How do you model data in Neo4j?

A node has labels and properties. A relationship has a type, a direction and properties, and it is stored as a direct pointer between nodes. The model should make the important traversals explicit rather than hiding them in join tables.

CREATE (asha:User {id: "u_7", name: "Asha"})
CREATE (ravi:User {id: "u_9", name: "Ravi"})
CREATE (asha)-[:FOLLOWS {since: date("2026-01-10")}]->(ravi)
CREATE (asha)-[:PURCHASED {at: datetime()}]->(:Product {id: "sku_1"});

CREATE INDEX user_id IF NOT EXISTS FOR (u:User) ON (u.id);

Modelling guidelines

• Use relationship types for the verb in the domain, such as FOLLOWS or PURCHASED
• Store properties on the relationship when they describe the connection
• Avoid a supernode that connects to millions of others; it makes traversal explode
• Index the property used to find entry nodes, then traverse from there
• Do not encode every attribute as a node; properties are cheaper for simple values

A common mistake is translating a relational schema directly into nodes and edges, including join tables. A join table often becomes a relationship with properties. Another mistake is using a graph for simple lookups that an index could serve, which adds complexity without a traversal benefit.