MongoDB in E-Commerce

Introduction

E-commerce is MongoDB's flagship use case. From eBay's product catalog to Forbes' commerce stack and countless Shopify-style platforms, MongoDB powers product search, cart, checkout, order history and analytics — all from a flexible document model that matches how product data actually looks.

The document model is a natural fit: every product has different attributes (size, color, voltage, ISBN, expiration date), every cart embeds line items, every order embeds shipping + payment snapshots that must never change post-fact.

Understanding the topic

The core e-commerce collections:

• products — embed variants, specs and price tiers; index on slug, categoryId, tags.
• inventory — separate collection updated atomically with $inc; avoids embedding stock in the catalog.
• carts — short-lived, embedded line items keyed by userId or anonymous sessionId.
• orders — immutable snapshots of cart + price + address + tax at checkout time.
• reviews — referenced (not embedded) so reviews can be moderated and paginated independently.
• search — Atlas Search index over products for typo-tolerant, faceted, real-time search.

Informative example

Atomic "add to cart with inventory check" — multi-document transaction:

const session = client.startSession();
await session.withTransaction(async () => {
  const stock = await db.inventory.findOneAndUpdate(
    { sku, available: { $gte: qty } },
    { $inc: { available: -qty, reserved: qty } },
    { session, returnDocument: "after" }
  );
  if (!stock) throw new OutOfStockError(sku);

  await db.carts.updateOne(
    { userId },
    { $push: { items: { sku, qty, price: stock.price } },
      $set:  { updatedAt: new Date() } },
    { upsert: true, session }
  );
});

Real-world use

Coinbase Commerce, ASOS, Sega Store and 60% of the Internet Retailer Top 1000 use MongoDB for catalog and order data. Atlas Search replaces Elasticsearch entirely for many of them.

Best practices

• Snapshot price + tax + shipping into the order document at checkout — never recompute.
• Inventory in its own collection, updated atomically — never $inc embedded stock.
• Use Atlas Search facets for filter UIs (brand, size, price range) — no separate search engine needed.
• Add a unique index on orders.orderNumber to make idempotent retries safe.