For learners from disadvantaged backgrounds with limited science fair-related skills, knowledge, and community support, investigating the advisability of various ways of arriving at a science fair project topic is needed. This mixed-methods case study with an ex post facto criterion-group design uses questionnaire data and science fair awards regarding the engagement of 23 grade 9-10 learners in a year-long university-led intervention. Data analysis was conducted for four post-hoc determined topic-origin groups of these learners' projects: internet, self, adult, and adult after the learner had displayed competence. Descriptive statistics and thematic analysis were used to determine whether each topic-origin group met the advisability criteria regarding learners' perceptions of value and facilitators' perceptions of feasibility. Each learner who was provided with a topic by an adult without first having displayed competence through designing a project of their choosing required unfeasibly high levels of facilitator input to remain in the programme. Generally, the greatest success was observed for the learners to whom an adult gave a topic after displaying competence. These findings and how the intervention was conducted are explained in terms of the four paradigms of learning and instructional design. A framework (Instruct-Expose-Explore-Formalise) for deriving science fair topics is suggested to enhance the efficiency and effectiveness of similar interventions.