This study intends to develop heuristic strategies in learning courses based on statistical data analysis metacognitive approach. This study was a quasi-experimental study with 2 x 3 factorial design. Techniques of data analysis, namely: balance test, test requirements analysis (test for normality and homogeneity test), test research hypotheses using two way analysis of variance with unequal cells. Based on the research results, it was concluded that that: 1) Students are provided with the learning of metacognitive have problem-solving ability is better than the conventional learning; 2) Students who have a high learning independence have problem solving skills better than students who have low learning independence, and students who have moderate learning independence as well as problem solving skills of students who have learning independence high and low; 3) Based on the categories of metacognitive and conventional learning, students who have a higher learning independence acquire mathematical problem solving ability is better than the lower learning independence, while the mathematical problem solving ability of students who have moderate learning independence as well as low learning independence; 4) Based on the category of high-independent learning, students who applied to acquire metacognitive learning mathematical problem solving abilities better than that applied by conventional learning. Based category learning independence medium or low, there is no difference mathematical problem solving ability of students who applied to the metacognitive learning and conventional learning.