🚀 IRVINS Vibe Coding Guidebook

AI-Powered Development Best Practices & Prompt Engineering for 2026

Transform Business Functions with Strategic AI Development

📘 What is Vibe Coding?

Vibe coding is the practice of building production-ready software through conversational AI interactions, using natural language prompts instead of traditional coding. It's not about writing less code—it's about thinking in products and letting AI handle implementation details while you focus on business logic and user experience.

Why This Matters for IRVINS

As we scale across markets (Singapore, USA, China) and channels (retail, e-commerce, modern trade), speed and consistency in tool development become competitive advantages. Vibe coding enables:

  • Rapid prototyping for new market features (1 day vs 1 week)
  • Cross-functional empowerment (commercial team builds their own dashboards)
  • Consistent patterns across markets and functions
  • Faster iteration cycles on AI transformation initiatives
💡 Golden Rule: Vibe coding works best when you combine clear business requirements with structured prompting. Ambiguity produces generic output; specificity produces production-ready tools.

🎯 6 Core Principles for Production Vibe Coding

1. Think in Products

Every prompt should describe a complete feature or user workflow, not just isolated code.

❌ Bad: "Create a sales chart"

✅ Good: "Build a Singapore sales dashboard showing weekly revenue by channel (retail, e-commerce, MT) with MTD vs target variance and export to PDF"

2. Integrated Stack

Specify your complete tech environment upfront to avoid compatibility issues.

For IRVINS: Google Sheets API, JavaScript/HTML, Tailwind CSS, Chart.js, FastAPI backends

3. Test as You Build

Request test cases and edge case handling in the same prompt that creates the feature.

Example: "Include validation for negative values, missing dates, and zero quantities"

4. Git Safety Net

Version control every iteration. Before major changes, create a checkpoint you can roll back to.

Habit: Commit after each successful feature addition

5. Full-Stack First

Design for both frontend and backend from the start, even if you're only building one initially.

Why: Prevents architectural rewrites later

6. Deploy Fast

Build deployment-ready from iteration 1. Use platforms that support one-click deployment (Vercel, Netlify, GitHub Pages).

Goal: Live test link within 1 hour of starting

📝 Prompt Engineering Templates

Anatomy of a Production-Ready Prompt

Every effective prompt contains these 7 elements:

  1. Role/Context: Who is the AI acting as? (Senior Data Engineer, Frontend Developer, etc.)
  2. Tech Stack: Specific technologies, libraries, versions
  3. Goal/Outcome: What user problem are we solving?
  4. Business Logic: IRVINS-specific rules, calculations, constraints
  5. Edge Cases: Error handling, validation, security
  6. Existing Code Context: What already exists that must integrate?
  7. Deliverables: Exact files, functions, or features expected

Master Template

TEMPLATE You are a [ROLE] for IRVINS. **Tech Stack:** [List technologies] **Goal:** [User problem + business outcome] **Business Context:** - [IRVINS-specific rule 1] - [IRVINS-specific rule 2] - [Data source/format] **Requirements:** 1. [Functional requirement 1] 2. [Functional requirement 2] 3. [UI/UX requirement] **Edge Cases:** - [Error scenario 1] - [Validation rule 1] - [Security concern] **Existing Code:** [Paste relevant code or describe integration points] **Deliverables:** - [File 1 with specific features] - [File 2 with specific features] - [Documentation/tests if applicable] **Initial Prompt:** "[One-sentence summary of what to build]"

Basic: Simple Data Transformation

BASIC PROMPT You are a data processing assistant for IRVINS. **Goal:** Convert weekly sales CSV to monthly summary **Input Format:** - CSV columns: date, market, channel, sku, units, revenue - Date format: YYYY-MM-DD **Output Requirements:** 1. Group by month and market 2. Sum units and revenue 3. Calculate average selling price (revenue/units) 4. Export as JSON **Edge Cases:** - Handle missing values (treat as 0) - Skip rows with invalid dates - Round ASP to 2 decimals **Deliverable:** Python script (process_sales.py)

