Running two separate two-stage randomized trials, we implement and test three variants of a small group instruction intervention aimed at improving mathematics competencies for the 20% lowest achievers in mathematics in grades 2 and 8 in Danish public schools. We calculate immediate impacts on math competencies as well as impacts in the medium term and conduct cost-effectiveness analyses of the different intervention arms. The context is the Scandinavian welfare state with an already high-quality public school system. Nevertheless, we find very large positive and cost-effective impacts of several intervention variants, some of which persist in the medium term. In particular, interventions in grade 2 tend to persist, while eighth-grade interventions tend to fade out in the medium term.