This is an article about graph databases in general.
This is an article about neo4j. From the ground up, maybe some reference material.
O'reilly Graph Databases
Graphs often map more naturally to the entities (nodes) and their relationships than 'traditional' RDBMS do. Business and technical sides can now communicate because the whiteboard model is the model. The questions you want to ask of your data will drive how to model your graph. Graphs, as opposed to other NOSQL databases, retain ACID transactions, fast lookups, and scalability. While there is no one-size-fits all database solution, graphs are a formidable option in any realm and really shine with densely connected datasets.