MongoDB Module 5: Schema Design and Validation

Step 1: Embed or reference

Embed when data is read together, belongs to one parent, and the embedded part is small and bounded (an address, line items in an order). Reference when data is shared by many parents, changes independently, or can grow without limit (all orders of a customer, comments on a popular post). Embedding gives faster reads and atomic updates of one document; referencing avoids duplication and oversized documents.

Embed:      order { items: [ {sku, qty, price} ] }        read together, bounded
Reference:  order { customerId: ObjectId(...) }           customer shared, changes alone

Step 2: Avoid unbounded arrays and huge documents

A document has a size limit (16 MB), and very large arrays slow updates and indexes. Do not store an ever-growing list inside a parent document, such as every click for a user. Store each event as its own document that references the parent instead.

// risky: grows forever
{ _id: 1, user: "asha", clicks: [ ... millions ... ] }

// better: one document per event
{ _id: ObjectId(), userId: 1, url: "/pricing", at: ISODate("2026-10-10T10:00:00Z") }

Step 3: Use common design patterns

The bucket pattern groups many small time-series measurements into one document per time window to cut document count. The subset pattern keeps only the most-read part (the latest ten reviews) inside the parent and the rest in another collection. The computed pattern stores precomputed values such as totals so reads stay fast. Pick a pattern only when you have a measured need. MongoDB also provides native time series collections for measurement data.

// bucket pattern: one document per sensor per hour
{ sensorId: "s1", hour: ISODate("2026-10-10T10:00:00Z"),
  readings: [ { t: 0, v: 21.4 }, { t: 60, v: 21.5 } ], count: 2, sum: 42.9 }

Step 4: Add schema validation

Validation lets you require fields and types while keeping flexibility elsewhere. Define a $jsonSchema validator when you create the collection or with collMod. Choose the validation action: error rejects invalid writes, warn only logs them.

db.createCollection("customers", {
  validator: { $jsonSchema: {
    bsonType: "object",
    required: ["name", "email"],
    properties: {
      name:  { bsonType: "string", minLength: 1 },
      email: { bsonType: "string", pattern: "^.+@.+$" },
      age:   { bsonType: "int", minimum: 0, maximum: 120 }
    }
  } },
  validationAction: "error"
})

Step 5: Plan for change

Add a schemaVersion field to documents so your code can handle old and new shapes while you migrate gradually. Make changes backward compatible where you can, and migrate in batches rather than one huge update. Design for your top queries: write them down before you design the documents.