01: Problem Statement
The gap between wellness intent and safe purchasing
Wellness ecommerce customers regularly purchase products that are incompatible with their health conditions, medications, or sensitivities: not out of carelessness, but because product pages do not surface this information meaningfully. A customer managing a chronic condition has no reliable way to filter a product catalogue against their health profile before purchasing. This creates risk (adverse reactions, wasted spend, eroded trust) and missed opportunity (higher confidence purchases, increased basket size, better retention). For wellness brands, it also represents an untapped differentiator: most competitors treat all customers as identical.
02: Goals
What success looks like
G1
Reduce health incompatible purchases: measurable reduction in returns and complaints citing ingredient conflicts within 90 days of launch
G2
Increase conversion on personalised recommendations: users who complete the health intake convert at a higher rate than those who browse without it (target: 20% lift within 60 days)
G3
Improve basket confidence: users who receive a personalised recommendation report higher purchase confidence vs. generic browse (target: 70% positive via post purchase survey)
G4
Prove rule-based matching feels personal, not generic: recommendation accuracy rated as relevant or very relevant by 80% of users in the first 30 days
G5
Support dual audience positioning: product is legible and compelling to both wellness/ecommerce audiences and healthtech stakeholders as health aware commerce infrastructure
03: Non Goals
What this is not
Clinical diagnosis or medical advice. WellMatch surfaces compatibility signals only. This is both a regulatory constraint and an intentional product boundary.
EHR/EMR integration. User health data is self reported via intake only. Connecting to clinical health records is out of scope for v1.
Prescription medication lookup. v1 uses a curated list of common conditions and sensitivities, not a full pharmacopeia. Drug to ingredient interaction logic is v2.
Real time inventory sync. Product catalogue is seeded manually or via static import for v1. Live Shopify inventory sync is a v2 integration.
User accounts and persistent profiles. v1 uses session based intake only. Saved health profiles require auth infrastructure and are a v2 feature.
04: User Personas
Who this is built for
Primary · Consumer
The Health Conscious Shopper
A wellness consumer aged 25 to 55 who actively manages their health. They research products carefully but find generic product pages insufficient. They want confidence, not just choice.
Would use health aligned product tool
57%
of consumers say they would use an app ensuring products align with their health priorities (NIQ, 2025)
Primary · Consumer
The Care Adjacent Buyer
A family member or caregiver purchasing wellness products on behalf of someone with health conditions. Stakes are higher. Expertise is lower. They need a filter, not a catalogue.
Adults aged 45 to 64 who are caregivers
24.4%
of adults 45 to 64 are caregivers and among the most powerful decision makers in wellness product purchasing
Secondary · B2B
The Wellness Brand Operator
A DTC wellness brand (Saje, Thrive, similar) evaluating WellMatch as a site layer that improves conversion and reduces returns. They care about integration simplicity and measurable lift.
Wellness beauty market annual sales
$200B
in wellness oriented beauty sales in the past year, growing 4% year over year (Circana, 2025)
Secondary · B2B
The Healthtech Evaluator
A product lead at a health platform (League, Dialogue, Maple) assessing WellMatch as infrastructure for a health aware commerce module. They care about the API, data architecture, and defensibility.
Prefer tech with health and wellness features
74%
of consumers prefer a tech product with health and wellness features over one without (NIQ, 2025)
05: User Stories
What each persona needs to accomplish
Health Conscious Shopper
As a health conscious shopper, I want to enter my conditions and sensitivities so that I only see products that are safe and relevant for me
As a health conscious shopper, I want to understand why a product is recommended so that I feel confident purchasing rather than just trusting an algorithm
As a health conscious shopper, I want to update my health profile mid session so that I can explore different scenarios, like gifting for someone else
Care Adjacent Buyer
As a caregiver, I want to filter products by the health conditions of the person I am buying for so that I do not accidentally purchase something incompatible
As a caregiver, I want a simple, non clinical intake experience so that I do not need medical expertise to get useful results
Wellness Brand Operator
As a brand operator, I want to embed WellMatch as a widget on my Shopify storefront so that customers get personalised recommendations without leaving my site
As a brand operator, I want to see aggregate data on the most common health profiles my customers report so that I can inform product development and content strategy
Healthtech Evaluator
As a healthtech product lead, I want to query the WellMatch recommendation engine via API so that I can surface compatible wellness products within my existing health platform
As a healthtech evaluator, I want documentation of how health intake data is handled so that I can assess regulatory and privacy compliance
06: Persona Research
Evidence base for each persona
Persona 1 · Health Conscious Shopper
Market validated demand
57% of consumers say they would use an app or screening device that ensures products align with their personal health priorities (NIQ, 2025). This is direct demand validation for WellMatch's core proposition.
The CDC reports 6 in 10 US adults have at least one chronic disease, creating a large base of consumers who need health aware product guidance.
Online retail now holds a 47% share of the health and wellness product market (2024), meaning the shopper is already purchasing online. The gap is health aware filtering, not channel adoption.
Online conversations about AI health tools increased 50% from 2023 to 2024, but 90% of sentiment is negative: consumers want health guidance but do not trust AI giving medical advice. WellMatch stays within the safer product filtering boundary.
