Learning Analytics: Meaning, Types, and a Simple Weekly Plan for Coaching Institutes to Use Student Data Well

What learning analytics means, in plain words

Every time a student comes to class, writes a test, watches a recorded lecture or asks a doubt, they leave a small trace of data. One trace tells you very little. A few weeks of traces, read together, tell you a lot: who is slipping, which topic the whole batch did not understand, and which lecture is too long or confusing.

Learning analytics is the work of turning those traces into decisions. The word "analytics" can sound technical, but the core idea is old. Good teachers have always looked at test results, noticed the quiet student at the back and changed their plan. Learning analytics does the same thing in a more regular and organised way, using the digital data that online tests and learning management systems now record for you.

Two definitions are widely quoted. The first International Conference on Learning Analytics and Knowledge, held in 2011, described it as the measurement, collection, analysis and reporting of data about learners and their contexts, to understand and improve learning and the places where it happens. In 2026 SoLAR updated its definition, adding interpretation and communication, and insights that are actionable. That last word matters most. A report that nobody acts on is not learning analytics. It is just a report.

A simple test: if a number on your screen cannot change what a teacher does next week, stop tracking it.

The four types of learning analytics, with coaching examples

SoLAR lists four methods: descriptive, diagnostic, predictive and prescriptive analytics. Each answers a different question. Most small institutes only need the first two, and do the last one themselves, as teachers.

Big platforms sell predictive and prescriptive features that use machine learning. A coaching centre with 200 students does not need them. You can predict risk well enough with two or three clear rules, as the worked example below shows, and your teachers know the students far better than any model.

Learning analytics, educational data mining and progress reports: the difference

You will see three terms used close together. They overlap, but they are not the same.

• Learning analytics is about using data to help real students and teachers, usually quickly, while the course is still running. The person stays in control and the data supports their decision.
• Educational data mining is the more technical, research side: building models and algorithms that find patterns in large sets of student data. The Handbook of Learning Analytics, published by SoLAR, notes that the two fields are closely linked but differ in aims, methods and how much they automate.
• A student progress report is one output of the process: a summary of one student's results that you send to the student or parent. Learning analytics is what happens before it, when you read the whole batch's data to decide what to teach and whom to help. For the report itself, see our guide to the student progress report.

For a coaching institute, the practical meaning is simple. Educational data mining is for researchers and very large platforms. Learning analytics is for you, every week.

The data your institute already has

Before you collect anything new, list what you already record. Most institutes are sitting on more useful data than they think, spread across registers, test sheets, WhatsApp groups and apps.

Notice that attendance and question-wise test results do most of the work. If you only have time for two sources, choose those two.

Six numbers worth tracking, and the ones to ignore

The biggest risk with learning analytics is tracking too much. A dashboard with thirty numbers gets opened once and then forgotten. Start with these six. The warning levels in the table are examples to begin with, not rules from any official body: set your own after a month, once you know what is normal for your batches.

Numbers that look useful but rarely are

• Total logins or app opens. A student can open the app ten times a day and learn nothing.
• Total time spent on a video. A video left playing while the student eats dinner counts the same as one watched with full attention.
• Batch averages alone. An average of 62 can hide three students at 20 who need help today.
• Rankings shared with the whole class. They tell weaker students they are weak, and often make them stop trying. Use ranks, if at all, privately.

How to start learning analytics in your institute, step by step

1. Pick one question to answer. For example: which students in Class 10 are at risk of failing maths this term? A clear question stops you from collecting data for its own sake. The trade-off: a narrow question is easy to answer but covers fewer students, so start narrow and widen it after a month.
2. Choose two or three measures from the table above that answer that question. For the Class 10 question, attendance, test score trend and question-wise accuracy are enough.
3. Decide where the data will live. A small tuition class can use a paper register and one spreadsheet. A larger institute with recorded courses is better served by a learning management system that records attendance, test results and video progress for you. The decision depends on how many students you have and how much time your staff can spend on typing marks.
4. Set your warning signs before you look at the data. Write them down, for example: below 75% attendance in four weeks, or a falling score in two tests in a row. Deciding in advance stops you from judging students by mood or by who is loud in class.
5. Fix a weekly review. Thirty minutes on the same day every week, with the batch teacher and whoever calls parents. The review produces a short list: who needs a call, which topic needs re-teaching, which lecture needs fixing.
6. Act, then check whether the action worked. If you re-taught factorising, did the re-test improve? If you called a parent, did attendance come back? This closing step is what turns numbers into learning. Without it, you are only keeping records.

If you run online tests, a tool that marks answers and shows question-wise results makes step 2 almost free. Our guide to online exam software explains what to look for, and the guide on how to make a question paper shows how to tag each question by topic, which makes question-wise analysis far more useful.

