How do you decide between embedding and referencing in MongoDB?

Short answer

Embed when the data is read together, owned by the parent, and bounded in size. Reference when it is large, shared between parents, independently queried, or can grow without limit.

Why?

Embedding gives you one read and one-document atomic updates, which is why it removes most transaction needs. Referencing avoids unbounded documents and shared mutable copies. Order line items are the classic embed. A user's orders are the classic reference. Start from 'how is it read?', not from an ER diagram.

Interview tip

Answer with the access pattern first. Interviewers want to hear you ask how it is read.

Common mistake

Copying the SQL ER diagram one collection per table, then $lookup on every request — or the opposite, one mega-document.