In the last two decades Dutch primary school students scored below expectation in international mathematics tests. An explanation for this may be that teachers fail to adequately assess their students' understanding of learning goals and provide timely feedback. To improve the teachers' formative assessment practice, researchers, curriculum experts and teachers worked together to develop a model for classroom formative assessment (CFA). In three pilot studies, six teachers from three different schools implemented the CFA-model and evaluated its feasibility together with the researchers by means of checklists. The CFA-model was primarily changed with regard to the assessment techniques. Teachers indicated that classroom management and preparation time were preconditions for an optimal implementation. Analysis of covariance was used to explore students' learning outcomes. The results showed that a correct implementation of the CFA-model might result in the enhancement of students' mathematical performance. The implications of the three pilots for the implementation of the CFA-model on a larger scale are discussed.