E-commerce is one of the most studied problem domains in software. Millions of online stores have been built, billions of transactions processed, and the architectural patterns that work have been refined over decades. This is good news for anyone planning a new online store: you don't need to invent the wheel.
This guide covers the key technical decisions in planning an e-commerce platform — from tech stack to payments to search.
Start By Defining Your Product Type
Not all e-commerce is the same. Your architectural choices should be driven by what you're selling and how:
- Simple storefront (physical goods, single vendor): Sell your own products directly. Standard inventory, checkout, shipping. The most common case.
- Digital products: Sell downloads, licenses, access codes, or subscriptions. No shipping, instant fulfillment. Different product schema and fulfillment logic.
- Marketplace (multi-vendor): Multiple sellers, one buyer-facing storefront. Adds seller accounts, commission management, and payout logic.
- Subscription box / recurring physical: Physical goods on a recurring subscription. Subscription billing is a meaningfully different problem from one-time purchases.
- B2B / wholesale: Different pricing for different buyer tiers, purchase orders, net terms, bulk discounts. More complex pricing and checkout logic.
Knowing your product type determines which platform decisions are relevant.
The Core Decisions: Platform vs. Custom
The first and biggest decision is whether to use an existing e-commerce platform or build from scratch.
Use a Platform (Shopify, WooCommerce, BigCommerce)
For most e-commerce businesses — especially those where the product and marketing are the differentiation, not the technology — using an existing platform is the right call.
Shopify is the dominant choice for good reason. It handles payments, inventory, tax, shipping, checkout optimization, and has thousands of apps for extending functionality. It's not cheap at scale but the operational overhead is minimal.
When to choose a platform:
- You want to launch in weeks, not months
- Your differentiation is not the e-commerce technology itself
- You're selling physical goods with standard fulfillment requirements
- You don't have dedicated engineering resources
When platforms don't work:
- You need custom checkout logic or highly custom product configurations
- You're building a marketplace with complex seller/buyer relationships
- Your pricing logic is complex (dynamic pricing, tiered B2B pricing)
- You need tight integration with existing systems
- Platform costs at your expected scale are prohibitive
Build Custom
Building custom gives you complete control at the cost of time, money, and ongoing maintenance. The custom route makes sense when your product is the technology, or when a platform's limitations would require more workarounds than building from scratch would.
The rest of this guide focuses on the custom build path.
Core Components of a Custom E-commerce System
Product Catalog
The product catalog is the foundation. Key data model decisions:
- Simple products vs. variants: T-shirts have variants (size, color). Books don't. Your product model needs to handle both.
- Categories and taxonomy: How products are organized for browsing and search.
- Attributes: Flexible product attributes (material, weight, dimensions) that vary by product type.
- Inventory: How stock is tracked. Per SKU (stock keeping unit, i.e., per variant), per warehouse, or both.
- Pricing: Is pricing uniform, or do different customer groups get different prices?
- Digital fulfillment: For digital products, how is the download link generated and delivered after purchase?
A clean product schema is worth spending real time on. It affects search, checkout, inventory management, and reporting.
Cart and Checkout
Cart and checkout are the most conversion-critical parts of any e-commerce system. Every friction point costs sales.
Key checkout decisions:
- Guest checkout vs. account required: Guest checkout is higher-converting. Account creation should be optional and offered after purchase.
- Address validation: Validate shipping addresses against a real address database to reduce failed deliveries.
- Tax calculation: Tax is complicated and jurisdiction-dependent. Use a service like TaxJar or Avalara rather than building your own tax engine.
- Shipping rate calculation: Real-time shipping rates from carriers (UPS, FedEx, USPS) or flat rates. Consider using a shipping API (Shippo, EasyPost) for rate shopping and label generation.
Payments
Use Stripe for payment processing in almost all cases. Alternatives (PayPal, Square, Braintree) are worth considering for specific markets, but Stripe has the best API, the most complete documentation, and the deepest ecosystem.
