Another question that arises quite often and early in the choice of graph model is whether a (categorical) characteristic should be stored as a node property, or whether it should have its own node.
Nodes are the following:
- Explicit: Nodes appear in the graph schema, which makes them more visible than properties.
- Performant: If we need to find nodes sharing the same characteristics, it will be much faster with a node than by filtering on properties.
On the other hand, properties are the following:
- Simpler: It's just a property with a value, not a complex node and relationship structure. If you don't need to perform special analysis on the property and just want it to be available when working on the node it belongs to, a property is just fine.
- Indexable: Neo4j supports indexing for properties. It is used to find the first node in a graph traversal and is very efficient (more details in Chapter 2, The Cypher Query Language).
As you can see, there is no universal answer to this question, and the solution will depend on your use case. It is usually recommended to list them and try to write and test the associated queries in order to make sure you are not falling into a Neo4j pitfall.
For more information and more examples on graph modeling, I encourage you to have a look at Neo4j Graph Data Modeling (see the Further reading section).
