E-learning research is plenty of difficulties, as also research in education is. Usually, the high number of features involved in e-learning processes complicates and masks the identification and isolation of the factors which cause the expected benefits, when they exist. At the same time, a bunch of threats are ready to weaken the validity of the research, for example, disregard of previous research, use of small samples, absence of randomization in the assignment to groups, ineffective designs, lack of objectivity in the measuring process, poor descriptions of the research in publications (which implies few possibilities of replication), wrong statistical procedures, inappropriate inference of results, etc. All of these obstacles accumulate and are carried along the whole research, resulting in low quality studies or irrelevant ones. This theoretical paper suggests a roadmap in order to face the most common problems in e-learning research. The roadmap informs about some cautions which must be considered at each stage of the research and recommendations to increase the validity and reproducibility of results. The roadmap and conclusions included in this paper have been obtained from our experience in educational and e-learning research, also from our long path as reviewers in key journals of these fields, and from readings of significant research handbooks. This is not a strict guide but a set of milestones on which it is necessary to stop and reflect.