Intermediate: Interactive Dashboard

INTERMEDIATE PROMPT You are a frontend developer building analytics for IRVINS. **Tech Stack:** HTML, JavaScript, Chart.js, Tailwind CSS **Goal:** Build TPM (Trade Promotion Management) ROI calculator **Business Logic:** - Baseline sales = average of 4 weeks pre-promo - Incremental units = promo period units - (baseline × promo weeks) - ROI % = ((incremental revenue - promo cost) / promo cost) × 100 - Break-even threshold = promo cost / (ASP - variable cost) **UI Requirements:** 1. Input fields: promo cost, promo weeks, discount %, historical sales data 2. Auto-calculate baseline from data input 3. Display: incremental volume, revenue lift, ROI %, break-even units 4. Color-code ROI (green >20%, yellow 10-20%, red <10%) 5. Export results as PDF **Edge Cases:** - Validate promo cost > 0 - Prevent division by zero - Alert if insufficient historical data (<4 weeks) **Deliverable:** Single HTML file with embedded CSS/JS

Advanced: Full-Stack Microservice

ADVANCED PROMPT You are a senior backend engineer for IRVINS. **Tech Stack:** FastAPI, PostgreSQL, SQLAlchemy, pytest, Docker **Goal:** Build market expansion readiness API **Business Context:** - IRVINS evaluates new markets using 6 criteria: market size, competition, regulatory, logistics, brand fit, ROI projection - Each criterion scored 1-5 - Minimum threshold: 24/30 to proceed - Weightings: market size (2x), ROI (1.5x), others (1x) **API Endpoints:** 1. POST /markets/evaluate - Accept market data, calculate weighted score 2. GET /markets/{market_id}/report - Return full assessment 3. GET /markets/compare?ids=[] - Compare multiple markets side-by-side 4. PUT /markets/{market_id}/update - Update scores **Data Model:** - Market: id, name, country, region, status - Evaluation: market_id, criterion, score, notes, evaluated_by, date - WeightedResult: market_id, total_score, recommendation, created_at **Security:** - JWT authentication - Role-based access (analyst can read, manager can write) - SQL injection prevention (parameterized queries) - Input validation (scores 1-5 only) **Edge Cases:** - Handle partial evaluations (not all criteria scored) - Prevent duplicate evaluations on same date - Archive old evaluations after market decision **Testing:** - Unit tests for weighted scoring logic - Integration tests for all endpoints - Test data fixtures for 3 markets **Deliverables:** - app.py (FastAPI routes) - models.py (SQLAlchemy models) - services.py (business logic) - schemas.py (Pydantic validation) - tests/test_markets.py (pytest suite) - README.md (API documentation) **Initial Prompt:** "Build market expansion evaluation microservice as specified above."

🏢 IRVINS Function-Specific Examples

Commercial Function

Example 1: Channel Performance Dashboard

Sales AnalyticsMulti-Channel

COMMERCIAL Build a Singapore channel performance dashboard. **Data Source:** Google Sheets API (sales_data_sg tab) **Channels:** Retail (own stores), E-commerce (direct), Modern Trade (NTUC, Cold Storage, Sheng Siong), Marketplaces (Shopee, Lazada, TikTok Shop) **Metrics:** - Weekly revenue by channel - Channel mix % (of total) - YoY growth rate by channel - Average basket size by channel **Visualizations:** 1. Stacked area chart: revenue trend (8 weeks) 2. Pie chart: current week channel mix 3. Table: top 5 SKUs per channel **Filters:** Date range, include/exclude channels **Business Rules:** - Marketplace revenue includes platform fees deduction - MT revenue is sell-in (to retailer), not sell-out - E-commerce includes shipping revenue **Tech:** HTML, Chart.js, Google Apps Script for data fetch

Example 2: Modern Trade Negotiation Simulator

Pricing StrategyWhat-If Analysis

COMMERCIAL Create a MT partnership negotiation calculator. **Goal:** Model P&L impact of different retailer terms **Inputs:** - Baseline: monthly volume, wholesale price, variable cost - Negotiation terms: listing fee, promotional support %, rebate %, payment terms (days) **Calculations:** 1. Gross margin = (wholesale price - variable cost) / wholesale price 2. Net margin = gross margin - (listing fee/volume) - promo support % - rebate % 3. Working capital impact = (volume × wholesale price × payment days) / 30 4. Break-even volume = fixed costs / net margin per unit **Scenarios:** Compare 3 retailers side-by-side with different terms **Output:** Recommendation (proceed/negotiate/decline) based on net margin threshold (>18%) **Visual:** Waterfall chart showing margin breakdown

