ROI Case File No.276｜'Aurora HealthTech's Small Experiments'

Chapter 1: The Crime of Feature Bloat—Aiming for Perfection, Ending in Failure

The week after resolving the Zenith Media PEST analysis case, a consultation arrived from North America regarding a healthtech company's development stagnation. Case File 276 of Volume 22 "The Pursuit of Reproducibility" is a story about abandoning perfectionism and acquiring the attitude of learning from small experiments.

"Detective, we spent 18 months developing the perfect health management app. We implemented all features and addressed every use case. But right after release, no users are using it."

David Park, Chief Product Officer of Aurora HealthTech from Seattle, visited 221B Baker Street unable to hide his disappointment. In his hands were thick feature specification documents contrasted sharply with stagnant app download graphs.

"We're a Washington state startup developing personal health management apps. To create the perfect product, we loaded every feature. Yet users say 'too complex' and churn."

Aurora HealthTech's Perfect Failure: - Founded: 2022 (healthtech startup) - Development period: 18 months - Development cost: 800 million yen - Implemented features: 127 features - Downloads 1 month after release: 8,500 - 1-week retention rate: 12% - Monthly active users: 320

David's expression held deep confusion.

"The problem is we pursued 'perfection' too much. Step counting, calorie calculation, sleep analysis, heart rate monitoring, meal logging, exercise logging, mental health checks, medical institution linkage, medication management, health goal setting, friend competition features... We implemented everything. But users say 'don't know what to do.'"

Feature Bloat: - Initial plan: 20 features - Added during development: "Need this too, need that too" - Final implementation: 127 features - User voices: "Too many features, don't know where to start"

Development Mindset: - "Competitors have this feature, so we need it too" - "Might need it in future, so build it now" - "Since we're building, make it perfect" - Result: 18 months development, 800 million yen investment, and failure

"We aimed for perfection and created something nobody uses."

Chapter 2: MVP and Learning Metrics—Build Small, Learn Fast

"David, did you ask users for opinions during development?"

To my question, David answered.

"We conducted market research before development. We surveyed 'what features do you want' and implemented all highly-requested features."

Development Process (Perfectionist Type): - Phase 1: Market research (1 month) → Create requested feature list - Phase 2: Specification (2 months) → Detailed design of all features - Phase 3: Development (12 months) → Implement all features - Phase 4: Testing (2 months) → Bug fixes - Phase 5: Release (1 month) → Users touch for first time - Result: After 18 months, discovered "nobody uses it"

I explained the importance of learning.

"Perfect plans don't exist. You can't know what users truly want without having them use it. Lean Startup—build small, learn fast, correct direction. This repetition creates truly valuable things."

⬜️ ChatGPT｜Catalyst of Concepts

"Perfection is the enemy. Build small, break fast, keep learning"

🟧 Claude｜Alchemist of Narratives

"Value is only a hypothesis. Only when validated does value become truth"

🟦 Gemini｜Compass of Reason

"LEAN is the science of learning. Hypothesize, experiment, judge with data"

The three members began analysis. Gemini deployed the "Healthtech-Specialized Lean Startup" framework on the whiteboard.

Lean Startup's 3 Principles: 1. MVP (Minimum Viable Product) - Validate value hypothesis with minimum features 2. Build-Measure-Learn - Accelerate the build, measure, learn cycle 3. Validated Learning - Determine direction through data-based learning

"David, let's transform Aurora's development into a lean learning process."

Chapter 3: Hypothesis Decomposition—From 127 Features to 3 Features

Phase 1: Value Hypothesis Identification (1 week)

First, we articulated the "value hypotheses" behind 127 features.

Question: "Why is this app needed?"

Initial Hypothesis (Vague): "People want to be healthy"

Clarified Hypotheses: 1. Hypothesis A: People want to establish exercise habits but can't sustain them 2. Hypothesis B: People want to manage diet but recording is tedious and doesn't continue 3. Hypothesis C: People want to improve sleep quality but don't know what to do

Phase 2: Customer Interviews (2 weeks)

To validate each hypothesis, we conducted deep interviews with 30 target users.

Interview Subjects: - Health-conscious 30-45 year-olds without exercise habits

Questions: "Have you thought about starting something for health? What was the obstacle?"

Interview Results:

Hypothesis A (Exercise Habits) Validation: - 23 people "want to exercise" - Obstacles: "Don't know what to do" "Can't continue" - Deep dive: Could continue with specific goals like "10,000 steps daily" - Validation result: Promising

Hypothesis B (Diet Management) Validation: - 18 people "want to improve diet" - Obstacles: "Recording every meal is tedious" "Calorie calculation is difficult" - Deep dive: Voices saying "if there was easy recording method" - Validation result: Somewhat promising (but recording effort is challenge)

Hypothesis C (Sleep Improvement) Validation: - 12 people "want to improve sleep quality" - Obstacles: "Don't know specifically what to do" - Deep dive: Can measure sleep time but improvement methods unclear - Validation result: Need exists but solution unclear

Phase 3: MVP Design (3 days)

We focused on most promising Hypothesis A and designed MVP.

