Enhancing the problem-solving skills of learners belonging to socio-economically disadvantaged groups is a great concern in general and particularly in the Indian context. Problem-solving skills are the most important skills of the 21st century for all learners to acquire necessary competencies. The main purpose of this work is to investigate the effect of metacognitive brainstorming strategy, gender, and residential areas, and their various interactions on problem-solving skills in physics for ninth-grade learners who belong to socio-economically marginalised groups. The pre-test--post-test, quasi-experimental design was applied to 107 learners from two government-sponsored schools of the same nature located in Kolkata, West Bengal, India. The 55 learners in the experimental group were given intervention through a metacognitive brainstorming strategy, whereas the 52 learners in the control group were taught the same lessons by simply using the conventional method. Self-developed and standard problem-solving skills test with a reliability of 0.89 and validity of 0.94 were employed on them as pre-test and post-test instances. The correlated t-test, independent samples t-test, and 2 x 2 factorial design analysis of covariance were applied for data analysis through Statistical Package for Social Science. The results revealed that the brainstorming instructional strategy had a huge effect on problem-solving skills due to its large effect size value (2.23 > 1.2). Moreover, the problem-solving skills of learners in physics were found to be independent of gender and residential areas, and also of their interaction with the treatment strategy, when the pre-problem-solving skill was taken as a covariate.