Back to Notes
How We Predict a Student’s Exam Result — A Week Before They Sit For It
How We Predict a Student’s Exam Result — A Week Before They Sit For It
How We Predict a Student’s Exam Result — A Week Before They Sit For It
Here’s a strange truth about education: for decades, we’ve measured learning the way you’d count lifeboats after the Titanic has already sunk. A student writes an exam, the marks come out, and only then do we discover they never really understood the subject. By that point the ship has gone down—the gap we just “found” had been widening for weeks.
Here’s a strange truth about education: for decades, we’ve measured learning the way you’d count lifeboats after the Titanic has already sunk. A student writes an exam, the marks come out, and only then do we discover they never really understood the subject. By that point the ship has gone down—the gap we just “found” had been widening for weeks.
Here’s a strange truth about education: for decades, we’ve measured learning the way you’d count lifeboats after the Titanic has already sunk. A student writes an exam, the marks come out, and only then do we discover they never really understood the subject. By that point the ship has gone down—the gap we just “found” had been widening for weeks.
That final grade is what analysts call a lagging indicator. It tells you what already happened; it can’t tell you what’s about to. At NxtWave, the Learning Outcome pod works across our products—Academy, Intensive and NIAT—each of which has its own learning portal, where the students enrolled in that product watch videos, solve MCQs, practise coding and sit exams. We decided to stop counting lifeboats—and start reading the early signals instead.
That final grade is what analysts call a lagging indicator. It tells you what already happened; it can’t tell you what’s about to. At NxtWave, the Learning Outcome pod works across our products—Academy, Intensive and NIAT—each of which has its own learning portal, where the students enrolled in that product watch videos, solve MCQs, practise coding and sit exams. We decided to stop counting lifeboats—and start reading the early signals instead.
That final grade is what analysts call a lagging indicator. It tells you what already happened; it can’t tell you what’s about to. At NxtWave, the Learning Outcome pod works across our products—Academy, Intensive and NIAT—each of which has its own learning portal, where the students enrolled in that product watch videos, solve MCQs, practise coding and sit exams. We decided to stop counting lifeboats—and start reading the early signals instead.
So we measure two things about every student, in real time: how hard they’re actually working, and whether that work is turning into real skill. Effort and value. Here’s how both work—and why the usual ways of measuring them quietly lie to you.
So we measure two things about every student, in real time: how hard they’re actually working, and whether that work is turning into real skill. Effort and value. Here’s how both work—and why the usual ways of measuring them quietly lie to you.
So we measure two things about every student, in real time: how hard they’re actually working, and whether that work is turning into real skill. Effort and value. Here’s how both work—and why the usual ways of measuring them quietly lie to you.
Effort You Can’t Fake
Effort You Can’t Fake
Effort You Can’t Fake
The old way of measuring effort was embarrassingly easy to game. Log in, keep the screen on, walk away and make a coffee—six hours “logged”, and almost nothing learnt. It’s the gym-membership illusion: sitting on the bench for an hour doesn’t grow a single muscle.
The old way of measuring effort was embarrassingly easy to game. Log in, keep the screen on, walk away and make a coffee—six hours “logged”, and almost nothing learnt. It’s the gym-membership illusion: sitting on the bench for an hour doesn’t grow a single muscle.
The old way of measuring effort was embarrassingly easy to game. Log in, keep the screen on, walk away and make a coffee—six hours “logged”, and almost nothing learnt. It’s the gym-membership illusion: sitting on the bench for an hour doesn’t grow a single muscle.
So we threw screen time out entirely and started measuring meaningful engagement instead—split into two kinds, each weighted by how much real work it demands.
So we threw screen time out entirely and started measuring meaningful engagement instead—split into two kinds, each weighted by how much real work it demands.
So we threw screen time out entirely and started measuring meaningful engagement instead—split into two kinds, each weighted by how much real work it demands.
Passive learning—watching a video, reading notes or a cheat sheet—carries a weight of 0.5. It’s genuine learning, but lower intensity, and it’s now very hard to fake: while a video plays, the system logs a “heartbeat” every 10 seconds, but only while that video is the active window. Switch tabs or minimise it and the count stops. Reading has a sensible time ceiling too—leave a page open all day and the extra hours simply don’t count.