In 7 years of clinical care, I regularly observed patients and families attempting to navigate wellness product choices without guidance, often making purchases that conflicted with medications or conditions. The problem is real and recurring.
Persona 2 · Care Adjacent Buyer
A primary purchasing segment, not a niche
24.4% of adults aged 45 to 64 are caregivers, and 1 in 4 women are caregivers (Credence Research, 2024). This is a substantial purchasing segment, not an edge case.
Research confirms that family caregivers have emerged as the most powerful decision makers in senior care product purchasing, making consequential product decisions on behalf of others.
Family members, friends, and caregivers play a crucial role in shaping the online purchasing behaviour of older customers, confirming this persona drives ecommerce volume.
At posAbilities and Villa Cathay, family members regularly asked clinical staff for product guidance before purchasing. They wanted a filter, not a catalogue. None existed.
Persona 3 · Wellness Brand Operator
A growing market actively seeking personalisation
Wellness oriented beauty products generated USD 200 billion in sales in the past year, up 4% year over year (Circana, 2025). The operator market is large and growing.
DTC brands are actively investing in AI driven personalised product recommendations. WellMatch extends this into health safety territory where no existing solution operates.
Online channels gained the most market share in health and wellness between 2020 and 2024. Operators need differentiation in an increasingly crowded online market.
Through LashPals I operated a beauty brand and understand the operator perspective: conversion, return rates, and customer trust are the metrics that matter. WellMatch speaks directly to all three.
Persona 4 · Healthtech Evaluator
Commerce as a natural extension of health platforms
74% of consumers prefer tech products with health and wellness features over those without (NIQ, 2025). Health platforms adding commerce capability meet real consumer demand.
The caregiver app market is projected to grow from USD 1.38 billion (2023) to USD 3.67 billion by 2031, at 15% CAGR. Health platforms are actively building adjacent services.
77% of Americans prioritise their mental health, and beauty is intrinsically tied to emotional wellbeing. Health platform members are also wellness consumers: a product recommendation layer serves both needs.
From clinical experience using PointClickCare, Bossnet, and ShareVision: health platforms that connect users to appropriate products and services create meaningfully better outcomes than those that treat health and commerce as separate domains.
| Persona | Core Need | Key Market Signal |
|---|---|---|
| Health Conscious Shopper | Safe, relevant product discovery | 57% want health aligned product tools (NIQ) |
| Care Adjacent Buyer | Filtering by another person's health needs | 24.4% of adults 45 to 64 are caregivers |
| Wellness Brand Operator | Conversion lift and return reduction | USD 200B wellness beauty market, +4% YoY |
| Healthtech Evaluator | Commerce API for health platform integration | USD 3.67B caregiver app market by 2031 |
07: Requirements
What we are building, in order of priority
P0 Must HaveMVP cannot ship without these
Health Intake FlowUser completes a structured intake covering common conditions, dietary restrictions, and ingredient sensitivities: 5 questions or fewer, skippable, with a progress indicator.
Acceptance Criteria
Given a user opens WellMatch, when they begin intake, then they complete the flow in under 90 seconds on average. Given a user skips intake, when they browse, then the full product catalogue is visible without restriction.Product Compatibility TaggingEach product is tagged with: relevant conditions, compatible dietary profiles, ingredient flags, and a confidence tier (safe, use with caution, consult a professional). Maintained in a structured data layer, not unstructured descriptions.
Acceptance Criteria
Given a product exists in the catalogue, when the API is queried, then structured health compatibility metadata is returned for every product.Recommendation EngineGiven a completed health profile, the engine returns a ranked list of compatible products with plain language explanations. Products flagged "use with caution" are surfaced but visually distinguished, not hidden.
Acceptance Criteria
Given a user with a completed profile, when recommendations are returned, then every product has an associated plain language reason. Given a product is flagged "use with caution", then it is visually differentiated with a clear label.Recommendation Display UIProducts display with: image, name, compatibility badge, and reason snippet. User can toggle between "recommended for you" and "all products". Mobile responsive.
