Firestore Data Modeling Patterns

Overview

Firestore is a NoSQL document database that requires careful modeling to optimize for query patterns, scalability, and cost. This skill provides patterns for structuring data effectively.

Subcollections vs. Root Collections

When to Use Subcollections

Pattern: users/{userId}/orders/{orderId}

Use When:

• •Data is tightly coupled to a parent (orders belong to a specific user)
• •Child data is always accessed via parent
• •You don't need to query children globally across all parents
• •Document hierarchy makes semantic sense

Example:

// User's private sessions (accessed only via user)
users/{userId}/sessions/{sessionId}

// User's notification preferences
users/{userId}/settings/notifications

Critical Limitation: Deleting a parent document does NOT delete subcollections. You must implement cleanup logic (e.g., Cloud Function).

When to Use Root Collections

Pattern: Separate users and posts collections with reference fields

Use When:

• •Need to query data globally (e.g., "all posts across all users")
• •Data has many-to-many relationships
• •Want simpler deletion semantics (no orphaned data risk)
• •Need flexibility for future access patterns

Example:

// posts collection
{
  id: "post1",
  authorId: "user123",  // Reference to users collection
  title: "...",
  createdAt: Timestamp
}

// Query all posts by a user
postsRef.where('authorId', '==', 'user123').get()

// Query all posts globally
postsRef.orderBy('createdAt', 'desc').limit(10).get()

Decision Matrix:

Document Structure Best Practices

Embedding vs. Referencing

Embed When:

• •Data is small and frequently accessed together
• •1-to-1 or 1-to-few relationships
• •Data doesn't change frequently

// User profile with embedded address
{
  id: "user1",
  name: "Alice",
  address: {
    street: "123 Main St",
    city: "NYC",
    zip: "10001"
  }
}

Reference When:

• •Data is large or changes frequently
• •1-to-many or many-to-many relationships
• •Need to query the related data independently

// Post references author
{
  id: "post1",
  title: "My Post",
  authorId: "user1", // Reference
  categoryIds: ["cat1", "cat2"] // Many-to-many
}

Document Size Limits

• •Max Document Size: 1MB
• •Max Array Size: 20,000 elements (but practical limit is much lower for performance)
• •Max Nesting Depth: 20 levels

Atomic Operations

Transactions (Read-Modify-Write)

Use When: Write depends on current document state

Example: Increment a counter

import { runTransaction } from 'firebase/firestore';

await runTransaction(db, async (transaction) => {
  const postRef = doc(db, 'posts', 'post1');
  const postDoc = await transaction.get(postRef);

  if (!postDoc.exists()) {
    throw new Error('Post does not exist');
  }

  const newViewCount = postDoc.data().viewCount + 1;
  transaction.update(postRef, { viewCount: newViewCount });
});

Characteristics:

• •Reads must precede writes
• •Automatic retries on conflicts
• •Fails if offline
• •Limited to 500 documents

Batched Writes (Write-Only)

Use When: Multiple independent write operations need atomicity

Example: Create user + settings document

import { writeBatch } from 'firebase/firestore';

const batch = writeBatch(db);

const userRef = doc(db, 'users', 'user1');
batch.set(userRef, {
  name: 'Alice',
  email: 'alice@example.com',
  createdAt: serverTimestamp(),
});

const settingsRef = doc(db, 'users', 'user1', 'settings', 'notifications');
batch.set(settingsRef, {
  emailNotifications: true,
  pushNotifications: false,
});

await batch.commit(); // All succeed or all fail

Characteristics:

• •Faster than transactions
• •Works offline (queued)
• •Up to 500 operations
• •No reads allowed

Decision Rule: Default to batched writes (simpler, faster). Use transactions only when reads are required.

Best Practices Summary

Do:

• •Denormalize data for read-heavy applications
• •Use root collections for flexibility
• •Create composite indexes for complex queries
• •Use batched writes for atomic multi-document updates
• •Keep documents under 1MB

Don't:

• •Embed large arrays or frequently-changing data
• •Use subcollections without cleanup strategy
• •Create excessive indexes (storage cost + write latency)
• •Assume parent deletion cascades to subcollections
• •Use transactions when batches suffice

Related Skills: zod-firestore-type-safety, firebase-nextjs-integration-strategies