Learning analytics involve the measurement, collection, analysis, and reporting of data about learners and their contexts, in order to understand and optimize learning and the environments in which it occurs. Since emerging as a distinct field in 2011, learning analytics has grown rapidly, and early adopters around the world are developing and deploying these new tools. This paper reports on a study that investigated how the field is likely to develop by 2025 in order to recommend actions to those concerned with the implementation of learning analytics. The study used a Policy Delphi approach, presenting a range of future scenarios to international experts in the field and asking for responses related to the desirability and feasibility of these scenarios, as well as actions that would be required. Responses were received from 103 people from 21 countries. Responses were coded thematically, inter-rater reliability was checked using Cohen's kappa coefficient, and data were recoded if kappa was below 0.6. The seven major themes identified within the data were power, pedagogy, validity, regulation, complexity, ethics, and affect. The paper considers in detail each of these themes and its implications for the implementation of learning analytics.