A
Lesson Content Schema
A modular structure for lessons, challenges, answers and hints, so topics slot into one format.
GCash serves tens of millions of people, many new to formal finance. The growth that matters is moving transactional users up into savings, investing, credit and insurance, and those products only get adopted when people understand them. That makes in-app education a direct activation lever rather than a side project.
The education that existed did not do that job. It lived as long, help-center style articles: formal, buried, and rarely opened. The few who started tended to leave before finishing. No feedback, no sense of progress, and no link to the moments where someone actually makes a money decision.
Measure learning completed rather than time spent. Every design decision followed from that.
One concept each, with clear stop points. Easy to start, easy to finish.
Completion, mastery and streaks. No infinite loops, no "just one more".
People see what is left, rather than being nudged onward by design.
Limited retries with fast feedback. Opt-in nudges, quiet hours, frequency caps.
The system was built so new topics and difficulty tuning do not need engineering every time. On top sits an observability layer that reads learning rather than raw minutes.
A modular structure for lessons, challenges, answers and hints, so topics slot into one format.
Levels, mastery gates, limited lives. The next topic unlocks only when the current one is understood.
A model built around outcomes: start, complete, mastery, retry, drop-off, notification response.
No-code updating of lesson content and difficulty, so the team tunes continuously without shipping code.
Opt-in nudges with quiet hours and frequency caps, governed centrally rather than per flow.
The discipline was not a one-off launch gate but a set of standing constraints that governed every change. The hardest was built into how experiments could ship.
Content, CX and compliance each had review gates. Experiments ran against hypothesis templates with thresholds tied to completion and mastery, never click-through. Session caps and stop-points were enforced across flows, so nothing became an infinite scroll.
A change could not increase time-spent at the expense of completion.
That single rule made the tempting engagement wins structurally off-limits. It is the same line drawn in Fig. II: learning had to go up, and more time alone did not count.
We did not manufacture numbers for this write-up. We would rather say what changed than dress it up.
Success moved from time-spent to learning-completed as the primary signal, and the team aligned on a weekly scorecard that surfaced mastery rather than clicks.
Intentionally bounded lessons improved user trust and return behavior, rather than stretching for one more minute of attention.
The team can add lessons and tune difficulty without heavy engineering, which is what makes a learning product sustainable rather than a one-off launch.
Across the wider set of financial products, using the same schema and progression rules.
Using the cohort and mastery data the system already captures to adapt sequencing and hints to the individual learner.
We design and build websites, web applications and automation systems for B2B companies. On this project our role was product management and product design: the learning system, the measurement model, and the guardrails that kept it trustworthy. We are careful about what we claim and honest about what we did.