Introduction of Credit Scoring in AccessHolding network banks

2012-10-31 - 2013-03-31
Germany, Azerbaijan, Madagascar, Tanzania, Nigeria, Liberia, Tajikistan, Zambia

To improve operational efficiency and move forward in line with recent developments in the digital finance sphere, automated and digital loan processing has recently been moved to the focus of attention in the AH network and digital finance has become a core part of the overall business strategy. Credit scoring as one of the most prominent tools of digital finance has been identified to serve the network in order to decrease the processing time of loan analysis and disbursement, streamlining and improving the quality of credit decisions, as well as reducing operational costs. Over the past year, several scoring models were designed and pilot tested for different institutions of the network, including AccessBank Azerbaijan, AccessBank Tanzania and Credo Georgia.
Credit Scoring at AccessBank Azerbaijan In August 2013, a scoring system was built for AccessBank Azerbaijan (ABA) in order to facilitate automated decision processes for loans with a positive loan history. The scorecard was designed by an external consultant with a strong track record in building such systems. The scoring model was developed based on historic data of over 300,000 loans disbursed by ABA. Based on the analysis of such data, 15 criteria with influence on the credit repayment behavior were identified and used as a basis for the scorecard. During a 3-months period, the scoring system was pilot tested. Based on the results of this phase, the project was rolled out to all branches. The scoring was implemented for two loan products, yet, without interlinking the system with the core banking system. Credit Scoring at Credo Georgia In 2012, Credo Georgia ventured into a project to build a credit scoring system to improve the accuracy and efficiency of risk assessments of potential agricultural loan clients and reduce the variability of credit decisions. In collaboration with an external consulting company with strong expertise in designing scoring models, Credo Georgia designed a scorecard to score loans for agricultural microenterprises. In 2015 the scorecard was further developed to provide an utmost accurate risk assessment of potential clients and derive an associated decision-making strategy. To build the scorecard, two data sets from the T24 core banking system were used, one on general loan information and another set on delinquency information. In detail, the different steps to build the scoring system included: Definition of good and bad loans Selection of the sample Analysis of variables Identification and selection of loan segments Development of the scoring formula Evaluation of the accuracy of the scoring formula Design of loan approval strategies In early 2016, the pilot testing of the scoring system at Credo Georgia started and over 62,000 loans have been scored for testing reasons to date. Disbursements will only happen after completion of the pilot test (in September 2016), based on which the implementation throughout the whole institution is planned. Credit Scoring at AccessBank Tanzania To provide a deeper understanding of how a scoring system can be built in practice, we present the details of the scoring approach and pilot test at AccessBank Tanzania in the following: The scoring model at ABT was built together with an US-based company with a strong track record in designing scoring mechanisms based on and interlinking data sets from different providers such as financial institutions or mobile network providers. The first phase of the pilot test was conducted between September 2015 and March 2016. During that period, over 700 loans were scored 300 of which were disbursed. With only two of those disbursed loans defaulting, the Portfolio at Risk >1 has remained below 1% to date.