Passive learning—watching a video, reading notes or a cheat sheet—carries a weight of 0.5. It’s genuine learning, but lower intensity, and it’s now very hard to fake: while a video plays, the system logs a “heartbeat” every 10 seconds, but only while that video is the active window. Switch tabs or minimise it and the count stops. Reading has a sensible time ceiling too—leave a page open all day and the extra hours simply don’t count.
Passive learning—watching a video, reading notes or a cheat sheet—carries a weight of 0.5. It’s genuine learning, but lower intensity, and it’s now very hard to fake: while a video plays, the system logs a “heartbeat” every 10 seconds, but only while that video is the active window. Switch tabs or minimise it and the count stops. Reading has a sensible time ceiling too—leave a page open all day and the extra hours simply don’t count.
Why give passive learning any weight at all? Because some students are concept-first learners—they watch and read to build a mental model before they ever touch a keyboard. Scoring that as zero would punish a perfectly valid way of learning, so half-weight respects the foundation.
Why give passive learning any weight at all? Because some students are concept-first learners—they watch and read to build a mental model before they ever touch a keyboard. Scoring that as zero would punish a perfectly valid way of learning, so half-weight respects the foundation.
Why give passive learning any weight at all? Because some students are concept-first learners—they watch and read to build a mental model before they ever touch a keyboard. Scoring that as zero would punish a perfectly valid way of learning, so half-weight respects the foundation.
Active learning—writing code in the playground, typing notes, solving MCQs, asking the AI tutor a doubt—carries the full weight of 1.0. Here too there’s a guard against faking it: 60 seconds of no keyboard or mouse activity on an MCQ (120 seconds in the coding playground) quietly pauses the timer. Sitting idle with the screen open doesn’t count.
Active learning—writing code in the playground, typing notes, solving MCQs, asking the AI tutor a doubt—carries the full weight of 1.0. Here too there’s a guard against faking it: 60 seconds of no keyboard or mouse activity on an MCQ (120 seconds in the coding playground) quietly pauses the timer. Sitting idle with the screen open doesn’t count.
Active learning—writing code in the playground, typing notes, solving MCQs, asking the AI tutor a doubt—carries the full weight of 1.0. Here too there’s a guard against faking it: 60 seconds of no keyboard or mouse activity on an MCQ (120 seconds in the coding playground) quietly pauses the timer. Sitting idle with the screen open doesn’t count.
Add the two together and you get a student’s Weighted Engagement Time—effort measured by what they actually did, not by how long a tab was open.
Add the two together and you get a student’s Weighted Engagement Time—effort measured by what they actually did, not by how long a tab was open.
Add the two together and you get a student’s Weighted Engagement Time—effort measured by what they actually did, not by how long a tab was open.
It’s a small shift with a big consequence: time spent is trivial to fake, but genuine engagement isn’t.
It’s a small shift with a big consequence: time spent is trivial to fake, but genuine engagement isn’t.
It’s a small shift with a big consequence: time spent is trivial to fake, but genuine engagement isn’t.
Rewarding the Comeback
Rewarding the Comeback
Rewarding the Comeback
Two students hit the same bug. One rage-quits and closes the tab; the other tries four different fixes until it works. Treating them identically would be absurd—so we don’t. Alongside time, we measure persistence directly.
Two students hit the same bug. One rage-quits and closes the tab; the other tries four different fixes until it works. Treating them identically would be absurd—so we don’t. Alongside time, we measure persistence directly.
Two students hit the same bug. One rage-quits and closes the tab; the other tries four different fixes until it works. Treating them identically would be absurd—so we don’t. Alongside time, we measure persistence directly.
A persistence factor rewards grit: retry a problem you failed instead of abandoning it, and you earn +0.25; come back to a task you’d walked away from, another +0.25. It rolls into one clean formula—Learning Effort = Weighted Engagement Time × (1 + Persistence Factor).
A persistence factor rewards grit: retry a problem you failed instead of abandoning it, and you earn +0.25; come back to a task you’d walked away from, another +0.25. It rolls into one clean formula—Learning Effort = Weighted Engagement Time × (1 + Persistence Factor).
