Struggling students are often unaware or unwilling to make contact for help and tutors do not always have time to observe or offer the help. In a distance learning university, it is harder to see what is happening and who may be getting into trouble. Suggested solution from the OU was to use predictive data to identify who may benefit from an intervention based on VLE engagement information, assessment information and demographic information in combination. The OU predictive model was tested on 3 modules.
VLE engagement information was related to their click history. The key issue was not the number of clicks that each student made, but a change in click behaviour. Assessment scores needed to remain above a pass threshold and they looked for any drop in assessment performance on subsequent assessments. Predictions were good over the 3 modules tested, particularly using assessment + changes in VLE engagement in combination. Demographic information important at the start as is VLE data. Later in a module, assessment data becomes the stronger predictor.