European Microfinance Week 2016

PRESENTATIONS

Moderator Philippe BREUL opened by explaining the objectives of the session: explain big data techniques in the context of financial inclusion, learn how big data techniques are used in practice and identify what the benefits of big data are for both financial service providers and customers. He explained how big data techniques can contribute to financial inclusion by creating impact to one of six value proposition dimensions: Branding, customer care, convenience, executional excellence, products and services, and value for money. He then introduced how the speakers would each focus on different value propositions dimensions in their presentations. 

Etienne MOTTET, innovation analyst at Business and Finance Consulting, presented two case studies to answer the question whether we actually need to mine big data in microfinance. In both case studies the challenge was to get the best yield and profit from the companies’ agricultural lands. However, the data used was quite different. He first presented the case of Al Rawafed Serbia. This was a large scale and highly mechanised company. The choice was made to first invest in better intelligence, such as sensor, GPS and tractor fleet guidance tools. With these technologies, data was collected and analysed allowing tractor fleet optimization and configuration of input usage automation. The company benefited by saving 15% on inputs, increasing income by 20% , and from better cost control and improved soil management. 

The second case concerned Tylek from Kyrgyzstan: a small-sized company which had to consider a new type of crop to maximise agricultural outputs and profit. During data collection it became apparent that the local sugar factory had under capacity. After careful analysis, beetroot seemed to have good opportunities for Tylek so they applied for a loan which they received after careful testing by the agro expert. At the regional level, it has resulted in one beetroot processing factory working at maximum capacity and a second being operational in 2017. When comparing the two case studies, more could have been done with big data in the second case, but the choice was made to leverage existing data in order to keep the project manageable for the small company. 

Alexis LEBEL, CEO at OpenCBS, showed the case of Agora Microfinance Zambia, an MFI with 12,000 clients, 70% of which live in rural areas. OpenCBS designed a data collection system for this MFI and digitalized their appraisal process based on its own free Core Banking System (CBS). The system speeds up the on-site collection of information as well as loan appraisal as it can be done paperless if there is a mobile network available. The system is customisable to the client’s appraisal procedures and can provide instant SMS notifications.

Yasser EL JASOULI SIDI of MFI Insight Analytics presented how he introduced a customer management system in a previously manual process to improve and accelerate decision-making processes to provide credit. The system can make a credit assessment using financial data and alternative data consisting of mobile data, credit bureau data, utilities rent data and social media data. The system provides a maximum loan size and advises the loan officer which products to offer to the client. The credit scoring tool assists in analysing the assets in order to eliminate borrowers that are not credit worthy and therefore may affect portfolio delinquency and default probability. In the case presented, delinquency went down 50%, probability default rate decreased with 30% and manual credit assessments decreased by 20%. The customer management system has allowed an increase in caseload per officer of 134% and a decrease in time used for the loan decision-making process from 72 hours to 6 hours. 

Simon PRIOLLAUD, Lead DFS Consultant at Inbox, shared his lessons learned on customer segmentation based on his experiences in Africa. In many cases, Finance Institutions think they have a clear understanding of their portfolio but often they only understand “some” of their clients and not the big picture. If they want to serve clients better, they should start auditing their management information system (MIS) in order to segment clients and use this information to distribute Digital Financial Services accurately. Priollaud concluded by providing some key lessons from his experiences. Most MFIs probably already have the data needed to conduct a segmentation but need to take time to assess their MIS. Furthermore, he recommends to start small as soon as possible and to go step by step for these complex projects. 

DISCUSSION

A participant in the audience noted that data is already available in most cases but not well organised. How can organisations improve this issue? Mottet suggested to keep it simple at first and increase the use of data from that starting point. Breul stressed that it is important to start with having the process of data collection, analysis and decision/making in place so that you can build your MIS to support that continuous process. 

Another participant mentioned that big data is often perceived as disruptive, whereas the case of Mottet presented a more gradual transition path. The participant asked if Mottet had also experienced disruptions. Mottet replied that in most cases you have to start with small improvements in data collection or else the project is doomed to fail as companies no longer feel they are in control. 

A member from the audience also asked Lebel who owns the data collected in his open source Core Banking System. He replied that MFIs using the applications are owners of the information. Information is not distributed or disclosed to other actors in the value chain. On a similar note, El Jasouli Sidi explained to the audience how to ensure ethical use when collecting data on social media. He mentioned that this data is carefully weighed by an algorithm. 

Another question raised by the audience related to the usefulness of micro segmenting. Priollaud advised MFIs to keep segmentation “simple” at first. If you make it too complicated (and do not have adequate resources), the segmentation could turn obsolete before segmentation goes live.