Adaptive Learning and Progress Design for a Game-Based Platform

Ed-Tech
Children's Learning
Adaptive Systems

How rebuilding SplashLearn's progression logic around learning theory reduced churn and laid the foundation for a new product feature.

Adaptive Learning and Progress Design for a Game-Based Platform

Ed-Tech
Children's Learning
Adaptive Systems

How rebuilding SplashLearn's progression logic around learning theory reduced churn and laid the foundation for a new product feature.

+0%
+0%
+0%

Retention, KS1 Cohort.

Retention, KS1 Cohort.

What it opened up

What it opened up

Findings from this project directly informed SplashLearn's daily learning path feature, now live for 2M+ learners across Pre-K to Grade 5.

Findings from this project directly informed SplashLearn's daily learning path feature, now live for 2M+ learners across Pre-K to Grade 5.

COMPANY

Splashlearn

Splashlearn

MY ROLE

Learning and Product Designer

Learning and Product Designer

TEAM

Children Pre-K to Grade 5 + Parents

Children Pre-K to Grade 5 + Parents

DELIVERABLE

Adaptive path system + Challenge Cave

Adaptive path system + Challenge Cave

The Problem

25% of learners were dropping off a path that was designed to hold them

SplashLearn's curriculum ran through sequential educational games: complete one, unlock the next. The path worked for learners who could tolerate repetition, but 25% were dropping off. The assumed cause was repetitiveness. The goal was a 10% reduction in churn.

My starting hypothesis: the linear model was not just repetitive, it was misaligned with how children actually learn. Rather than iterating on the existing path, I proposed rebuilding the progression logic using competence-based knowledge space theory.

The Problem

A product built for urban, English-literate users was failing a new majority

SplashLearn's curriculum ran through sequential educational games: complete one, unlock the next. The path worked for learners who could tolerate repetition, but 25% were dropping off. The assumed cause was repetitiveness. The goal was a 10% reduction in churn.

My starting hypothesis: the linear model was not just repetitive, it was misaligned with how children actually learn. Rather than iterating on the existing path, I proposed rebuilding the progression logic using competence-based knowledge space theory.

The Framework

Why knowledge space theory, not just adaptive sequencing

The design needed a principled basis for three decisions: what to skip, what to surface, and when. Knowledge space theory provided that.

Competence-Based Knowledge Space Theory

KST models learning as a network of dependent skills rather than a fixed sequence. It maps which skills a learner has mastered and which are prerequisites for others, then adapts the path: skipping what has been demonstrated, returning to what has not. For a platform with 100+ games across five grade levels, this means genuinely different paths through the same curriculum — not the same path at different speeds.

research

The insight came from watching how people used their phones,not the product

The design needed a principled basis for three decisions: what to skip, what to surface, and when. Knowledge space theory provided that.

Competence-Based Knowledge Space Theory

KST models learning as a network of dependent skills rather than a fixed sequence. It maps which skills a learner has mastered and which are prerequisites for others, then adapts the path: skipping what has been demonstrated, returning to what has not. For a platform with 100+ games across five grade levels, this means genuinely different paths through the same curriculum — not the same path at different speeds.

Research

Two users. Two completely different needs. One solution.

Online video call interviews with children from Kindergarten and Grade 1 (75% of the user base), followed by separate interviews with their parents. Each child completed a 5-game adaptive session while I shadowed and took observational notes.

01

Children

Variety to sustain attention.
Celebration to sustain motivation.

After an initial peak of interest, learners found the path monotonous around games 5 and 6 and clicked off. When they beat games of increasing difficulty, they expressed genuine delight and shared it with their parents proudly.

Need: Break + Celebrate

02

Parents

Transparency over the algorithm.
Involvement over automation.

Parents did not trust the algorithm to manage their child's progress unsupervised. They wanted to know when games were being skipped and why.They frequently asked children to repeat skills rather than move on.

Need: Visibility + Control

These two sets of needs pointed toward the same solution: a structured moment of challenge and visibility, appearing at the right point on the adaptive path.

Design Decisions

Design Decisions

Challenge Cave: one feature, four purposes

Challenge Cave: one feature, four purposes

Challenge Mode appears after a learner has progressed along the adaptive path. It was designed to serve both user groups simultaneously.

Challenge Mode appears after a learner has progressed along the adaptive path. It was designed to serve both user groups simultaneously.

Break Monotony

A distinct experience with different rules, stakes, and rewards pulls the learner out of the routine path at the right moment.

Celebrate Achievement

Challenge Mode only appears when the learner is performing well. It is earned, not imposed — which is what makes it feel like an achievement.

Surface Revision

It draws on skipped games and weaker skills, testing real gaps rather than reinforcing what the learner already knows.

Parent Checkpoint

Parents can review performance across multiple skills in one session, without needing to track individual games on the adaptive path.

The Octo mechanic

Themed around "Challenge Cave" with an octopus character named Octo. Lose, and Octo offers one more chance at a cost. Win, and the whole treasure chest is yours

Variations:

Pre-K / Kindergarten: Rewards only, no timer

Grade 2 / 3: Timer introduced, streaks optional

Grade 4 / 5: Full stakes: timer, streaks, bonus rounds

outcome

The brief was a 10% reduction. The finding was bigger.

+0%
+0%
+0%

Daily Retention

Daily Retention

What the results revealed

The more significant finding was behavioural: children on the adaptive path did not stick to linear sequences even when the path adapted to them. They explored, revisited, and moved laterally. This informed a conversation with the product team about designing for that behaviour rather than against it.

+0%
+0%
+0%

Weekly retention

Weekly retention

The result was the daily learning path feature:

An adaptive sequence presented as a daily quest, removing the expectation of linear completion entirely. Instead of increasing retention for the day, the learning path focuses on bringing players back the next day for a new path.

Create a free website with Framer, the website builder loved by startups, designers and agencies.