Data shape determines query and index shape

Schema design and indexing are not separate phases. The location of a field—top-level, nested, inside an array, or in a referenced collection—changes how the application queries it and which index can support that query.

Embedded field index

For documents shaped like:

{
  _id: ObjectId("65a000000000000000000080"),
  customerId: ObjectId("65a000000000000000000010"),
  address: {
    country: "IN",
    postalCode: "560001"
  }
}

an index can target the nested field:

db.customers.createIndex({ "address.postalCode": 1 });

Array index and multikey behavior

If tags is an array, an index on tags becomes a multikey index:

db.products.createIndex({ tags: 1 });
db.products.find({ tags: "mongodb" });

Array indexing has restrictions, especially when indexing multiple array fields in the same compound index for one document. Understand multikey index behavior before modeling a query that depends on multiple arrays.

Compound indexes should follow real queries

Suppose the application frequently retrieves a customer's orders by status, newest first:

db.orders.find({
  customerId,
  status: "shipped"
}).sort({ createdAt: -1 });

A candidate index is:

db.orders.createIndex({
  customerId: 1,
  status: 1,
  createdAt: -1
});

The best order depends on the query shapes, selectivity, sort requirements, and other operations. Test with explain("executionStats") and representative data.

Referenced child collections

If comments are separate from posts, a post's document shape cannot by itself provide an efficient paginated comment list. The comments collection needs an index designed for the query, such as { postId: 1, createdAt: -1, _id: -1 }.

Projection and data locality

Embedding may reduce the need for an extra read, but a very large embedded document can make fetching a small subset of fields less efficient than a purpose-built collection. Evaluate actual document size, projection, working set, index size, and network payload.

db.orders.createIndex({
  customerId: 1,
  status: 1,
  createdAt: -1
});

db.orders.find({
  customerId: ObjectId("65a000000000000000000010"),
  status: "shipped"
})
.sort({ createdAt: -1 })
.limit(20)
.explain("executionStats");