Operations Function

Example 3: Production Planning Optimizer

Supply ChainForecasting

OPERATIONS Build production schedule optimizer for IRVINS factory. **Inputs:** - Demand forecast (next 4 weeks by SKU) - Current inventory levels - Production capacity (kg/day by production line) - Shelf life by SKU (days) - Minimum batch size (kg) **Business Logic:** 1. Priority = (forecast - inventory) / shelf life (produce urgent items first) 2. Production quantity = MAX(forecast + safety stock - inventory, minimum batch) 3. Safety stock = 7 days of average demand 4. Line allocation based on SKU type (salted egg vs salmon skin different lines) **Constraints:** - Don't exceed daily capacity - Avoid stock-outs (inventory > 3 days demand) - Minimize waste (produce only what's needed within shelf life) **Output:** - Daily production schedule (4 weeks) - Projected inventory levels - Alerts for potential stock-outs or excess **Tech:** Python script with pandas, output to Google Sheets

Marketing Function

Example 4: Campaign ROI Tracker

Digital MarketingAttribution

MARKETING Create multi-channel marketing ROI dashboard. **Channels:** Meta ads, Google ads, TikTok ads, influencer partnerships, email **Data Sources:** - Ad platform APIs (Meta, Google, TikTok) - Google Analytics for web traffic - Shopify for e-commerce conversions - Manual input for influencer costs **Metrics:** 1. Spend by channel 2. Impressions, clicks, CTR 3. Conversions attributed to each channel 4. CAC (customer acquisition cost) = spend / new customers 5. ROAS (return on ad spend) = revenue / spend 6. LTV:CAC ratio (use 180-day LTV estimate) **Attribution Model:** Last-click (from GA4) **Visualizations:** - Funnel chart: impressions → clicks → conversions by channel - Scatter plot: CAC vs ROAS by campaign - Trend: weekly spend vs revenue **Alerts:** Flag campaigns with ROAS < 2.5 or CAC > $25 **Tech:** JavaScript dashboard with Google Apps Script integration

Finance Function

Example 5: Weekly P&L Flash Report Generator

Financial ReportingAutomation

FINANCE Automate weekly P&L flash report for IRVINS leadership. **Data Sources:** - Revenue: Google Sheets (sales by channel) - COGS: ERP export (production costs) - Opex: Google Sheets (fixed cost tracker) **Calculations:** 1. Gross profit = revenue - COGS 2. Gross margin % = GP / revenue 3. EBITDA = GP - opex (excl. depreciation) 4. EBITDA margin % = EBITDA / revenue **Report Format:** - Week-on-week comparison (W1 vs W2) - MTD actual vs budget variance - YTD actual vs budget variance - Color-coding: green (on/above target), red (below target) **Breakdown:** By market (SG, USA, China), by channel **Distribution:** Auto-email PDF to leadership every Monday 9 AM SGT **Tech:** Python script (schedule with cron), fpdf for PDF generation

Retail Function

Example 6: Store Inventory Reorder Alert System

Inventory ManagementAlerts

RETAIL Build store inventory monitoring and reorder system. **Stores:** Jewel, Takashimaya, Orchard Central, Plaza Singapura **Data:** Real-time inventory from POS system (API or daily CSV) **Reorder Logic:** 1. Reorder point = (daily sales rate × lead time days) + safety stock 2. Daily sales rate = 7-day moving average 3. Lead time = 2 days (warehouse to store) 4. Safety stock = 3 days of sales **Alerts:** - Email store manager when SKU hits reorder point - Escalate to operations manager if stock < 1 day sales - Weekly summary of all reorders **Dashboard View:** - Current inventory levels vs reorder points by store - Stockout risk score (red/yellow/green) - Inventory turnover by SKU **Edge Cases:** - New SKUs with no sales history (use category average) - Promotional periods (increase safety stock to 5 days) **Tech:** FastAPI backend, email via SendGrid, web dashboard

