Backend Platform for Home Goods Marketplace

Designed and developed a backend system for a growing marketplace. Ensured stable order processing, catalog and pricing management. The platform handles 50,000+ concurrent users and is ready for significant growth.

Industry
E-commerce
Format
B2C
Duration
4 months
Stack
Node.js, PostgreSQL, Redis, Kubernetes

About the Client

A home goods marketplace focused on designer interior items. The platform hosts 500+ sellers with 35,000 monthly active users.

Fast-growing e-commerce segment requires reliable infrastructure for order processing without losses or downtime during peak periods.

Venture-funded startup

Challenge & Problems

  • Growing product catalog with complex attribute structure and dependencies
  • High order processing load during peak periods
  • Complex pricing logic with discounts, promo codes, and seller conditions
  • Need for reliable payment and logistics integrations
  • Risk of downtime and order loss during traffic spikes
  • No technical team for backend development and maintenance

Why standard solutions didn't work

Off-the-shelf solutions (Shopify, WooCommerce) didn't provide needed catalog flexibility and couldn't implement specific multi-vendor marketplace business logic.

Project Goals

Ensure stable order processing

24/7 without losses during peak periods

Scale platform for audience growth

up to 100,000 MAU

Stabilize API response time

under 100ms under load (p95)

Automate catalog and inventory sync

real-time updates

Our Solution

Developed a modular backend platform with centralized order processing and catalog management services. Architecture designed for external system integration and fault-tolerant operation under load.

Catalog Service

Product, attribute, and inventory management. Full-text search and filtering.

Order Service

Order processing, cart, delivery calculation. Status tracking and history.

User Service

Authentication, profiles, favorites, purchase history.

Payment Gateway

Integration with Stripe and local payment systems. Payment and refund processing.

Notification Service

Email, SMS, and push notifications for order statuses.

Architecture

Modular structure with unified API Gateway. Async processing via message queues. Caching to reduce database load.

Fault Tolerance

Data replication, graceful degradation on external service failures, automatic recovery.

Monitoring

Centralized metrics and log collection. Alerts on anomalies. Request tracing for diagnostics.

Development Process

1

Architecture Design

Business requirements analysis, API and data structure design. 2 weeks.

2

Core Services Development

Catalog, orders, and user services implementation. 6 weeks.

3

Integrations

Payment systems, logistics, and notification connections. 3 weeks.

4

Infrastructure Setup

Kubernetes deployment, CI/CD and monitoring configuration. 2 weeks.

5

Load Testing

Performance verification under load, bottleneck elimination. 2 weeks.

6

Phased Launch

Production deployment, metrics monitoring, adjustments. 1 week.

Technology Stack

Backend

Node.js 20
NestJS
TypeScript
GraphQL

Databases

PostgreSQL 16
Redis 7
Elasticsearch 8

Infrastructure

Kubernetes
Docker
Helm
ArgoCD

Monitoring

Prometheus
Grafana
Jaeger
Sentry

Cloud

AWS EKS
RDS
ElastiCache
S3

Results

Measurable Results

45ms (p95)

API response time

under working load

50,000+

Concurrent users

without degradation during peak periods

1.2M/day

Request processing

current load after launch

99.95%

Availability

over first 6 months of operation

Qualitative Improvements

  • Stable order processing without losses during peak periods
  • Ability to quickly add new features without architecture rework
  • Client team independently develops product on the platform
  • Infrastructure ready for significant load growth

Business Value

Platform enabled on-time product launch. Stable backend operation helped secure next investment round. Infrastructure cost — $800/month at current loads.

Current Usage

Platform serves 35,000 MAU and 500+ sellers. System operates 24/7 with minimal operations team involvement.

Scaling Opportunities

Architecture designed to scale to 1M MAU. Adding new regions and sellers requires no core system changes.

Challenges & Learnings

Peak Loads During Sales

Problem

During marketing campaigns, load increased 5–7x. Order service couldn't cope, users received errors.

Solution

Implemented order processing queue with prioritization. Configured automatic horizontal scaling based on load metrics.

Learning

Designing for peak loads is mandatory for e-commerce platforms. We apply this architectural approach to all marketplace backend projects.

Data Consistency Between Services

Problem

Under high load, order and inventory services became out of sync. Customers placed orders for out-of-stock items.

Solution

Implemented Saga pattern for distributed transactions. Added inventory reservation mechanism with TTL and automatic rollback.

Learning

Data consistency matters more than speed. For multi-vendor marketplaces, this is a critical architectural aspect.

Want the Same Results for Your Business?

Describe your task — we'll propose architecture, timeline and cost.