DOL English Platform - Discovery Log (Working Notes)

Updated: 2026-02-15 Status: in-progress

Confirmed decisions

1. Success priority (90-day)

1. Sustainable learning habit (highest priority)
2. Learning plan completion
3. Score increase (when user takes mock tests)

2. Sustainable habit definition

• User studies >= 4 days/week for 4 consecutive weeks.
• Each valid day requires at least 1 completed + submitted exercise.
• Streak is tracked as a motivation signal.

3. Streak validity rule

• A day counts if user completes and submits any exercise.
• No constraint on exercise difficulty, completion time, or accuracy.

4. Streak freeze rule

• If user misses 1 day after having streak, system grants exactly 1 freeze day.
• If user still does not submit during freeze day, streak resets.
• No extra freeze/rescue mechanism for now.

5. Post-submission flow

• After submitting, user goes to Result page first.
• Result page includes performance + explanations.
• On exit, user can move to Practice Management (and/or exercise bank based on context).

6. Practice Management recommendation section

• Show recently completed exercise.
• Provide recommendation options:
  • If result is weak: recommend easier corresponding practice.
  • If result is good: recommend 1 equivalent + 1 harder practice.
• Also provide options to switch to other skills.

7. Performance tier model (latest)

• Chua dat: <20%
• Can cai thien them: 20-59%
• Kha: 60-80%
• Gioi: 81-99%
• Xuat sac: 100%

8. Writing/Speaking async scoring behavior

• AI scoring can complete quickly (<1 minute) or slower (5-10 minutes).
• While pending, user may:
  • Return to exercise bank, or
  • Go to Practice Management.
• Practice Management must show pending-scoring item state.
• While score is pending, recommendation must NOT use score.
• Pending recommendation uses: level + topic + recent skill intent.
• User gets options for equivalent exercises and switching skill.

9. Async scoring ready notification behavior

• When AI scoring is finished, create an in-app notification entry.
• Show a small, non-blocking popover on screen (toast-like).
• Popover actions:
  • Quick open to the newly scored result page.
  • Dismiss/close.
• If user takes no action, popover auto-closes after a few seconds.
• If user ignores the popover, do NOT repeat reminder.
• Keep the result item only in Notification center for later access.

10. Self-study learning plan completion (V2)

• Self-study does not use a fixed, universal curriculum plan.
• Completion status is goal-based and mock-test-based:
  • If mock score is below user’s target selected at onboarding -> not completed / not achieved yet.
  • If mock score reaches or exceeds onboarding target -> achieved learning goal.
• Supporting progress signals:
  • Number of completed exercises in a time window.
  • Quality trend of exercise results over time.
  • Overall score trend from exercises and/or mock tests.
• Product philosophy for success measurement: “Better every day.”
• If later performance is better than earlier baseline, it is considered positive progress.

11. Version scope clarification

• This discovery and product definition are for V2 (not V1 legacy).

12. Evidence options for “user is improving” (V2)

• Improvement can be recognized through one or multiple evidence tracks:
  • Track A: 2 consecutive mock tests with score improvement.
  • Track B: average score of this week compared to last week.
  • Track C: average score of this month compared to last month.
  • Track D: average score of this 3-month cycle compared to previous 3-month cycle.
  • Track E: at least 5 recent learning days/attempts show an upward trend without sharp drop.
• “No sharp drop” rule is hybrid:
  • Use both percentage-based and program-scale thresholds.
  • System applies the stricter condition.
• Rollout strategy:
  • Start with shared threshold baseline for all programs.
  • After 4 weeks of real usage data, calibrate thresholds per program (IELTS/TOEIC/SAT).