E-Commerce Function

Example 7: Marketplace Performance Analyzer

Marketplace AnalyticsMulti-Platform

E-COMMERCE Build unified marketplace performance dashboard. **Platforms:** Shopee SG, Lazada SG, TikTok Shop SG, Amazon US **Metrics Per Platform:** - GMV (gross merchandise value) - Net revenue (after platform fees, shipping subsidies) - Units sold - Average order value - Conversion rate - Return rate - Seller rating / reviews **Comparative Analysis:** - Which platform has highest margin? - Which SKUs perform best on each platform? - Platform fee comparison - Shipping cost as % of GMV **Data Integration:** - Shopee API - Lazada Seller Center export - TikTok Shop analytics CSV - Amazon Seller Central API **Automation:** Refresh data daily at 8 AM SGT **Insights:** - Flag declining seller ratings - Identify underperforming SKUs (low conversion) - Suggest platform-specific promotions **Tech:** Python backend (FastAPI), HTML dashboard, Chart.js

HR / Team Leadership

Example 8: Sales Team Performance Scorecard

Team ManagementKPI Tracking

HR / LEADERSHIP Create sales team scorecard for IRVINS retail team. **Team:** 12 sales associates across 4 stores **KPIs:** 1. Sales target achievement % 2. Average transaction value 3. Units per transaction 4. Customer satisfaction score (from feedback forms) 5. Attendance rate **Weights:** Sales target (40%), ATV (20%), UPT (20%), CSAT (15%), Attendance (5%) **Scoring:** - 100+ points: Exceeds expectations - 80-99: Meets expectations - <80: Below expectations **Features:** - Individual scorecards - Team ranking - Month-over-month trends - Manager comments section **Gamification:** Top 3 performers each month highlighted **Privacy:** Each associate sees only their own scores; managers see all **Distribution:** Auto-generate and email individual PDFs end of each month **Tech:** Google Apps Script for data processing, automated PDF generation

Market Expansion / Strategy

Example 9: New Market Entry Feasibility Model

Strategic PlanningMarket Analysis

STRATEGY Build market entry evaluation framework for IRVINS. **Evaluation Criteria:** 1. Market size (addressable snack market $) 2. Competitive intensity (# of premium snack brands, market share concentration) 3. Regulatory complexity (food import rules, labeling requirements) 4. Logistics feasibility (cold chain availability, lead times) 5. Brand cultural fit (premium positioning acceptance, flavor preferences) 6. Financial viability (5-year NPV, payback period) **Scoring System:** - Each criterion scored 1-5 - Weighted total: market size (2x), financial (2x), brand fit (1.5x), others (1x) - Threshold: 28/35 to proceed to pilot **Data Inputs:** - Market research reports (manual input or API to Euromonitor) - Regulatory database - Logistics quotes - Financial model (revenue projection, capex, opex) **Output:** - Weighted score with breakdown - Side-by-side comparison of up to 5 markets - Recommendation: Go / No-Go / Pilot - Risk assessment summary **Visualization:** Radar chart of 6 criteria for each market **Tech:** Web-based tool (HTML/JS) with export to PDF for board presentations

Product Development

Example 10: NPD (New Product Development) Pipeline Tracker

Product ManagementInnovation

PRODUCT Create NPD pipeline management dashboard for IRVINS. **Pipeline Stages:** 1. Ideation (concept submitted) 2. Feasibility (R&D testing) 3. Costing (financial model) 4. Consumer Testing (taste panels, surveys) 5. Production Trial (factory run) 6. Launch (market introduction) **Data Points Per Project:** - Product name, category, flavor profile - Stage, stage entry date, days in stage - Owner (R&D, commercial, operations) - Target launch date - Estimated margin % - Risk level (low/medium/high) **Views:** 1. Kanban board (drag-drop between stages) 2. Gantt chart (timeline view) 3. Portfolio summary (# products per stage) 4. Risk dashboard (high-risk projects flagged) **Alerts:** - Projects in stage > 30 days (stalled) - Projects at risk of missing launch date - Margin below threshold (35%) **Integrations:** - Google Sheets for data storage - Email notifications for stage transitions **Tech:** React/JavaScript frontend, drag-drop library (react-beautiful-dnd), Google Apps Script backend

📋 Vibe Coding Cheat Sheet

Quick Command Patterns

Pattern Use Case Example
You are a [role] Set expertise context "You are a data scientist for IRVINS"
Tech stack: [list] Define environment "Tech stack: Python, pandas, FastAPI, PostgreSQL"
Goal: [outcome] Specify business value "Goal: Reduce inventory stockouts by 30%"
Edge cases: [list] Handle errors "Edge cases: null values, negative numbers, division by zero"
Do NOT [constraint] Negative constraints "Do NOT use external libraries beyond specified stack"
Include tests for [scenarios] Request testing "Include tests for empty data, extreme values, API failures"
Existing code: [context] Integration context "Existing code: function fetchSalesData() returns array of objects"
Return full files: [list] Specify deliverables "Return full files: app.py, models.py, README.md"

