A Comparative Study of Relational, Graph, Wide Column, Key-Value, and Document Models in Retail Industries

Aim/Background

The study addresses the growing need for efficient database management systems in the rapidly expanding online retail sector, especially post-COVID-19. It aims to identify the most suitable data model for handling retail operations by comparing five types: relational, graph, wide-column, key-value, and document databases. The focus is on evaluating their performance in retail-specific functionalities such as order processing and Customer Relationship Management (CRM).

Methodology

Researchers created test environments using five database platforms-Oracle APEX, ArangoDB, AstraDB, Redis, and CouchDB. Each was populated with 200–1000 records and assessed under two retail scenarios: order management and CRM systems. Data was inserted via JSON files and evaluated based on CPU usage, processing speed, and memory consumption. Performance data was collected using both database logs and system monitoring tools.

Results

Oracle APEX (relational) demonstrated the most consistent performance across all tested metrics, including data integrity, query speed, and memory efficiency. ArangoDB excelled in handling complex relationships but required tuning. AstraDB showed high throughput and scalability, while Redis performed best in speed for simple operations. CouchDB offered schema flexibility but lagged in complex query handling due to its reliance on MapReduce.

Discussion

Relational databases proved most effective in structured data environments where strict access control and data integrity are essential. In contrast, NoSQL models provided better flexibility and scalability but lacked advanced features like views and joins. While some prior research favors No SQL, this study highlights the continued relevance of SQL models in structured, high-volume retail applications.

Conclusion

The research concludes that relational databases, particularly Oracle APEX, are the most balanced and reliable option for online retail systems. Despite the appeal of No SQL models for specific use cases, SQL databases offer superior overall performance, especially in maintaining data consistency and supporting complex queries in dynamic retail environments.