A persistence factor rewards grit: retry a problem you failed instead of abandoning it, and you earn +0.25; come back to a task you’d walked away from, another +0.25. It rolls into one clean formula—Learning Effort = Weighted Engagement Time × (1 + Persistence Factor).
A quick example. A student watches 20 minutes of video (passive → 10 points) and codes for 10 minutes (active → 10 points), for an engagement time of 20. During the coding they failed twice and came back both times—a persistence factor of 0.5. So 20 × 1.5 = 30. Their score jumps from 20 to 30, purely for not giving up.
A quick example. A student watches 20 minutes of video (passive → 10 points) and codes for 10 minutes (active → 10 points), for an engagement time of 20. During the coding they failed twice and came back both times—a persistence factor of 0.5. So 20 × 1.5 = 30. Their score jumps from 20 to 30, purely for not giving up.
A quick example. A student watches 20 minutes of video (passive → 10 points) and codes for 10 minutes (active → 10 points), for an engagement time of 20. During the coding they failed twice and came back both times—a persistence factor of 0.5. So 20 × 1.5 = 30. Their score jumps from 20 to 30, purely for not giving up.
Why We Removed the Camera
Why We Removed the Camera
Why We Removed the Camera
Older systems watched students through the webcam—face tracking, eye tracking, the works. We removed all of it, for two reasons. Trust: surveilling every blink corrodes the relationship between a learner and the product. And accuracy: staring at a screen isn’t learning—you can have your eyes open and be fast asleep.
Older systems watched students through the webcam—face tracking, eye tracking, the works. We removed all of it, for two reasons. Trust: surveilling every blink corrodes the relationship between a learner and the product. And accuracy: staring at a screen isn’t learning—you can have your eyes open and be fast asleep.
Older systems watched students through the webcam—face tracking, eye tracking, the works. We removed all of it, for two reasons. Trust: surveilling every blink corrodes the relationship between a learner and the product. And accuracy: staring at a screen isn’t learning—you can have your eyes open and be fast asleep.
Keystrokes and clicks are far more honest signals, and far harder to fake than a face pointed at a lens. We’d rather reward the students who genuinely do the work than the ones who have simply learnt to look busy.
Keystrokes and clicks are far more honest signals, and far harder to fake than a face pointed at a lens. We’d rather reward the students who genuinely do the work than the ones who have simply learnt to look busy.
Keystrokes and clicks are far more honest signals, and far harder to fake than a face pointed at a lens. We’d rather reward the students who genuinely do the work than the ones who have simply learnt to look busy.
Effort Isn’t the Same as Progress
Effort Isn’t the Same as Progress
Effort Isn’t the Same as Progress
You can push against a wall all day—sweat, strain, exhaust yourself—and the wall won’t move an inch. Effort, zero result. Students are no different: hours studied don’t automatically become skill. So alongside effort we measure Concept Mastery Evidence, and instead of one flat number it reads every concept along three dimensions.
You can push against a wall all day—sweat, strain, exhaust yourself—and the wall won’t move an inch. Effort, zero result. Students are no different: hours studied don’t automatically become skill. So alongside effort we measure Concept Mastery Evidence, and instead of one flat number it reads every concept along three dimensions.
You can push against a wall all day—sweat, strain, exhaust yourself—and the wall won’t move an inch. Effort, zero result. Students are no different: hours studied don’t automatically become skill. So alongside effort we measure Concept Mastery Evidence, and instead of one flat number it reads every concept along three dimensions.
Depth: can the student merely define a concept (level 1), or build a real project with it and teach it to someone else (level 7)? Independence: did they solve it themselves, lean on the AI tutor’s hints at every step, or copy the solution outright—because under exam pressure, only what you did yourself survives? Retention: is the knowledge still fresh, starting to fade, or already gone? We all learn things and forget them two weeks later; now we can see it happening.
Depth: can the student merely define a concept (level 1), or build a real project with it and teach it to someone else (level 7)? Independence: did they solve it themselves, lean on the AI tutor’s hints at every step, or copy the solution outright—because under exam pressure, only what you did yourself survives? Retention: is the knowledge still fresh, starting to fade, or already gone? We all learn things and forget them two weeks later; now we can see it happening.
