MongoDB Schema Design

Unlike relational databases, MongoDB has a flexible schema — but that flexibility demands intentional design. The golden rule: design your schema around your application's access patterns, not around the data relationships.

Embedding vs Referencing

Embedding — store related data inside a single document

// User with embedded addresses
{
  "_id": "user_1",
  "name": "Alice",
  "addresses": [
    { "type": "home", "city": "New York", "zip": "10001" },
    { "type": "work", "city": "Boston",   "zip": "02101" }
  ]
}

✅ Use when: data is always read together, one-to-few relationships, no independent access to nested data.

Referencing — store a foreign key (ObjectId) and join in application code or via $lookup

// Order referencing User
{ "_id": "order_1", "userId": "user_1", "total": 99.99 }

✅ Use when: data is large/unbounded, shared across many documents, accessed independently.

Decision Matrix

Situation	Embed	Reference
One-to-few	✅
One-to-many	depends	✅
One-to-millions		✅
Data changes frequently		✅
Data read together 90%+	✅
Max document size concern (16MB)		✅

The Bucket Pattern — for time-series data

Instead of one document per event (too many docs), group events into time buckets:

{
  "sensorId": "sensor_42",
  "date": "2024-01-15",
  "readings": [
    { "time": "00:00:00", "temp": 22.1 },
    { "time": "00:01:00", "temp": 22.3 }
    // ... up to 60 readings per bucket
  ],
  "count": 60,
  "avgTemp": 22.2,
  "minTemp": 21.8,
  "maxTemp": 22.5
}

The Outlier Pattern — handle rare large documents

// Most blog posts have < 100 comments → embed
// Viral posts have 10,000+ comments → use overflow flag
{
  "_id": "post_1",
  "title": "...",
  "comments": [ /* first 100 */ ],
  "hasOverflow": true  // flag that more comments are in separate collection
}

Schema Versioning

Add a schemaVersion field to handle migrations gracefully:

// v1 document
{ "schemaVersion": 1, "name": "Alice Smith" }

// v2 document (migrated)
{ "schemaVersion": 2, "firstName": "Alice", "lastName": "Smith" }

// Handle both versions in application code
public User fromDocument(Document doc) {
    int version = doc.getInteger("schemaVersion", 1);
    if (version == 1) {
        String[] parts = doc.getString("name").split(" ");
        return new User(parts[0], parts[1]);
    }
    return new User(doc.getString("firstName"), doc.getString("lastName"));
}

Document Size Limit

MongoDB documents have a 16MB limit. For large arrays (comments, events), use referencing or the bucket pattern to stay within the limit.

Interview Tips

The core principle: design for access patterns, not for data relationships.
Explain the embedding vs referencing trade-off clearly — this is the most common MongoDB schema question.
Know that unbounded arrays are an anti-pattern — they can grow past 16MB and hurt index performance.