Designing learning systems where progress isn’t immediately visible

Led product design for a sensory substitution platform translating sound into touch.

Client Neosensory
Role Product Designer
Scope iOS and Android, sensory substitution, learning system design
Neosensory — sensory substitution and learning design

Context

Neosensory translates sound into haptic signals.

Instead of amplifying hearing, users must learn a new sensory language.

Early adoption revealed a critical issue:

Users dropped off within the first weeks.

The system worked.
The learning did not.

The problem

Early experiences felt indistinguishable.

Users described it as: “everything feels the same.”

This ambiguity was interpreted as failure, not learning.

Engineering focused on signal quality.

I reframed the problem:

The issue was not accuracy.
It was expectation.

Users didn’t know what progress should feel like.

The shift

I shifted the focus from signal performance to learning design.

From:

  • accuracy → perception
  • output → progression
  • correctness → confidence

Retention depended on perceived competence, not actual accuracy.

The system

The product was restructured as a staged learning system:

  1. 1
    Onboarding

    Normalize ambiguity and set expectations

  2. 2
    Differentiation

    Help users detect meaningful differences

  3. 3
    Immersion

    Transition learning into real-world use

The goal was not faster accuracy.

It was sustained engagement through uncertainty.

Design decisions

  • Perceived competence over accuracy

    Users disengage when they feel they’re failing.

    We prioritized confidence signals alongside performance.

    Result: increased retention and continued engagement.

  • Staged exposure over signal complexity

    Reduced early signal variety and increased repetition.

    Result: faster differentiation and lower cognitive load.

  • Expectation framing over silent onboarding

    Made ambiguity explicit and expected.

    Result: confusion interpreted as progress, not failure.

  • Timing over constant feedback

    Reduced feedback frequency and improved timing.

    Result: deeper internal calibration and learning.

Behavioral signals

We measured learning through behavior, not just performance:

  • Perceived competence — Strong predictor of retention between Weeks 2–4
  • Time to first success — Day 3 milestone correlated with continuation
  • Emotional interpretation — Framing confusion as learning reduced drop-off
  • Return behavior — Increased session consistency after early progress

Tradeoffs

Designing for learning required restraint:

  • Slower initial progress vs faster perceived progress
  • Less feedback vs clearer signals
  • Simpler early experience vs full system exposure

We prioritized progression over completeness.

Impact

The system improved both retention and learning outcomes:

  • Reduced early-stage drop-off
  • Increased 3-month retention
  • Higher 12-week completion rates
  • Increased NPS and user confidence

The key shift:

Retention moved from fragile to structured.

Key insight

Users don’t drop off because systems fail. They drop off because they believe they are failing.

Reflection

This project changed how I think about learning systems.

Users don’t disengage because they can’t improve.
They disengage because they don’t recognize progress.

Key takeaways:

  • perceived competence drives retention more than accuracy
  • early micro-success determines long-term engagement
  • ambiguity must be framed, not removed

Designing for learning means designing for belief, not just performance.