Worked example: one month of learning analytics in a Class 10 maths batch

This is a made-up example to show the method. The institute, the students and the numbers are invented. Imagine a tuition centre with a Class 10 maths batch of 24 students. The teacher holds 12 classes a month and gives a short 10-question test every Saturday. The question for the month is: who is at risk, and which topics need re-teaching?

Week 4: the student sheet

The teacher keeps one row per student. Here are six of the 24 rows at the end of week 4. The warning signs set at the start were: attendance below 75%, or a score that fell in two tests in a row.

Look at what the sheet shows. Student B is the clear risk: missing classes and falling every week. Student E scores well but is missing a lot of classes. That is worth a call too, because good marks today can hide a student who is about to leave. Student D attends nearly every class yet keeps falling, so absence is not the cause: something in the topics is. Student F is low but not falling, which needs a different kind of help from Student D.

Week 4: question-wise results

Next, the teacher looks at the Saturday test question by question, for the whole batch of 24.

The three weakest questions are all about quadratic equations. That is a batch problem, not a student problem, and it explains why Student D, who attends every class, keeps falling: the batch moved on to quadratics in week 2. Before re-teaching Q9, the teacher reads the question again, because a 25% score can also mean the question was badly worded.

The decisions from a 30-minute review

• Call the parents of Students B and E this week. For B, ask about the missed classes and offer a catch-up session. For E, ask simply whether everything is all right.
• Run one 40-minute revision class on factorising and the discriminant, with fresh examples, before moving to the next chapter.
• Sit with Student D for ten minutes after class to find out where the confusion starts.
• Give Student F easier practice sheets that build up step by step, rather than more of the same test.
• Re-test quadratics next Saturday, with new questions, and compare.

The next Saturday, the teacher checks whether the share of the batch getting quadratics questions right has gone up, and whether B and E came back to class. That check is the whole point. If nothing changed, the plan changes, not just the numbers.

A ready weekly review template you can copy

Use this template in Google Sheets, Excel or a notebook. It fits on one page and takes about half an hour a week for one batch.

Sheet 1: the student sheet (one row per student)

• Columns: student name, batch, classes held, classes attended, attendance %, last three test scores, flag, action taken, date of action, result after one week.
• Formula for attendance: classes attended divided by classes held. Format the cell as a percentage.
• Colour the warning signs automatically. In Google Sheets, select the attendance column, open Format, then Conditional formatting, and add a rule such as Less than 75 with a red fill. Google's own help page uses the same idea as its example: a teacher highlighting test scores below a set mark.

Sheet 2: the question sheet (one row per test question)

• Columns: test date, question number, topic, number correct, number who attempted, share correct, action.
• If you give tests as Google Forms quizzes, you do not have to count by hand. Google's help page says quiz responses get automatic summaries, including frequently missed questions, graphs marked with the correct answers, and the average, median and range of scores.
• Sort by share correct, lowest first. The top three rows are your re-teaching list.

The weekly review checklist

1. Update attendance and the latest test scores.
2. List every student who crossed a warning sign this week.
3. List the three questions the batch got wrong most often, and their topics.
4. Check last week's actions: did they work?
5. Decide this week's actions, with one name against each.
6. Write the date and the action in the student sheet, so the next review can check it.

Learning analytics tools: what you need at each size

You do not need special learning analytics tools to start. What you need depends on how many students you have and whether you teach in class, online or both.

If you are choosing a learning management system, look for three things on the reporting side: question-wise test results, progress for each student and not only averages, and a way to download the data so you are never locked in. Our guide to choosing a learning management system covers the rest.

Sikhami's free LMS for coaching institutes includes course analytics for video courses: how many students enrolled, who is active this week, average progress, who finished, a table of test results and a chart of where students stop watching. You can also download the student list as a CSV file. For classroom batches, the institute dashboard shows attendance over the last month next to today's classes and fees. The course analytics help page explains each number.

Student data and the law: what India's DPDP rules say about children

Learning analytics uses personal data about students, and most coaching students in India are children. The Digital Personal Data Protection Act, 2023 (the DPDP Act) and the Digital Personal Data Protection Rules, 2025 set the rules. This section sums up what the official texts say. It is not legal advice; if you are unsure how the rules apply to you, ask a lawyer.

What the Act says

• Under the Act, a child is anyone who has not completed 18 years of age. So most school-level coaching students are children in law.
• Section 9(1): before processing a child's personal data, the organisation must get verifiable consent from a parent or lawful guardian.
• Section 9(2): it must not process a child's data in a way that is likely to cause any detrimental effect on the child's well-being.
• Section 9(3): it must not do tracking or behavioural monitoring of children, or targeted advertising directed at children.
• Section 8: the organisation must take reasonable security safeguards to prevent a personal data breach, and must erase data when the purpose is over or consent is withdrawn, unless a law requires it to be kept.

