Relational - (PostgreSQL, MySQL, SQL Server, Oracle)

• Tabular - (like an Excel spreadsheet)
• Structured - (rows and columns)
• Consistency - (datatypes & constraints. Limit allowed input to improve quality)

Best when

• Clearly defined data structures & relationships. Can be easily Normalised
• Data quality matters or needs constraining

Non-Relational - (MongoDB, Cassandra, Redis, Couchbase, DynamoDB)

• Less structured - (flexibility for inconsistent keys or arrays)
• Data largely unrelated - (no schemas, tables cannot be reliably joined)

Best when

• The data structure cannot be clearly defined
• If data inputs are variant and flexibility matters

Graph Database - (AWS Neptune, Neo4J, HyperGraphDB, ArangoDB)

• Nodes & Edges - (every node is related to other nodes via edges)
• No structure - (relationships at record level)

Best when

• Everything is related to everything else
• Deep-dive network analysis is required (e.g. fraud detection)