Use of ERIAN for the optimization of the instruments used in claim recovery (Vostok Finance Group, Ukraine)
The ERIAN system is an original BRMS system developed at KOMIX. The historically first implementation of ERIAN was for automated risk assessment of customs declarations. Since then, its versatility and tested performance have made ERIAN gain additional implementations in various areas, both thematic and geographic. This case study describes the implementation of ERIAN for the optimization of the instruments used in claim recovery at Vostok Finance Group (VFG), one of the biggest financial institutions in Ukraine.
Vostok Finance Group - initial situation
The subject matter of business of VFG is purchasing of various types of debts and their collection. Various tools are used for collecting the debts, including notifications by means of text messages and e-mail, automated and manual telephone calls, mailing of legal letters and personal visits. The use of each of the instruments has a probability of success depending on the various attributes of the different cases. The use of each instrument also entails certain costs, and each of the tools also have a certain limited capacity. The purpose and objective of the business of Vostok Finance Group is to optimise the debt collection process in terms of maximisation of the debts collected while minimising the expended costs. An instrument has to be identified for each case that provides the maximum probability of success at the moment and entails the least costs.
VFG used to employ its own solution for the support of this process, but its performance capacity no longer sufficed, and it did not enable implementation of more complex decision-making models the implementation of which VFG was considering. In 2013, VFG conducted an international tender with the objective to supplant the existing system for optimising the debt collection process with a new system that would be better, more powerful and able to assess millions of cases several times a day.
Traditional vendors strove for the delivery for VFG, such as FICO and Xperian, who offered their standard solutions in the area. Competing with these traditional specialised systems, ERIAN too applied for the client’s favour in the area of debt collection; it could not boast such numbers of existing installations, but based on experience in other areas, it promised a noticeably higher operating capacity, greater options for future expansion and, above all, significantly lower acquisition and operating costs than the traditional specialised systems.
In the end, ERIAN’s flexibility, operating capacity and costs of ownership outperformed the other competing systems, so that ERIAN was selected as the new system for optimisation of VFG’s debt collection.
ERIAN is a versatile and scalable platform, which has to be slightly modified for each type of implementation. The typical customisation of ERIAN consists of a definition of input structure, definition and implementation of output rule actions, decision (assessment) output structure and structure of statistics stored for monitoring behaviour of rules. In addition, the case of VFG was the first implementation in Ukraine and so required a language localisation. Another standard service provided as part of a system implementation is the analysis of decision-making practice in the area and design of the best way to establish the starting rules in the ERIAN system. This is followed by training for future users and writing of training materials for future development of users’ own rules.
In terms of project management, the typical implementation of ERIAN consists primarily of a roughly several-week phase of user need analysis and designing of necessary adjustments, followed by a two-to-four-week phase of implementation of the adjustments designed. The actual installation of ERIAN in the client’s environment then takes typically about a week, followed by integration with the client’s existing systems and testing operation at a duration as needed. The integration was atypical in the case of VFG, because the requirement to achieve maximum performance capacity led to a replacement of the standard ERIAN interactive interface by means of a web-based service and SOAP with a high-capacity batch interface by means of database tables.
The integration with the client’s systems included a configuration of ERIAN so that it matched as much as possible the performance characteristics of the production operation. At VFG, the ERIAN system was implemented on a virtual server with 16 CPU cores, and the ERIAN configuration managed to achieve the processing of 1.1 million requests in about 3.5 minutes, i.e., 5200 requests per second. With this capacity, the critical point was the reading of requests and registering of results in database tables, and the load of the available CPUs was around 35%. A further capacity increase could be achieved by tuning the database parameters and the communication infrastructure if necessary. However, as the performance achieved was exceeded the client’s needs and expectations many times; further optimisation was therefore abandoned. This capacity enables VFG to scan its entire portfolio of cases up to several times a day, and thus flexibly respond to current operating conditions, such as changes in loading of call centres. That said, the comprehensive rules contained decision trees with a depth in excess of 80 levels. The fast and successful implementation led to an accelerated elicitation of trust in the new technology.
At present, the ERIAN system is implemented in routine operation and requires only minimal operating support. We continue to develop additional specific functionalities together with VFG. The project of implementation of ERIAN in an area of optimisation of a debt collection process was highly successful and serves as inspiration for other similar applications, such as optimisation of marketing campaigns.