Depth: can the student merely define a concept (level 1), or build a real project with it and teach it to someone else (level 7)? Independence: did they solve it themselves, lean on the AI tutor’s hints at every step, or copy the solution outright—because under exam pressure, only what you did yourself survives? Retention: is the knowledge still fresh, starting to fade, or already gone? We all learn things and forget them two weeks later; now we can see it happening.
When Effort Is High but Value Is Low
When Effort Is High but Value Is Low
When Effort Is High but Value Is Low
Picture a student who studies three hours a day—great engagement, high persistence—but the three-dimensional picture shows them stuck at level 2, leaning on hints, and forgetting fast. Plenty of effort, barely any value. The old instinct would be to write them off as “not smart”.
Picture a student who studies three hours a day—great engagement, high persistence—but the three-dimensional picture shows them stuck at level 2, leaning on hints, and forgetting fast. Plenty of effort, barely any value. The old instinct would be to write them off as “not smart”.
Picture a student who studies three hours a day—great engagement, high persistence—but the three-dimensional picture shows them stuck at level 2, leaning on hints, and forgetting fast. Plenty of effort, barely any value. The old instinct would be to write them off as “not smart”.
The system does the opposite. It reads the signal as a hard-working learner who is struggling—the effort is real, the approach is off—and responds like a mentor, not a judge: serving easier material, or flagging a teacher that this student needs personal attention. That distinction, effort-rich but value-poor, is one you simply can’t see from a final grade.
The system does the opposite. It reads the signal as a hard-working learner who is struggling—the effort is real, the approach is off—and responds like a mentor, not a judge: serving easier material, or flagging a teacher that this student needs personal attention. That distinction, effort-rich but value-poor, is one you simply can’t see from a final grade.
The system does the opposite. It reads the signal as a hard-working learner who is struggling—the effort is real, the approach is off—and responds like a mentor, not a judge: serving easier material, or flagging a teacher that this student needs personal attention. That distinction, effort-rich but value-poor, is one you simply can’t see from a final grade.
How Monday Predicts Friday
How Monday Predicts Friday
How Monday Predicts Friday
With a full picture of each student, we can forecast this week’s exam result days early—scoped only to what that week’s exam actually covers, not the whole course. Think of it as a chain with four links: if any one link snaps, the student is flagged as at-risk. We check them in order.
With a full picture of each student, we can forecast this week’s exam result days early—scoped only to what that week’s exam actually covers, not the whole course. Think of it as a chain with four links: if any one link snaps, the student is flagged as at-risk. We check them in order.
With a full picture of each student, we can forecast this week’s exam result days early—scoped only to what that week’s exam actually covers, not the whole course. Think of it as a chain with four links: if any one link snaps, the student is flagged as at-risk. We check them in order.
First, concept coverage: has the student touched at least 75% of the week’s syllabus—watched, read, attempted? If they’ve opened only 30%, we stop right there; no other signal can rescue an unread syllabus.
First, concept coverage: has the student touched at least 75% of the week’s syllabus—watched, read, attempted? If they’ve opened only 30%, we stop right there; no other signal can rescue an unread syllabus.
First, concept coverage: has the student touched at least 75% of the week’s syllabus—watched, read, attempted? If they’ve opened only 30%, we stop right there; no other signal can rescue an unread syllabus.
Second, independent mastery: of what they’ve learnt, is at least 50% done without hints? Hold someone’s hand through every practice problem and they can’t run alone in the exam hall.
Second, independent mastery: of what they’ve learnt, is at least 50% done without hints? Hold someone’s hand through every practice problem and they can’t run alone in the exam hall.
Second, independent mastery: of what they’ve learnt, is at least 50% done without hints? Hold someone’s hand through every practice problem and they can’t run alone in the exam hall.
Third, retention risk: is less than 10% of what they learnt slipping toward “aging” or “expired”? Forgotten material doesn’t resurface under pressure.