The exemption for educational institutions

Rule 12 and the Fourth Schedule of the DPDP Rules, 2025 exempt some organisations from Sections 9(1) and 9(3). One of them is an educational institution, which the Schedule defines as an institution of learning that imparts education, including vocational education. The condition is strict: the processing must be restricted to tracking and behavioural monitoring for the educational activities of the institution, or in the interests of the safety of children enrolled with it. Section 9(2), the duty not to harm a child's well-being, is not in the exemption.

In plain words: tracking attendance, test results and lecture progress to teach your enrolled students better looks like the kind of use the exemption describes. Using the same data to target advertising at children, or for anything outside teaching and safety, does not. Whether a particular coaching centre counts as an educational institution under this definition is a legal question the rules do not spell out for coaching, so take advice if it matters to you.

When the rules apply

The rules were notified on 13 November 2025. Most of the duties, including the rules on consent, children's data and security safeguards, come into force eighteen months after their publication in the Gazette, which is in the middle of 2027. Rule 6 on security safeguards asks, at the minimum, for measures such as encryption or masking, access control, logs of who accessed the data, backups, and keeping those logs for one year. It is sensible to build these habits now rather than in a rush later.

Simple habits that keep student data safe

• Tell students and parents, in the admission form, what you record and why: attendance, marks and progress, to help teaching.
• Collect only what you will use. If you never act on a number, do not record it.
• Give each staff member access only to the batches they teach.
• Do not post marks or ranks in a WhatsApp group with every parent in it. Share each student's results with that student's family only.
• Delete the data of students who left long ago, unless a law needs you to keep it.

What research says about learning analytics

Most research on learning analytics comes from universities in other countries, so read it as a guide for your institute, not a promise.

• Early warnings shown as simple signals are an old idea. The Handbook of Learning Analytics describes Course Signals at Purdue University as an early, famous example: it showed students their predicted outcome as a traffic light, based on grades so far, time on task and past performance. A red, amber and green column in your sheet uses the same idea.
• Dashboards are meant to help people decide, not decide for them. The Handbook's chapter on dashboards says they aim to promote awareness, reflection and sense-making, and compares a dashboard to a car with a human driver who stays in control. That is why the worked example ends with a teacher's judgment, not an automatic message.
• Prediction is not the same as explanation. The Handbook's chapter on predictive modelling separates models that predict what will happen from models that explain why. Its chapter on measurement adds that facts like a learner's age or sex may predict results but give weak grounds for action, because they are not the cause and cannot be changed. Act on causes you can change, such as missed classes or a misunderstood topic.
• Ethics and privacy help, they do not get in the way. The Handbook's ethics chapter points to research that sees ethics and privacy as enablers of learning analytics rather than barriers. Students and parents who trust you with their data are more likely to act on what you tell them.

India's National Education Policy 2020 points the same way. It calls for continuous formative and adaptive assessment to track and support each student's learning, and lists assessment analytics among the things that bodies like PARAKH will design for online assessment. Our guide to formative assessment shows how to build the small, frequent checks that give learning analytics its best data.

Common mistakes, and how to avoid them

• Collecting data and never acting on it. Fix the weekly review in the timetable, and end it with named actions.
• Judging students by one test. Look at the trend over three tests and at attendance together.
• Blaming the student for a batch-wide gap. If more than half the batch got a question wrong, the teaching or the question needs fixing first.
• Trusting a bad question. Before re-teaching a topic, read the question again and check the answer key.
• Labelling students as weak in front of others. Use flags privately, to start a kind conversation, not to sort children.
• Measuring teachers by raw batch averages. Batches start at different levels. Compare a batch with itself over time.
• Buying a tool before knowing the question. Decide what you want to find out first; the tool comes second.
• Sharing data loosely. Results in open WhatsApp groups and spreadsheets shared with anyone who has the link are the most common leaks in small institutes.

When learning analytics is not for you, yet

Be honest about where you are. A formal learning analytics routine may not be worth it if:

• You teach fewer than about ten students yourself. You already know each one well, and a notebook does the job.
• You do not give regular tests. Without regular checks, there is little data to read. Start with weekly short tests first.
• Nobody has time to act on what the review finds. Data with no follow-up only creates guilt. Free up one hour a week first, then start.
• Your problem is not learning but something else, such as fees, timings or the building. Fix that first; analytics will not.

For everyone else, start small: one batch, three measures, one review a week. After a month you will know which numbers matter in your institute, and you can grow from there. If you teach online as well, our guide to asynchronous learning explains how to keep self-paced students moving, which is where course progress data helps most.