P1 Nice to HaveSignificantly improves experience, not blocking
Ingredient level transparency overlay: user taps any ingredient to see a plain language explanation relevant to their profile
Shareable recommendation summary: user can copy or share their personalised product list without an account
"Shop for someone else" mode: user can switch intake context mid session without restarting
Basic analytics dashboard for brand operators: top conditions reported, recommendation click through rate, conversion by profile type
P2 FutureExplicitly out of scope for v1
Saved health profiles with user authentication
Shopify app store listing for self serve brand integration
Full medication to ingredient interaction logic (requires pharmacopeia integration and clinical review)
EHR/EMR data import via FHIR standard (requires clinical partnership and regulatory review)
White label API for healthtech platform integration
08: Success Metrics
How we know it worked
| Metric | Target | Tool | When |
|---|---|---|---|
| Leading Indicators: measure at 2 and 4 weeks | |||
| Intake completion rate | >65% | Pendo | Week 2, 4 |
| Recommendation click through rate | >40% | Pendo | Week 2, 4 |
| Time to first recommendation | <2 minutes | Session recording | Week 2 |
| Mobile usability score | >80 Lighthouse | Lighthouse CI | Pre launch |
| Lagging Indicators: measure at 60 and 90 days | |||
| Conversion rate lift vs. non personalised | +20% | A/B test | Day 60 |
| Returns citing health incompatibility | Reduction vs. baseline | Support data | Day 90 |
| Post purchase confidence survey | 70% confident | In product survey | Day 60 |
| Recommendation relevance rating | 80% relevant | Post recommendation prompt | Day 30 |
09: Open Questions
What needed to be resolved
| Question | Resolution | Status |
|---|---|---|
| What is the source of truth for ingredient safety data: curated internal list, third party API, or clinical review? | Curated internal list for v1, built from publicly documented ingredient contraindications (e.g. retinoids and salicylic acid in pregnancy, common allergen classes). A commercial launch would require formal clinical review or a licensed third party API before making safety claims. That gate is v2, not v1. | Resolved |
| Should "use with caution" products be shown by default or require an explicit "show anyway" action? | Shown by default with a clear "use with caution" badge, never hidden. Matches the P0 acceptance criteria: caution items are visually differentiated, not suppressed. | Resolved |
| What is the legal review requirement for displaying health related compatibility information? | Not required for a portfolio project. Mitigated with a persistent disclaimer ("compatibility signals only, not medical advice") in the footer and Non-Goals section. A commercial launch would require legal and regulatory review before enabling purchase-blocking or explicit health claims. | Resolved |
| Should the intake be shown as a modal on landing, an opt in widget, or a dedicated page? | Default landing experience rather than a blocking modal, always skippable via a persistent "Skip intake" action, with the full catalogue remaining visible if skipped. | Resolved |
| How granular should condition tagging be: broad categories (autoimmune) or specific diagnoses (lupus, MS)? | Broad categories for v1 (eczema, psoriasis, diabetes, hypertension, pregnancy/breastfeeding, autoimmune condition, etc.), consistent with the "not a full pharmacopeia" non-goal. Diagnosis level granularity is v2, gated on clinical review. | Resolved |
10: Technical Considerations
Stack and architecture (v1)
Frontend
ReactViteTailwind
Recommendation Engine
Deterministic rule-based matcherNo external API calls
Product Data
Ingredient-level JSON (v1)Shopify API (v1.5)
Auth
None in v1Session based only
Hosting
Vercel
Analytics
Pendo
The recommendation engine is a deterministic, rule-based matcher, not an LLM call. Each product is defined by its actual ingredient list; every ingredient carries structured health metadata (conditions to avoid, conditions to flag, conditions it helps, allergen tags, vegan status). The engine scores each ingredient against the user's profile and rolls the result up to a product-level compatibility tier, with plain language rationale generated from the specific ingredients that drove the match.
Why rules instead of an LLM: this is a deliberate choice, not a placeholder for one. When the output is a health compatibility signal, every recommendation needs to trace back to an explicit, reviewable rule, not a probabilistic generation that could vary between requests or state something the underlying ingredient data doesn't actually support. A rules engine can't hallucinate a contraindication that isn't in the data; a language model can. It also runs free, at any volume, with no per-request latency or cost, which matters once a widget like this is embedded across every product page on a brand's storefront. The honest tradeoff: rules don't scale as gracefully to a thousand-SKU catalogue or handle free-text health input the way a language model could. For a curated, safety-adjacent recommendation surface, correctness and auditability mattered more than that flexibility.
11: Timeline
Phased delivery to demo ready MVP
0
Phase 0 · Data Foundation
Build product catalogue seed data
Build a health-metadata database of individual ingredients (avoid/caution/good-for conditions, allergen tags, vegan status), then compose 20 seed products from real ingredient lists rather than flat per-product tags. This is the foundation everything else sits on.
1 week
1
Phase 1 · MVP
Intake flow + recommendation engine + display UI
Build the health intake (5 questions), a rule-based ingredient-matching engine for recommendation logic, and the recommendation display with compatibility badges and plain language explanations.
2 to 3 weeks
2
Phase 2 · Polish
Ingredient overlay, mobile optimisation, operator analytics
Add ingredient level transparency, optimise for mobile, and build the basic operator analytics view. This phase makes the MVP demo ready for both consumer and B2B audiences.
1 to 2 weeks
3
Phase 3 · Integration
Shopify product sync + embeddable widget
Connect to the Shopify API for live product data and build the embeddable widget format for brand operator integration. This phase makes WellMatch a real B2B product, not just a demo.
2 weeks
Total estimated time to demo ready MVP: 3 to 4 weeks
12: Portfolio Context
Why this project exists
WellMatch is a portfolio project demonstrating the intersection of ecommerce product management and healthtech product thinking. It reflects real problems observed across both domains: wellness shoppers making uninformed purchases, and health platforms lacking commerce infrastructure that respects clinical context. The project is designed to be legible and compelling in interviews at wellness brands (Saje, Thrive), ecommerce platforms (Cymax, Shopify merchants), and health technology companies (League, Dialogue, Maple, PointClickCare).