Advanced Techniques

Micro-Prompt Chaining

Break complex features into 3-5 sequential mini-prompts:

  1. "Write the function signature for TPM ROI calculator"
  2. "Implement baseline calculation logic"
  3. "Add incremental volume calculation"
  4. "Create UI for inputs and outputs"
  5. "Add unit tests for edge cases"

Negative Constraints

Tell AI what NOT to do for clearer direction:

  • "Do NOT use Redux; use React hooks for state"
  • "Do NOT hard-code API keys; use environment variables"
  • "Do NOT generate sample data; connect to actual API"

Context Injection

Paste existing code or schemas to maintain consistency:

Existing schema: {date: "YYYY-MM-DD", market: string, revenue: number}

AI will match this format automatically.

"What's Missing?" Prompt

After initial generation, ask:

"Given this dashboard and the goal to track SG sales, what error handling, validation, or security checks are missing?"

AI will audit and suggest improvements.

Common Pitfalls to Avoid

⚠️ Avoid These Mistakes:
  • Vague goals: "Make a sales tool" → "Build a weekly sales variance report comparing actual vs budget by SKU and channel"
  • Missing business logic: AI doesn't know IRVINS-specific calculations. Always specify formulas.
  • No edge case handling: Explicitly state: "Handle missing data, invalid inputs, API timeouts"
  • Accepting first output: Always iterate. Ask "What could break?" and "How can this be improved?"
  • No version control: Save iterations. GitHub, Google Drive, or simple file versioning.
  • Ignoring security: Request "Prevent SQL injection, validate inputs, use parameterized queries"

Debugging Prompts

Issue Debug Prompt
Code breaks with edge case "The function fails when input is [X]. Add validation to handle [scenario]."
Performance is slow "This query takes 10+ seconds with 10k rows. Optimize using indexing or reduce loops."
Output doesn't match business logic "The ROI calculation should be [(incremental revenue - cost) / cost] × 100, not [current formula]. Fix this."
UI looks broken on mobile "Make this dashboard mobile-responsive. Use Tailwind CSS breakpoints for sm, md, lg screens."
Missing documentation "Add docstrings to all functions and a README explaining how to run this app."

⚙️ End-to-End Workflows

Workflow 1: Building a New Dashboard (0 to Deploy)

Phase 1: Requirements (15 min)

  1. Define user problem: "Sales managers can't see daily channel performance"
  2. List must-have metrics: revenue, units, channel mix, top SKUs
  3. Sketch wireframe: use Figma, paper, or describe layout in text
  4. Identify data source: Google Sheets, API, CSV export

Phase 2: First Prompt (10 min)

PHASE 2 You are a frontend developer for IRVINS. **Tech:** HTML, JavaScript, Chart.js, Tailwind CSS **Goal:** Build daily sales dashboard for Singapore market **Metrics:** Revenue, units sold, channel mix (retail, e-commerce, MT, marketplaces), top 5 SKUs **Data Source:** Google Sheets API (read-only, sheet name: "sg_daily_sales") **Layout:** Header with date selector, 4 KPI cards, channel mix pie chart, top SKUs table **Deliverable:** Single HTML file Build this dashboard.

Phase 3: Test & Iterate (20 min)

  1. Run the generated HTML file locally
  2. Test with real data (or sample data)
  3. Identify gaps: "Date selector doesn't filter data" or "Chart colors aren't IRVINS brand colors"
  4. Refinement prompts:
    • "Add date range filter that updates all metrics and charts"
    • "Use IRVINS brand colors: teal (#21808d), cream (#fcfcf9), slate (#13343b)"
    • "Handle case when API fails: show error message and use cached data"

Phase 4: Deploy (10 min)

  1. Commit to GitHub repository
  2. Deploy to GitHub Pages, Netlify, or Vercel
  3. Share link with stakeholders for feedback

Total Time: ~1 hour from concept to live dashboard

Workflow 2: Refactoring Legacy Code with AI

Scenario:

You have a messy Python script that processes sales data but it's hard to maintain and has no tests.