MVP Definition: "Minimum features for people wanting exercise habits to achieve daily step goals and continue"

Features to Include in MVP (only 3): 1. Automatic step recording (using smartphone sensors) 2. Daily goal setting (default 8,000 steps) 3. Achievement status visualization (simple pie chart)

Features NOT to Include in MVP (124 features): - Calorie calculation, meal logging, sleep analysis, heart rate monitoring, friend competition, medical institution linkage... - Reason: Unnecessary for validating value hypothesis

Development Period: 2 weeks
Development Cost: 8 million yen (1% of conventional)

David asked anxiously.

"Can we call it a product with just 3 features?"

"This isn't a product, it's an experiment. We validate whether users truly feel value at minimum cost."

Chapter 4: Measurement and Learning—The Truth Data Teaches

Phase 4: MVP Release and Measurement (4 weeks)

We provided the MVP developed in 2 weeks to 500 test subjects.

Measurement Indicators (Build-Measure-Learn): - Retention rate (after 1 week, 2 weeks, 4 weeks) - Goal achievement rate (days achieving 8,000 steps) - Qualitative feedback (weekly surveys)

Results after 4 weeks:

Retention Rate: - After 1 week: 78% (127-feature version was 12%) - After 2 weeks: 62% - After 4 weeks: 51%

Goal Achievement Rate: - 3+ days/week achievement: 68% - 5+ days/week achievement: 42%

User Voices: - "Simple and understandable. Seems sustainable" (85%) - "Want more features" (38%) - Top request: Friend sharing feature - 2nd: Past record graphs - 3rd: Achievement notifications

Learning: 1. Hypothesis was correct - Need to "establish exercise habits" truly exists 2. Simplicity is value - Retention rate 6x by focusing features 3. Next hypothesis emerged - Users want "sharing with friends"

Phase 5: Next MVP Based on Learning (2 weeks)

Based on learning, we designed the next MVP.

MVP v2 (2 additional features): 1. Friend invitation and step sharing 2. Past 7-day record graph

Development Period: 2 weeks
Cumulative Development Period: 1 month (6% of conventional 18 months)

Phase 6: MVP v2 Validation (4 weeks)

Results: - Retention rate (after 4 weeks): 51% → 72% - Friend invitation usage rate: 68% - Average invitations per person: 2.3 - Viral coefficient: 0.68 × 2.3 = 1.56 (1 person invites 1.56 people)

New Learning: - Sharing with friends is key to continuation - "Encouragement" more important than competition - Next hypothesis: "Group challenges" enhance value

Phase 7: Gradual Feature Addition (6 months)

We repeated Build-Measure-Learn, adding features based on learning.

Added Features (priority order): 1. Friend group step challenges (retention +15%) 2. Achievement praise messages (satisfaction +0.8) 3. Weekly reflection reports (goal achievement +12%) 4. Simple meal logging (photo only) (23% usage) 5. Sleep time recording (18% usage)

Features NOT Added (learned unnecessary): - Heart rate monitoring, calorie calculation, medical institution linkage, medication management...

Final Product after 6 months: - Implemented features: 15 features (12% of initial 127) - Development period: 6 months (33% of initial 18 months) - Development cost: 120 million yen (15% of initial 800 million yen)

Results:

User Metrics: - Downloads (6 months): 120,000 - 4-week retention rate: 72% - Monthly active users: 68,000 - NPS (Net Promoter Score): +58

Business Metrics: - Premium subscription rate: 18% - Monthly revenue: 24 million yen - Investment recovery: Expected in 5 months

Comparison with Original 127-feature Version: - Development cost: 120 million yen vs 800 million yen (85% reduction) - Retention rate: 72% vs 12% (6x) - Active users: 68,000 vs 320 (over 200x)

Chapter 5: The Detective's LEAN Diagnosis—Everything Except Value is Waste

Holmes compiled the comprehensive analysis.

"David, the essence of Lean is 'learning.' Perfect plans are illusions. You can't know what users want without building, measuring, and learning. Build small, fail fast, keep learning. Only that repetition creates truly valuable things."

Final Report after 12 months:

Aurora HealthTech was evaluated as "the most user-loved app" in the North American healthtech market.

Final Achievements: - Downloads: 380,000 - Monthly active users: 220,000 - Retention rate (3 months): 68% (industry average 12%) - Annual revenue: 1.2 billion yen - Investment round: Raised 2.5 billion yen in Series A

David's letter contained deep gratitude:

"Through Lean Startup, we transformed from 'a company creating perfection' to 'a company continuously learning.' Most important was understanding it's not the number of features but the magnitude of value. Now we always experiment small-scale before adding new features. We implement only features data shows as 'valuable.' We understood features that don't generate value are all waste."

The Detective's Perspective—Perfection is Evil, Learning is Good

That night, I contemplated the essence of value creation.

The true value of Lean Startup lies in humility. We can't predict the future. We can't perfectly understand what users want. Then we should bet small, learn fast, and correct direction.

Rather than spending 18 months aiming for perfection, building imperfectly in 2 weeks and improving for 16 months generates far more value.

"Everything except value is waste. And only users can teach us what value is."

The next case will also depict a moment when learning and experimentation open a company's future.

"Perfect plans are beautiful. But imperfect execution has far more value"—From the detective's notes

Related files