1. Current Pain Points
Most consumers face two structural issues when selecting their first serum. The first is information overload without a decision-making model: The market is flooded with various ingredient claims, brand stories, and KOL recommendations, yet no one provides a logical framework for selection. The second issue is opaque trial and error costs: A single serum can easily start at two thousand, and purchasing the wrong one results not only in financial loss but also potential skin sensitivity due to incompatible active ingredients, leading to further time and budget spent on recovery.
From a business perspective, brands and distributors often concentrate their marketing budgets on high-margin star ingredients, such as the peptides, hyaluronic acid, and retinol that have gained popularity in recent years. However, these ingredients can be too concentrated or irritating for novices who have never used serums before. Consequently, consumers end up purchasing well-reviewed products but abandon them due to insufficient skin tolerance, resulting in wasted inventory. This cycle lacks a layered guidance mechanism: No one is at the forefront of the sales funnel to help beginners establish “safe entry” selection criteria.
Another overlooked pain point is the low level of structured product information. Most e-commerce platforms or beauty community review systems only provide star ratings and scattered user experiences, making it difficult to quickly find “which products have a pH value between 5-6,” “which brands have relatively mild preservative systems,” or “which ingredient combinations are suitable for sensitive skin as a first trial.” The absence of such structured data forces consumers to rely on luck or blindly follow trends rather than making precise matches based on their skin conditions and needs.
2. Decomposing the Underlying Logic
To address the selection problem, it is essential to break down what constitutes “gentle” into quantifiable parameters. From a formulation engineering perspective, a serum suitable for beginners should meet three criteria: low irritability, high stability, and clear yet moderate efficacy.
Low irritability can be assessed from two dimensions: first, the concentration range of active ingredients. For example, vitamin C derivatives (such as MAP, SAP) are gentler than pure L-ascorbic acid because they penetrate more slowly and have looser pH requirements. Second, the preservative and solvent systems are crucial; while parabens are stable, they may irritate some sensitive skin types. In contrast, formulations using polyols (such as pentylene glycol, hexylene glycol) as preservative enhancers tend to be milder.
High stability relates to packaging design and storage conditions. Vacuum pump bottles and opaque glass containers effectively reduce the oxidation of active ingredients, extending product shelf life. If a serum noticeably changes color or separates within three months of opening, it indicates insufficient formula stability, presenting an additional barrier to entry for beginners.
Clear yet moderate efficacy means not pursuing multiple claims in the first bottle. Beginners should prioritize products with single or dual functions, such as “moisturizing + repairing” or “brightening + antioxidant,” rather than a complex formula that claims “anti-aging + whitening + firming + oil control” all at once. The latter’s complex ingredient stacking increases the likelihood of intolerance.
From a data perspective, a three-tier scoring model can be established: the first tier is the ingredient safety score (based on EWG or CIR databases), the second tier is the formula gentleness score (based on pH value, types of penetration enhancers, preservative systems), and the third tier is user feedback on tolerance (extracted from reviews by analyzing the frequency of keywords like “stinging,” “redness,” “peeling”). By cross-referencing these three dimensions, a list of products truly suitable for beginners can be filtered out.
3. AI Automation Solutions
Executing this selection logic manually would consume significant time on ingredient research, cross-referencing, and review analysis. However, by utilizing AI automation stacking, the entire process can be compressed to a matter of minutes.
The first step involves API integration with ingredient databases. Platforms like CosDNA and EWG Skin Deep offer public or semi-public data interfaces that can be accessed via web scraping or APIs to obtain ingredient lists, safety scores, and irritation indicators. Next, an NLP model (such as GPT-4 or a fine-tuned version of BERT) can be employed to structurally extract information from product descriptions and official website content, converting descriptors like “suitable for sensitive skin,” “fragrance-free,” and “dermatologist-tested” into comparable labels.
The second step is sentiment and keyword analysis of review texts. By scraping user reviews from e-commerce platforms (such as Shopee, Momo, or PTT Beauty), an AI model can identify the frequency of negative keywords like “stinging,” “allergy,” and “breakouts,” calculating each product’s tolerance risk index. This index can serve as a filtering criterion, allowing the system to automatically exclude high-risk products.
The third step involves a personalized recommendation engine. Users only need to input three parameters: skin type (dry/oily/combination/sensitive), primary concern (moisturizing/brightening/anti-aging), and budget range. The system can then automatically filter the top five recommended products from the database, providing ingredient analysis, safety scores, and contextual usage explanations for each product. This entire process can be embedded in a LINE official account or web chatbot, enabling consumers to complete their selections through conversation.
From a technical stack perspective, a Python + FastAPI backend service can be established, paired with PostgreSQL for storing structured ingredient data, while the frontend can utilize React or Vue.js for interactive interfaces. For SEO-driven traffic, each recommended product can automatically generate a static review page, utilizing Next.js or Nuxt.js for server-side rendering, allowing search engines to index detailed analyses of each product.
4. Revenue Expectations
Once this system is launched, revenue can be generated from three channels. The first is affiliate marketing commissions: each time a sale is made through a referral link, a commission of 5%-15% can be earned. Assuming 1,000 clicks per month with a conversion rate of 3% and an average order value of 1,500, with a commission rate of 10%, the monthly revenue would be approximately 4,500. If traffic sources are expanded to include SEO, social media, and LINE official accounts, there is potential to increase monthly revenue to over 15,000 within three months.
The second channel is paid consulting services. For consumers who prefer not to conduct their own research, a “one-on-one product analysis report” service can be offered, charging 300-500, providing personalized ingredient analysis, product comparison charts, and usage sequence recommendations. If 20 clients are served monthly, this could yield an additional income of 6,000-10,000. This aspect can utilize Google Forms or Typeform to collect demands, with AI automatically generating draft reports, which can then be fine-tuned manually, keeping the production cost for each report under 30 minutes.
The third channel involves content collaboration and data licensing with brands. Once sufficient user behavior data is accumulated (e.g., which ingredient combinations are most accepted by sensitive skin, which price ranges have the best conversion rates), brands will be willing to pay for these insights to optimize product development or marketing strategies. Such collaborations typically range from 30,000 to 100,000 per instance, depending on the sample size and depth of analysis.
In terms of time costs, the initial setup of the entire system will require approximately 40-60 hours, including database establishment, API integration, and frontend interface development. After launch, only 2-3 hours per week will be needed to update product data and optimize recommendation logic, with the rest operating automatically. If this model is replicated across other categories (such as sunscreen, cleansing oils, masks), the marginal cost will decrease, while revenue scales can grow linearly.
Leave a Reply