Critical payment considerations:
- PCI compliance: Never store card numbers yourself. Stripe's client-side tokenization (Stripe.js) handles this. Your servers never see raw card data.
- Stripe Checkout vs. custom UI: Stripe Checkout (hosted payment page) is the fastest to implement and PCI-compliant by default. A custom payment form (Stripe Elements) gives more control over design.
- Strong Customer Authentication (SCA): Required for EU customers. Stripe handles 3D Secure automatically if you use their recommended integration patterns.
- Webhooks for fulfillment: Trigger order fulfillment from Stripe's
payment_intent.succeededwebhook, not from the client-side redirect. Client-side redirects are unreliable — users close tabs, connections drop.
Search
Search is a first-class feature in most e-commerce sites. PostgreSQL's built-in full-text search works adequately for small catalogs (< 10,000 products). For larger catalogs or when search quality matters (fuzzy matching, typo tolerance, faceted filtering), consider:
- Meilisearch: Open-source, self-hostable, excellent typo tolerance. Free if self-hosted.
- Algolia: Managed, fast, excellent developer experience. Pricing scales with usage.
- Elasticsearch / OpenSearch: Powerful, but operationally complex. Worth it at large scale.
Key search features for e-commerce: keyword search, category filtering, attribute facets (filter by size, color, price range), relevance tuning, and "out of stock" handling.
Order Management
Order management covers the lifecycle from purchase to delivery:
- Order creation and confirmation emails
- Inventory decrement on purchase
- Fulfillment (picking, packing, shipping label generation)
- Shipment tracking integration
- Returns and refunds
For small operations, this can be manual or semi-automated. At scale, an order management system (OMS) or a fulfillment integration (ShipBob, Fulfillment by Amazon) becomes necessary.
Recommended Tech Stack
Frontend: React with Next.js is the standard choice. Next.js gives you:
- Server-side rendering for fast initial page loads and SEO (critical for product pages)
- Static generation for category pages and the homepage
- Image optimization built in (crucial for product photos)
- API routes for simple backend logic
Backend API: Node.js (Express or Fastify) or Python (FastAPI) for the custom business logic layer.
Database: PostgreSQL for the primary database. Strong relational data model suits e-commerce well (orders → line items → products → variants, all relational).
Search: Meilisearch for self-hosted, Algolia for managed.
Payments: Stripe.
Email: Resend or Postmark for transactional emails (order confirmation, shipping notification, receipt).
File/image storage: Cloudflare R2 or AWS S3 for product images. Use a CDN (Cloudflare, CloudFront) to serve them fast globally.
Infrastructure: Vercel for the Next.js frontend, Railway or Render for the API, Supabase or Neon for PostgreSQL.
The SEO Imperative
E-commerce SEO is uniquely important because organic search is the highest-ROI acquisition channel for most stores. Architecture decisions that hurt SEO — JavaScript-only rendering, blocking crawlers from product pages, slow page load times — can be fatal.
Requirements:
- Product and category pages must be server-rendered or statically generated (no client-side-only rendering)
- Every product must have a canonical URL
- Structured data (JSON-LD) for products enables Google Shopping integration and rich snippets
- Page speed matters: Core Web Vitals affect Google rankings
- Sitemaps for product catalog pages
Planning Your Build Phases
A reasonable build sequence for a custom e-commerce platform:
Phase 1 (Weeks 1–6): Product catalog, user accounts, basic cart and checkout with Stripe, order confirmation emails, basic admin for product management.
Phase 2 (Weeks 7–12): Inventory management, shipping rate integration, tax calculation, customer order history, basic search.
Phase 3 (Weeks 13–20): Advanced search with faceted filtering, promotions and discount codes, returns management, enhanced analytics.
Phase 4 (ongoing): Performance optimization, A/B testing on checkout, fulfillment integrations, mobile app if relevant.
Blueprint AI generates a complete e-commerce system architecture including the database schema, API design, development roadmap, and cost estimates — tailored to your specific product type and scale — in under a minute.