MongoDB stores documents (BSON, max 16 MB). Model around how the application reads data: data that is read together should usually be stored together. Embed (sub-documents / arrays) Reference (store ids) one-to-few (addresses on a user) one-to-many with unbounded growth (comments on a viral post) data always read with the parent data read independently atomic updates of parent + children in one write many-to-many relationships child has no meaning on its own shared entities (a product referenced by many orders) Useful patterns Extended reference — copy a few frequently-read fields (e.g. customer.name) next to the reference to avoid a lookup. Subset — embed the latest 10 reviews, keep the rest in their own collection. Bucket — group time-series readings per hour/device (or use native…