Third, retention risk: is less than 10% of what they learnt slipping toward “aging” or “expired”? Forgotten material doesn’t resurface under pressure.
Third, retention risk: is less than 10% of what they learnt slipping toward “aging” or “expired”? Forgotten material doesn’t resurface under pressure.
Fourth—and deliberately last—weekly effort. This surprises people; shouldn’t effort come first? But some students grasp things fast and clear the first three links with barely any effort, and penalising a clearly capable student for “not enough hours” would be nonsense. So effort acts as a tie-breaker—it matters most when mastery isn’t there yet, to check whether the student is at least trying.
Fourth—and deliberately last—weekly effort. This surprises people; shouldn’t effort come first? But some students grasp things fast and clear the first three links with barely any effort, and penalising a clearly capable student for “not enough hours” would be nonsense. So effort acts as a tie-breaker—it matters most when mastery isn’t there yet, to check whether the student is at least trying.
Fourth—and deliberately last—weekly effort. This surprises people; shouldn’t effort come first? But some students grasp things fast and clear the first three links with barely any effort, and penalising a clearly capable student for “not enough hours” would be nonsense. So effort acts as a tie-breaker—it matters most when mastery isn’t there yet, to check whether the student is at least trying.
The Point Isn’t to Judge. It’s to Help in Time.
The Point Isn’t to Judge. It’s to Help in Time.
The Point Isn’t to Judge. It’s to Help in Time.
None of this exists to stamp a “pass” or “fail” on anyone. When a student lands in the amber zone, the whole purpose is to answer one question early: why? Haven’t covered the syllabus? Too dependent on help? Forgetting what they learnt? Catch the reason before the exam and you can fix it while there’s still time.
None of this exists to stamp a “pass” or “fail” on anyone. When a student lands in the amber zone, the whole purpose is to answer one question early: why? Haven’t covered the syllabus? Too dependent on help? Forgetting what they learnt? Catch the reason before the exam and you can fix it while there’s still time.
None of this exists to stamp a “pass” or “fail” on anyone. When a student lands in the amber zone, the whole purpose is to answer one question early: why? Haven’t covered the syllabus? Too dependent on help? Forgetting what they learnt? Catch the reason before the exam and you can fix it while there’s still time.
That’s the real shift. Learning stops being a grade you read after the fact and becomes a live signal you can act on—genuine effort, honestly measured; real grit, rewarded; and a student’s privacy left fully intact. It turns out you can tell who’s going to struggle a week early. Not with a crystal ball—but with the right signals, read at the right time.
That’s the real shift. Learning stops being a grade you read after the fact and becomes a live signal you can act on—genuine effort, honestly measured; real grit, rewarded; and a student’s privacy left fully intact. It turns out you can tell who’s going to struggle a week early. Not with a crystal ball—but with the right signals, read at the right time.
That’s the real shift. Learning stops being a grade you read after the fact and becomes a live signal you can act on—genuine effort, honestly measured; real grit, rewarded; and a student’s privacy left fully intact. It turns out you can tell who’s going to struggle a week early. Not with a crystal ball—but with the right signals, read at the right time.
Want to work together?
Feel free to reach out for collaborations, inquiries, or just to say hello.
Want to work together?
Feel free to reach out for collaborations, inquiries, or just to say hello.
Want to work together?
Feel free to reach out for collaborations, inquiries, or just to say hello.
Let's Be Friends
Feel Free to Hit Me Up!
I always enjoyed product discussions and If you’re a startup founder or PM/Growth person and interested to chat! Hit me up on any social media platforms.
Crafted with ❤️ on Framer, All Rights Reserved © 2026 Guruprakash.
Let's Be Friends
Feel Free to Hit Me Up!
I always enjoyed product discussions and If you’re a startup founder or PM/Growth person and interested to chat! Hit me up on any social media platforms.
Crafted with ❤️ on Framer, All Rights Reserved © 2025 Guruprakash.
Let's Be Friends
Feel Free to Hit Me Up!
I always enjoyed product discussions and If you’re a startup founder or PM/Growth person and interested to chat! Hit me up on any social media platforms.
Crafted with ❤️ on Framer
All Rights Reserved © 2025 Guruprakash.