Step 1: Audit Prompt

AUDIT You are a senior Python engineer. **Context:** I have a sales processing script (below) that works but is messy. [Paste existing code] **Task:** Analyze this code and tell me: 1. What are the main issues (complexity, readability, bugs)? 2. What edge cases are not handled? 3. How should this be refactored for production use?

Step 2: Refactor Prompt

REFACTOR Based on the analysis, refactor this script into: 1. **data_loader.py:** Functions to load CSV data 2. **processor.py:** Functions for calculations (modular, testable) 3. **exporter.py:** Functions to export results 4. **main.py:** Orchestrate the workflow **Requirements:** - Add docstrings to all functions - Add type hints (Python 3.10+) - Handle errors (file not found, invalid data) - Add logging for debugging - Include unit tests in tests/test_processor.py Return full refactored code.

Step 3: Test & Validate

  1. Run new code with sample data
  2. Run tests: pytest tests/
  3. Compare output with legacy script (should match)
  4. Commit refactored version

Workflow 3: Rapid Prototyping for Stakeholder Demo

Scenario: Leadership meeting in 2 hours, need to demo market expansion model.

Lightning Prompts (10 min each):

  1. Prompt 1: "Create HTML form for 6 market criteria (market size, competition, regulatory, logistics, brand fit, ROI). Each 1-5 slider."
  2. Prompt 2: "Add weighted calculation: market size × 2, ROI × 1.5, others × 1. Display total score out of 35."
  3. Prompt 3: "Add recommendation logic: >28 = Go, 20-28 = Pilot, <20 = No-Go. Color-code green/yellow/red."
  4. Prompt 4: "Add comparison view: 3 markets side-by-side with radar chart using Chart.js."
  5. Prompt 5: "Add export to PDF button using html2pdf.js library."
  6. Prompt 6: "Style with IRVINS brand colors and make mobile-responsive."

Result: Functional prototype in 1 hour, 30 min for testing and polish, demo-ready in 90 min.

🎓 Learning Path for Teams

Week 1: Basics

  • Understand the 7-element prompt structure
  • Build 1 simple calculator (TPM ROI or margin)
  • Practice iterating through 5 refinement prompts

Week 2: Intermediate

  • Build dashboard with API integration
  • Add error handling and validation
  • Deploy to GitHub Pages

Week 3: Advanced

  • Create FastAPI microservice
  • Write unit tests with AI assistance
  • Implement authentication (JWT)

Week 4: Production

  • Refactor existing tool to production standards
  • Set up CI/CD pipeline
  • Document API and create user guide
💡 Pro Tip: Create a team "prompt library" in Google Docs or Notion. When someone writes a great prompt that works, save it as a template for the team. This builds institutional knowledge.

🔗 Resources & Next Steps

Recommended Tools for IRVINS

  • AI Coding: Claude (Anthropic), Perplexity Computer, GitHub Copilot
  • Deployment: Vercel, Netlify, GitHub Pages, Railway
  • Backend: FastAPI (Python), Node.js + Express
  • Frontend: HTML + Tailwind CSS + Chart.js (quick), React (complex)
  • Data: Google Sheets API, Airtable, Supabase (PostgreSQL)
  • Testing: pytest (Python), Jest (JavaScript)
  • Version Control: GitHub, GitLab

Integration with IRVINS AI Transformation

Vibe coding accelerates our AI transformation roadmap by enabling:

  • Agentic workflows: Build AI agents that automate data processing, reporting, and decision support
  • Self-service analytics: Empower commercial, operations, and marketing teams to build their own tools without IT dependency
  • Rapid experimentation: Test new business models, pricing strategies, or channel approaches with quick prototypes
  • Scalable systems: Create microservices architecture that grows with IRVINS expansion (Singapore → USA → China → beyond)

What to Build Next

Priority projects for IRVINS teams:

  1. Commercial: Real-time channel performance dashboard (all markets)
  2. Operations: Demand forecasting model with production scheduler
  3. Marketing: Unified campaign ROI tracker (all digital channels)
  4. Finance: Automated weekly P&L flash report generator
  5. Retail: Store inventory alerts and reorder system
  6. E-Commerce: Marketplace performance analyzer (Shopee, Lazada, TikTok, Amazon)
  7. Strategy: Market expansion evaluation framework
  8. Product: NPD pipeline tracker with stage gates