Organizations with very high transaction volumes, and more so when data flows between multiple systems, will experience some percentage of records that cannot be committed during processing. There are a number of reasons for this undesirable situation; for example, reference tables might not be coherent in one or more of the involved systems.
Identifying the rejected records and resolving these cases, in particular when the transaction volume is high, is an arduous task. Thus enters EMS, the system that reduces revenue leaking and labor required to resolve such processing errors. EMS beautifully organizes rejected records into cases that can be remedied with a single action. The system's Machine Learning module learns from past actions and automatically applies corrective actions. Below a certain probability threshold, the organization can set a policy that asks that EMS will only suggest actions that will be approved by analysts.
The first EMS system was invented by Erez Sverdlov in Amdocs Israel. It was tailored for each customer, in particular because of the high degree of integration with the involved systems. Naturally these systems are radically different between each client and there exists a high-degree of variability in the rejected records structure. I took responsibility for the system to redesign it as a generic, single source system. This was a task that I completed while working at Amdocs' development center in St. Louis.