Higher education institutions (HEIs) increasingly employ learning analytics (LA) - popularly defined as the measurement, collection, analysis, and reporting of data about learners to understand and optimise learning environments. However, while LA promises benefits, its implementation often overlooks the perspectives and concerns of students, who are the primary data subjects. This doctoral research explored undergraduate student perceptions of LA during an educational intervention in which they participated and which focused on developing their critical data literacy (CDL). This qualitative case study highlights concerns from the students involved regarding agency, fairness, transparency, and trust within institutional data practices. Key concerns which arose included the status of students within LA as ‘data subjects’, not data owners or interpreters (Prinsloo, 2022). Students often lack awareness regarding the extent of data collection, its subsequent use, and the potential implications for their educational journey and future opportunities. This passive positioning limits their ability to critically engage with, question, or influence the LA systems that can monitor and potentially shape their learning experiences. LA, therefore, risk operating with insufficient consideration for student autonomy and critical consent, creating limitations related to understanding, control, and perceived equity. This qualitative case study suggests that developing CDL - an awareness and critical understanding of data collection, use, and potential misuse - is crucial for empowering students to navigate these increasingly datafied educational landscapes. This research employed a qualitative case study approach involving 18 undergraduate education students at DCU. Participants engaged in a short course intervention designed to foster CDL. A key component involved providing each student with a personalised report visualizing their own VLE activity data, conceptualised as a ‘reflective tool’ or “an object to think with” (Papert, 1980). This direct engagement with their own LA data was coupled with exploratory seminars discussing datafication, data ownership, agency, and criticality. Data regarding student experiences of the educational intervention and perceptions of LA limitations were gathered through semi-structured interviews and analysed using reflexive thematic analysis (Braun & Clarke, 2022). Among the findings were revealed limitations of LA from the student viewpoint, primarily centered around themes of ‘Agency’ and ‘Fairness’. Within ‘Agency’, students expressed feelings associated with being passive ‘consumers’ of data processes and instances of 'digital resignation' - a sense of powerlessness regarding data collection - highlighting a fundamental limitation in student control and autonomy within LA ecosystems. While the intervention fostered increased ‘awareness’ and prompted ‘changes in practice’, these developments stemmed from an initial deficit in agency and understanding. The theme of ‘Fairness’ encapsulated multiple perceived limitations. Students began ‘questioning why’ certain data were collected and how they were used, probing the ethical dimensions of LA practices. A critical limitation identified was the lack of ‘data transparency’, with students expressing a desire for clearer communication regarding institutional data handling. This opacity contributed to feelings of ‘trust and distrust’ towards the institution's use of their data. Furthermore, students emphasised the need for more ‘human-centred’ approaches, suggesting an important role for their lecturer in interpreting LA data around engagement and prediction, as they have a pre-existing relationship with the learner and can provide vital personal context which data alone cannot. This research illuminates critical limitations inherent in non-human-centered LA implementations. From the perspectives of the students, these limitations prominently include a lack of personal agency and control, concerns about fairness and algorithmic bias, lack of data transparency leading to trust issues, and a general feeling of being subjects within opaque systems rather than empowered participants. The findings suggest that for LA to be truly effective and ethical, its design and deployment must move beyond technical optimisation to foster genuine student empowerment and privilege human interpretation of LA.