Fail, flip, fix, feed: impact on engagement and learning outcomes

Background Flipped classrooms and productive failure foster active learning by reversing the traditional sequence of instruction and problem-solving. Despite strong evidence for each approach in secondary mathematics, their combined implementation remains untested. Aims Building on a previous framework (fail, flip, fix, feed), this study evaluates two variants that differ in whether collaboration occurred during problem solving or instruction, assessing their effects on learning and engagement. Sample The sample consisted of 221 students (ages 12–16) from 12 secondary school classes in Switzerland. Methods The study is based on a randomized control group design with two performance and four engagement measurements. Both test batteries are commonly used in productive failure research. Results Learning outcomes were statistically equivalent across both conditions after the first two phases (TOST; equivalence bounds ±0.5 SD). After the full intervention, three-sided equivalence testing yielded inconclusive results, with no reliable performance differences between conditions. Engagement levels did not differ between conditions activities, p = .449. Engagement predicted performance on both posttests, accounting for up to 18% of the variance: β = 0.760, p = .003 on Day 1, and β = 0.860, p = .001 on Day 2. Mental effort showed no reliable effect on performance: β std. = −0.039, p = .782. Conclusions The findings highlight the need for learning designs that foster student self-reported engagement. However, contrary to the original goals of flipped classroom approaches, the presence and guidance of the teacher did not result in higher student engagement compared to independent study time. The results further suggest that short-term outcomes in 4F-type designs may not be strongly determined by the phase placement of collaboration alone.