Learning analytics: definition, data types and challenges for training managers

In brief: Learning analytics refers to the full set of techniques for collecting, analysing and exploiting data generated by learners and training programmes. For digital learning managers, L&D directors and instructional designers, it is a strategic lever: detect disengagement before learners drop out, measure training ROI, and produce evidence of effectiveness exploitable during quality audits.

Definition of learning analytics

Learning analytics refers to the measurement, collection, analysis and interpretation of data about learners and their learning contexts, with the aim of optimising learning pathways and pedagogical environments.

Unlike a simple scored assessment, learning analytics looks at the learner's full behaviour within the system: time spent on each module, completion rates, error patterns, progression over time, attendance at synchronous sessions. It is a data-driven approach that does not reduce to final results but observes the journey.

The rise of digital learning and LMS platforms has considerably enriched the volume and granularity of available data. Every click, every answer, every log-in produces a signal that analytics tools can interpret. The question is no longer whether data exists, but how to exploit it relevantly and ethically.

Types of learning analytics

Three levels of analysis are generally distinguished, from the simplest to the most predictive:

• Descriptive analytics. They answer "what happened?": log-in rates, average time per module, assessment scores, drop-out rates by step. This is the most accessible entry point for training organisations.
• Predictive analytics. They answer "what will happen?": modelling drop-out risk, predicting final scores, identifying learners in difficulty before they quit. These tools often rely on deep learning and require sufficient data volumes to be reliable.
• Prescriptive analytics. They answer "what should be done?": automatic content recommendations, difficulty adjustment, suggestion of complementary resources. This is the basis of adaptive learning.

Data sources in training

Learning analytics draws on several complementary sources:

• LMS platforms. Completion data, scores, time on task, navigation paths. The xAPI and SCORM standards structure data exchange between tools.
• Attendance tracking tools. Presence data for synchronous sessions, punctuality, attendance across in-person and remote sessions. Often overlooked, yet essential for funding body declarations and audits.
• Questionnaires and assessments. Scores from formative and summative assessments, immediate and delayed satisfaction feedback, qualitative returns. Online questionnaires produce a qualitative data layer that LMS platforms alone cannot capture.
• Collaboration tools. Forum participation, collaborative spaces, virtual classroom engagement.

The richness of analytics depends directly on the quality and completeness of collection. A training programme that does not trace attendance, collect assessments or surface completion data produces impoverished analytics.

Tools and platforms

The training analytics tools market is mature. LMS platforms now integrate native dashboards (Moodle, 360Learning, Docebo, Talentsoft). In parallel, specialist solutions like Learning Locker or BI layers (PowerBI, Tableau) allow training data to be cross-referenced with other HR sources.

The xAPI (Tin Can) standard is particularly important: it allows any tool to "tell the story" of what happened during a training ("Learner X completed activity Y with score Z"), independently of the host platform. It is the protocol that makes interconnection between an LMS and a third-party tool like Edusign possible.

For training organisations subject to quality certification, quality managers seek ready-to-use analytics, exportable and timestamped. The ability to produce evidence of effectiveness on demand is now an operational requirement, not a bonus.

Limits and GDPR issues

Learning analytics raises serious regulatory and ethical issues that training managers cannot ignore:

• GDPR compliance. Training data is personal data. Its collection, processing and hosting must comply with European regulation: legal basis, limited retention period, learner access and correction rights, EU-based hosting.
• Transparency. Learners must be informed of the data collected and its purpose. A system collecting without their knowledge creates legal risk and erodes trust.
• Algorithmic bias. Predictive models can amplify existing biases (underrepresented profiles poorly modelled). Human oversight remains necessary, especially for high-impact decisions (selection, guidance).
• Source content quality. Prescriptive analytics does not improve poor pedagogical content: it amplifies it. Design work remains indispensable.

How Edusign produces exploitable learning analytics

Edusign is not an LMS, but a pedagogical and administrative management suite that produces data complementary to learning platforms:

• Remote and in-person attendance signing: timestamped and signed presence data, per learner and per session. Exportable for audits, integrable into training dashboards.
• Online questionnaires: immediate evaluations, satisfaction feedback, positioning questionnaires. Results consolidated by cohort, exploitable to demonstrate adaptation to learner needs (quality certification criterion).
• AI and automation: automatic reminders for non-connected learners, alerts on recurring absences, detection of pathways running late before learners disengage.

The goal: training managers have a complete view of their cohorts, without having to manually collect data scattered across paper sheets, emails and Excel files. Analytics produced by Edusign are ready to use for audits, funding body reports and annual pedagogical reviews.