difference in weight loss at 6-months between personalized and one-size-fits-all low-fat diets. This study aimed to determine subgroup characteristics that would predict weight loss success. Methods: Data were analyzed on adults with abnormal glucose metabolism and obesity from a 6-month randomized clinical trial. Participants were randomized to a low-fat diet (<25% energy intake; Standardized) or a personalized diet that predicts post-prandial glycemic response to foods using a machine learning algorithm (Personalized). Both arms received identical behavioral counseling and were instructed to self-monitor dietary intake using a smartphone app. The gradient boosting machines method was used to determine which baseline variables (age, gender, race, ethnicity, income, education, BMI, metformin use, self-efficacy, glycemic variability) can predict successful weight loss (>5%) at 6-months in each study arm separately by repeated 5-fold cross validation. Results: A total of 204 adults were randomized 199 participants (Personalized: n=102 vs. Standardized: n=97) contributing data (mean standard deviation: age, 58 11 y; 67% female; body mass index, 33.9 4.8). Standardized and Personalized arms achieved 0.59 and 0.62 average AUCs, respectively. In the Standardized arm, participants who had a higher BMI (p=0.01), higher self-efficacy (p=0.006), were older (p=0.001) and were not on metformin (p=0.003) favored weight loss. In the Personalized arm, participants who had higher self-efficacy (p=0.0003) and were older (p=0.0001) favored weight loss. Conclusions: Baseline self-efficacy and age are significant contributors to weight loss success in a behavioral weight loss intervention.