This study focuses on investigating quantitative methods that can be used to improve academic achievement at course level, with a specific emphasis on online learning scenarios. The focus on improvement implies concentrating on techniques with a high interpretative capacity, where understanding the process over time plays a fundamental role. Two methodologies have demonstrated significant utility: survival analysis to comprehend and reduce dropout, and educational process mining to analyse learning paths and assess deviations that may contribute to low performance. An innovative methodological proposal, linked to the latter technique, stands out for the modelling of learning paths based on the use of skeletons. The results, published in eight articles, represent a transfer of techniques more commonly associated with other disciplines than the field of learning analytics. They provide a methodological proposal to identify vulnerable groups, quantify the impact of risk factors, assess adherence to a learning path, or detect divergences from it. Consequently, these findings constitute tools of high interest in the design of academic interventions.