Part One: Current Pain Points and Underlying Logic
Each month, you spend 3,000 TWD on vitamins, B-complex, collagen, OPC grape seed, and probiotics, diligently maintaining this routine for three months, six months, or even a year, yet you see no results. Your skin does not improve, fatigue persists, and muscle mass remains unchanged. The issue lies not in the quality of the supplements but in a fundamental engineering mistake: blind investment without a feedback loop.
This represents a classic “black box system” problem. The logic followed by most consumers is: advertisements claim effectiveness → purchase → continuous use → hope for results. Throughout this process, there is zero data, zero measurement, and zero personalization. In engineering terms, this means: no sensors, no monitoring points, and no feedback control mechanisms.
Underlying Issue One: Individual Variability in Bioavailability
A fact often overlooked by consumers in nutrition is that the absorption rate of the same nutrient can vary by 5 to 10 times among different individuals. This is not an exaggeration from advertisements but a physical fact determined by genetics, gut microbiome composition, and metabolic enzyme activity.
For instance, the metabolism of vitamin D involves the CYP2R1 and CYP24A1 genes. Some individuals may inherently express these genes insufficiently; thus, even with high doses of vitamin D supplements, their serum D levels rise slowly. The absorption of vitamin B12 requires the presence of intrinsic factor, which significantly declines in the elderly and those with gastrointestinal disorders, resulting in an oral absorption rate of only 5% to 30%.
Iron, calcium, and zinc face similar issues. Your genetics determine your absorption ceiling, rendering the advertised “high absorption” claims meaningless for certain individuals.
Underlying Issue Two: Gut Microbiome Imbalance Leading to Metabolic Disorders
The second major issue is the gut microbiome. The proteins, fibers, and polyphenols you consume must be fermented and broken down by gut bacteria to be effectively utilized. Modern diets are generally characterized by an imbalance in gut microbiota—long-term high sugar, high fat, and low fiber intake, coupled with antibiotic use, has led to a significant decline in beneficial bacteria.
The result? No matter how high-quality the supplements are, they cannot be broken down or absorbed, ultimately leading to mere bowel movements. Essentially, the money spent is purchasing “expensive feces.”
Underlying Issue Three: Lack of Personalized Dosage and Combinations
The supplement market operates under an unspoken rule—products must cater to “the majority,” leading to standardized dosages. However, nutritional needs are highly personalized: a 50 kg woman and a 90 kg man have different vitamin D requirements; a person with normal kidney function and one with impaired kidney function have entirely different potassium intake standards.
Moreover, many individuals make the mistake of stacking supplements blindly. Taking calcium, iron, zinc, and vitamin C simultaneously leads to competition for absorption in the gut, resulting in decreased efficiency for each. This is akin to running 100 threads on a server competing for the same resource, ultimately reducing overall throughput.
Part Two: Technical Architecture of AI Automation Solutions
Step One: Establishing a Personal Metabolic Profile (Genetic + Biochemical Testing)
The traditional approach is to assertively say, “Trust me.” The scientific approach involves conducting a comprehensive personal metabolic baseline scan. This includes three components:
- Genetic Testing: Focus on SNP loci related to nutrient metabolism (CYP2R1, MTHFR, ACE, VDR, etc.), costing approximately 500 to 2,000 TWD, and valid for a lifetime.
- Biochemical Testing: Fasting blood glucose, lipid levels, B12, D, ferritin, homocysteine, and 20 to 30 other indicators, costing around 1,500 to 3,000 TWD to establish a baseline.
- Gut Microbiome Testing: 16S ribosomal RNA sequencing to understand your beneficial bacteria ratio, costing about 800 to 1,500 TWD.
The total cost is approximately 3,000 to 6,500 TWD, a one-time investment. In contrast, the traditional method involves spending 3,000 TWD monthly on random purchases, making this investment far more worthwhile.
Step Two: AI Model Building for Personalized Recommendation Engine
With baseline data in hand, AI has the “training material” it needs. The logic is straightforward—using machine learning models (such as gradient boosting machines or neural networks) to learn the mapping relationship between “your genes + your microbiome + your lifestyle” and “optimized nutritional formulations.”