13. Onboarding target guardrail by program (V2)

• Each program has a hard maximum target ceiling.
• User cannot select a target above that program ceiling.
• Example:
  • IELTS max target = 9.0.
  • User cannot set 10.0.
• Each program should also define:
  • minimum selectable target
  • fixed step size for target selection

14. Onboarding goal selection behavior (V2)

• Goal question appears in first-time onboarding right after registration.
• Goal selection is optional (user can skip to reduce signup friction).
• If user skips and has no goal:
  • they cannot compare results against target.
  • target-based progress comparison is unavailable.
• User can return later to set/update goal.
• Product should use non-forced nudges to encourage goal completion over time.
• For users without goal:
  • remove only goal-comparison layer and target-gap metrics.
  • keep normal learning analytics:
    • exercise activity statistics
    • score/result trend over time
    • progression signal (improving or stagnating)
    • diligence/consistency usage indicators
  • users can still understand whether they are improving from practice history.
• Diligence metric in V2 is hybrid:
  • active submission days per week
  • completed exercises per week
  • active learning time per week
• Initial rollout weight:
  • 40% active submission days
  • 40% completed exercises
  • 20% active learning time
• Active learning time counting rule:
  • Count only when user has real interactions (click/type/scroll/audio interactions).
  • Auto-pause when inactivity is detected to avoid idle/background inflation.
• User experience preference:
  • Inactivity timeout should be lenient (not too strict) to avoid interrupting normal study flow.
• V2 timeout selection:
  • Auto-pause active learning time after 90 seconds of inactivity.
• Resume behavior:
  • Active time resumes automatically on the first new user interaction.

15. Goal-completion encouragement surfaces (V2)

• Use multiple lightweight UI encouragement patterns (not forced):
  • Profile completion percentage pattern:
    • completing onboarding/goal selection increases profile completion toward 100%.
  • Profile-area subtle attention marker:
    • light visual indicator to drive curiosity and click-through.
    • indicator is removed once user completes goal selection.
  • Achievement mechanic:
    • user receives an achievement badge after selecting goal.
  • Practice Management reminder:
    • in goal-related section, remind user they have not set goal yet and show value unlock messaging.
• Overall tone: encourage and reward, not block or force.
• Goal-selection achievement strategy:
  • Main “goal selected” badge is granted one time only.
  • No milestone logic for later goal updates (removed for simplicity).
• Goal update feedback behavior:
  • When user updates target later, show a quick confirmation popover:
    • “Mục tiêu của bạn đã được thay đổi.”
• Operational boundary by learning mode:
  • CS/Teacher intervention applies to course-management mode only.
  • Self-study mode is fully automated (no human intervention workflow).

16. Self-study retention nudges and reactivation (V2)

• Reactivation should be staged by inactivity severity:
  • stage 1: user loses streak
  • stage 2: user has not studied for 1 week
  • stage 3: user has not studied for 1 month
  • stage 4: long-term dormant milestone (1 year)
• Each stage has its own dedicated reminder type (optimized per stage).
• Stage reminders are one-time triggers at each milestone (not daily repetitive reminders).
• Additional periodic reminder:
  • if user remains inactive for longer term, send recurring reminder every 3 months.
• Stage overlap assumption:
  • “lost streak” and “1 week inactive” are treated as sequential milestones in timeline, not simultaneous collisions.
• Primary objective for all stages:
  • remind user about platform and motivate return to studying.
• CTA strategy is diversified to avoid boredom:
  • explore new exercises
  • quick practice
  • start practicing now
  • other discovery/engagement actions related to learning content
• Default reminder channels (initial V2):
  • email
  • web browser push notification
  • Zalo message
• Channel control:
  • user can configure notification settings and enable/disable channels.
• Delivery note:
  • for now, simultaneous push across available channels is acceptable.
• Content personalization by inactivity duration:
  • streak-loss reminder: include lightweight recap of recent exercises, then comeback prompt.
  • 1-week+ inactivity: prioritize new/trending exercises because old personal data may be stale.
• Message frequency cap during inactive periods:
  • Maximum 1 reminder notification per day.
• Reactivation reset rule:
  • once user returns and completes 1 exercise, reactivation chain resets immediately and a new streak begins.
• Measurement priority:
  • strict reactivation success metrics are not a current priority.

Open question

• For the recurring 3-month reminder cycle, should it continue forever until user returns, or stop after a maximum number of attempts?