Kamol’s Profile
  1. Topics/

Personal Start-Up

Personal Start-Up Journey #

A journey of building a business from the ground up.

Starting Point #

  • One of the founders introduced me to a fortune teller who wanted to create a platform that impacts a large number of people using his core knowledge.
  • Before the teacher met me, he had a conversation about a 3M Baht deal with an existing horoscope application, but he did not have that amount of money.
  • I suggested starting on the LINE platform because:
    • No need to force users to install a new mobile application (less user friction)
    • No need to have two development teams for iOS and Android (much more cost-effective)
    • Thai people are already familiar with the LINE platform
  • The main business logic also aligns perfectly with LINE’s notification feature.
  • First code commit was on 2019-10-27.
  • First revenue was generated on 2020-05-13.

Platform Nature #

  • It is a read-intensive system when notifying users of horoscope updates.
  • It can become write-intensive when our influencers post on their Facebook accounts.
  • The website assets don’t change frequently.

Tech Stack / Implementation Concepts #

  • Backend Evolution:

    • PHP7.2 + Phalcon3.4 → PHP7.4 + Phalcon4.0 (current)
    • Future migration plan: PHP8.0 Hyperf+Swoole
    • Gradually migrating backend APIs to Go

  • Process Automation:

    • Tesseract 4.1.1 for bank transfer slip OCR to automate order approval
    • Custom Docker image bundling Nginx + PHP-FPM using supervisord

  • Infrastructure as Code:

    • DigitalOcean CLI (doctl) for infrastructure management
    • Kubernetes cluster creation and configuration
    • Load balancer provisioning and management

  • Kubernetes Deployment:

    • Zero downtime deployment strategy
    • Secure environment configuration management
    • Horizontal and vertical scaling capabilities

  • Security and Performance:

    • Cloudflare as security proxy and CDN
    • Infrastructure information hiding
    • DDoS mitigation
    • Content caching and delivery optimization

  • Cloud-Native Architecture:

    • Following 12factor methodology
    • Assets stored in DigitalOcean Space (CDN)
    • Centralized user sessions in Redis
    • No local file logging for better scalability
    • Truly stateless application design

  • CI/CD Pipeline:

    • Jenkins for continuous integration and deployment
    • Automated payment slip approval
    • Scheduled horoscope update generation
    • Automated notification scheduling
    • Regular database maintenance and backups
    • Streamlined Docker image building and deployment

  • Performance Optimization:

    • Minimized blocking I/O operations
    • APCu for resident memory caching
    • Multi-level caching: Database → Redis → Memory
    • Sophisticated cache invalidation and warming
    • SQL query optimization through database logs

  • LINE Platform Integration:

    • Comprehensive LINE SDK implementation
    • Messaging API, LINE Login, LIFF (LINE Front-end Framework)
    • Dynamic rich menu capabilities
    • Interactive flex messages

  • Cost-Effective Solutions:

    • Google Sheets as a lightweight admin interface
    • Regular horoscope content updates through collaborative spreadsheets

System Architecture #

Designed to be simple and maintainable for a small team.

System Architecture Diagram

Platform Performance Metrics #

Cloudflare Analytics #

  • Monthly Overview (December 2023)

    Cloudflare Monthly Stats

  • Peak Request Period (December 2023)

    Peak Request Period

LINE Platform Metrics #

  • Total Friends: 572,695 (as of November 2023)

    LINE Friends Growth Chart

Internal Platform Metrics #

  • User Acquisition: 45,368 new friends in November 2023 (averaging 1,512 per day)
  • Payment Processing: 2,092 total top-ups in November 2021 with 935 (44.69%) automatically approved

OCR Technology Performance #

  • Successful OCR Example:

    Successful OCR Example

  • Challenging OCR Example:

    Challenging OCR Example

  • Peak User Activity (during horoscope updates, 30-minute window):

    Peak User Activity Chart

Customer Segmentation & Marketing #

Our database enables precise customer segmentation for targeted communications:

  • Active vs. expired users
  • Users who have never made a purchase
  • Previous customers with expired subscriptions
  • Users with referral networks (with/without downlines)
  • Point-based segmentation
  • Referral network level grouping
  • Many additional custom segments

This targeted approach provides more effective and cost-efficient marketing compared to traditional advertising platforms.