The system will output:
- Which nutrients you need (not everyone is deficient in D, iron, or calcium; some may even have excess)
- Personalized dosages for each nutrient
- Optimal timing and combinations for intake (which nutrients absorb better together, which should be taken separately)
- Expected time to see effects and target indicators
This engine can be built using open-source frameworks (XGBoost, LightGBM) or cloud AI services (Azure ML, AWS SageMaker), with costs depending on scale, but for personal use, monthly costs should not exceed 100 TWD.
Step Three: Automated Feedback and Dynamic Adjustment
The killer feature of the AI solution is “continuous learning.” Every 4 to 8 weeks, users perform a simple retest (using at-home blood testing kits), and the system automatically tracks changes in core indicators. It then retrains the model with the latest data, automatically fine-tuning the nutritional plan.
This is akin to “Continuous Integration and Continuous Deployment” (CI/CD) in DevOps—not a one-time deployment but an ongoing iterative optimization process.
The entire process can achieve over 95% automation. Users only need to periodically upload testing data and complete a simple questionnaire (regarding sleep, stress, and exercise), with the system automatically generating new supplementation plans and pushing them to the app.
Part Three: Business Model and Revenue Expectations
Option One: B2C Health Profile Management SaaS (Subscription Model)
This targets individual users with a monthly fee structure. Services include: genetic testing guidance, AI recommendation engine, dynamic adjustments, and data dashboards.
- Monthly fee: 299 to 999 TWD (depending on feature richness)
- User target: reach 100,000 users within 1 to 5 years
- Annual revenue: 299 TWD × 100,000 users × 12 months = 358.8 million TWD (conservative estimate)
- Gross margin: SaaS typically ranges from 70% to 85%
Option Two: Collaboration with Supplement Brands (B2B Alliance)
The challenge for traditional supplement manufacturers is low conversion and repurchase rates—users feel no effect, leading to poor word-of-mouth. By integrating with the AI system, they can offer “personalized recommendation plans,” potentially increasing conversion rates by 3 to 5 times and repurchase rates from 30% to 40% up to 60% to 80%.
Business model:
- Charge supplement manufacturers a commission (8% to 15% of the recommended sales)
- Or directly license the technology, charging a monthly fee (50,000 to 200,000 TWD)
Assuming collaboration with 50 small to medium-sized supplement brands, with an average annual revenue of 20 million TWD each and a commission rate of 10%, the revenue would be: 50 brands × 20 million TWD × 10% = 100 million TWD annually.
Option Three: Corporate Health Management Platform (B2B2C)
Large enterprises, gyms, and clinics are willing to pay for personalized nutrition plans for their employees/members. A white-label system can be established and licensed to these organizations.
- Licensing fee: 50,000 to 500,000 TWD per organization (based on user count)
- Target: 100 organizations onboarded, with an average monthly fee of 100,000 TWD
- Annual revenue: 100 × 100,000 TWD × 12 = 120 million TWD (1.2 billion)
Part Four: Execution Challenges and Competitive Moat
The largest challenge is not the technology but the volume of data. The accuracy of the model depends on the amount of training data. You need at least 10,000 to 50,000 real users’ complete cycle data on “genes + tests + supplementation plans + effects” to train a good model.
This time cost is 24 to 36 months. Initially, one must accept the model’s imperfections, accumulating data while operating and iterating improvements.
The competitive moat lies in:
- Exclusive ownership of user health data assets (which are difficult to replicate)
- AI models that have been optimized through numerous iterations (far exceeding initial accuracy)
- Deep collaborative relationships with supplement and medical institutions
Conclusion: From Blind Supplementation to Scientific Optimization
The fundamental reason for the ineffectiveness of supplements lies not in the products themselves but in the decision-making process. The traditional approach is “advertisement-driven → blind purchasing → hoping for results.” The scientific approach is “data collection → AI analysis → personalized recommendations → continuous validation.”
This illustrates the power of automation—replacing human intuition and luck with the cold precision of machines. Once the system is established, the marginal cost of scaling to 100,000 users is nearly zero, yet it can yield a 100,000-fold scale effect.
If you are still blindly stacking supplements, you are merely paying for the precise targeting of advertisers. By using AI to reconstruct this decision-making process, you are tailoring solutions to your genetics and gut microbiome. The speed of results determines your confidence; confidence dictates your actions, and actions determine your health.
Leave a Reply