What Is Learning Analytics? – Data That Improves Training

Your learning management system can export a spreadsheet of who completed what and when. It's tidy, it's accurate and it rarely changes a single decision. Meanwhile, you suspect the data hides more useful answers. Which questions trip people up? Where do they quit? Does the training connect to anything on the job? What is learning analytics, and how can it actually improve your training?

Learning analytics is the practice of collecting, analyzing and reporting data about learners and their learning, so you can understand what's working and improve it. In corporate training, that means going beyond completion counts to look at how people move through courses, how they perform on assessments, where they struggle and whether their work changes afterward. The data comes from your learning platforms, assessment results, surveys and business systems. Good learning analytics starts with a specific question, uses only the data needed to answer it and ends with a decision, such as rewriting a module, changing a question or targeting extra support.

Where Does the Definition of Learning Analytics Come From?

The most common definition comes from the research community that formed around the field in the early 2010s. The Society for Learning Analytics Research (SoLAR) describes it as the measurement, collection, analysis and reporting of data about learners and their contexts, for the purpose of understanding and improving learning and the environments in which it happens.

The last part of that definition is the important one. The goal isn't reporting for its own sake. It's improvement. A dashboard nobody acts on isn't learning analytics. It's decoration.

What Are the Four Types of Learning Analytics?

The four types of learning analytics are descriptive, diagnostic, predictive and prescriptive. They build on each other, and most training teams get the biggest gains from the first two.

• Descriptive: what happened. Completions, scores, time spent, drop-off points.
• Diagnostic: why it happened. Which question confused people, which module lost them, which group struggled.
• Predictive: what's likely to happen. Which learners are at risk of not finishing or not passing, based on early patterns.
• Prescriptive: what to do about it. Recommending a review module, a coaching session or a different path.

Here's an honest observation: most organizations jump to wanting predictions before they've done good diagnosis. Diagnosis is cheaper, faster and usually more useful. Start there.

What Data Does Learning Analytics Use?

Learning analytics uses data from wherever learning and performance leave a trace. The richer and better connected the data, the more questions you can answer.

Course and Platform Data

Your learning management system (LMS) records enrollments, completions, scores and time. Courses published with SCORM (Sharable Content Object Reference Model) usually send a limited set of data: status, score and time, sometimes individual question responses. Courses using xAPI (the Experience API) can send much more detailed records, such as "learner chose option B in scenario 3," to a learning record store (LRS).

Assessment Data

Item-level assessment data is one of the most valuable and least used sources. It shows which questions most people miss, which ones everybody gets right and which wrong answers are most popular. That tells you about both the content and the questions themselves.

Survey and Feedback Data

Post-course surveys capture relevance, confidence and barriers. Free-text comments often explain patterns the numbers only hint at.

Business and Performance Data

Quality metrics, sales figures, safety records, support tickets and time to proficiency live in other systems. Linking them to training data is how you connect learning to results. It's also the hardest step, because it usually requires cooperation from other departments.

How Does Learning Analytics Improve Training in Practice?

Learning analytics improves training by pointing to specific, fixable problems. The table below shows common questions, the data that answers them and the kind of action that follows.

Here's a quick example with illustrative numbers. Item analysis shows that 70% of learners pick the same wrong answer on one question. That's not 70% of learners failing. It's a sign that either the question is misleading or the course teaches that point badly. Either way, you know exactly which screen or which question to fix.

Another example: a 30-minute course shows a sharp drop-off at a long, text-heavy policy section. Splitting that section into a short summary and a reference document can reduce the drop-off without touching the rest of the course.

Which Training Metrics Are Vanity Metrics?

Vanity metrics are numbers that look good in a report but don't lead to any decision. Training is full of them, and they crowd out the measures that matter.

• Total enrollments. A big number, but it says nothing about learning.
• Completion rate on mandatory courses. If people must finish, a high rate mostly measures compliance with a deadline.
• Hours of training delivered. More hours can mean more value, or just longer courses.
• Average satisfaction score. Useful as a warning light, but people can enjoy training that changes nothing.
• Number of courses in the catalog. A larger library can be harder to use, not easier.

None of these is useless. Each one can flag a problem. The test is simple: if the number went up or down by 20%, would you do anything differently? If not, it belongs at the bottom of the report, not the top.

Compare that with a metric like "share of new hires who reach full productivity within eight weeks." If it moves, someone has to act. That's the kind of number learning analytics should help you find, track and improve.

How Do You Get Started With Learning Analytics?

You get started with learning analytics by picking one course, asking one question and using the data you already have. Big platforms and data warehouses can come later, if they're needed at all.

1. Choose a course that matters: high enrollment, high stakes or frequent complaints.
2. Write the question you want answered, such as "Why do so many people fail the final assessment?"
3. Pull the data you already have: item-level scores, drop-off, survey comments.
4. Look for patterns, then check them with a few learners or managers.
5. Make one change, then measure again.

Once that cycle works, you'll know what data you wish you had. That's the right moment to consider xAPI, an LRS or connections to business systems. Buying tools before you know your questions usually produces expensive dashboards and few decisions.

What About Privacy and Trust?

Privacy and trust matter because learning data is personal. People are more willing to learn, make mistakes and try again when they trust how their data will be used.

• Tell learners what you collect and why.
• Use aggregated data for course improvement whenever you can.
• Be careful about using training data in performance reviews. If people think practice mistakes will be held against them, they'll stop practicing honestly.
• Follow your organization's data policies and any regulations that apply.
• Limit access to individual-level data to people who need it.

A simple rule helps: use learning data mainly to improve learning. When you need to use it for other purposes, say so openly.

Why Do Assessments Matter So Much for Analytics?

Assessments matter because they're usually the richest learning data you have, but only if they measure real skill. If a test checks recall of trivia, its data tells you who remembers trivia. If it presents realistic scenarios, its data tells you who can make good decisions on the job.

That's why analytics and assessment design go hand in hand. Better questions produce better data, and better data produces better decisions. If your assessments need work, our guide to writing an eLearning assessment that measures real skill is a good next step.

Learning Analytics Starts With a Question, Not a Dashboard

Learning analytics, in short, is the practice of using data about learners and learning to understand what works and improve it. It moves from describing what happened to diagnosing why, then to predicting and prescribing as your data matures.

Start small: one course, one question, the data you already have. Fix what the data points to, measure again and build from there. That loop, repeated, does more to improve training than any dashboard ever will.

← All articles