ARCHIVES

Automated decision-making has relied on rule-based systems for many enterprise applications for a number of years; however, many practical implementations have limitations related to rigid schemas, non-disclosed evaluation logic and a lack of configuration tools available to business domain experts without the need for developers. We describe here the design and development of a Rule Based Order Evaluation Engine (RBOEE), a modular web system that evaluates bulk order data provided in Microsoft Excel format based on a set of defined conditional rules by the administrator. The engine accepts an Excel workbook containing order records with various attributes (i.e., order value, quantity, user type, region and payment method) and evaluates each order against configurable rules using compound logical conditions (i.e., AND and OR operators) and allows for nested sub-condition groups to be defined to support any level of nesting. The architecture of the engine is represented by a layered framework using a React.js front-end, Node.js/Express RESTful back-end and MongoDB persistence layer enabling separation of concerns and independent horizontal scaling for each layer of the architecture. The evaluation engine operates with O(n x r) complexity for n order records and r rules; benchmark testing indicates that the short-circuit optimized implementation can evaluate 1,000,000 order records in approximately 2.87 seconds with 20 active rules. An essential level of explainability via matching establishes an important audit trail based on all conditions and actual data used for matching each rule. This ability is extremely critical for auditing purposes across many applications in pricing, fraud investigation, and logistics. In addition, we’ve built a extensible architecture for production purposes which is well-suited to the operational environment of e-commerce, banking and supply chain management.