IFC-World Bank SME Productivity Launchpad Program ALGORITHMS FOR INCLUSION: DATA DRIVEN LENDING FOR WOMEN OWNED SMEs June 2017 Authors: Salman Alibhai, Mengistu Bessir Achew, Rachel Coleman, Anushe Khan and Francesco Strobbe KEY MESSAGES: 1. All over the world, women have less access to credit than men. Because of both discriminatory property laws and unwritten social customs, women are less likely than men to own high-value assets that can be used as collateral to secure loans. 2. Financial institutions in developing countries rely on heavy collateral requirements because they don’t have enough information about their borrowers. New technologies – many emerging from financial technology (fintech) startups in the Silicon Valley – have the potential to generate data on borrowers that can replace traditional collateral requirements, and unlock finance for women. 3. In Ethiopia, we explored introducing fintech that could harness the data that financial institutions are already sitting on. The technology focuses on digitizing hard-copy loan application files of previous borrowers to identify trends and characteristics associated with repayment, and predict creditworthiness of new borrowers. 4. Fintech solutions can viably address the collateral constraint for women borrowers, and can work even in low tech environments. But technology adoption isn’t easy, and assessing the readiness of financial institutions to adopt fintech and embark on technological change is a critical first step. Can we find a viable solution to harness data to help financial institutions get around the collateral constraint to reach more women borrowers? A team from the Finance and Markets Global Practice, the Financial Institutions Group, and the Gender Innovation Lab, supported by the SME Launchpad Program, set out to understand the feasibility of digitizing customer data and introducing data- driven lending to improve access to finance for women owned SMEs in two microfinance institutions in Ethiopia. With limited information on their borrowers, financial WHAT CAN DATA institutions utilize heavy collateral requirements to minimize DRIVEN LENDING DO? their risk and exposure. However, many of these financial In the context of this pilot project, data driven lending institutions hold a lot of customer data in paper application involves using available data on previous borrowers and their forms. A single client has dozens of pages of application loan repayment history to identify patterns to understand materials, containing information on everything from the which customer profiles are likely to repay their loans. sector and size of their business, inventory and cashflows, Through this analysis a process can be developed called an assets in the household, education and employment history, algorithm that includes a set of rules to be followed when marital status, number of children and where they live. calculating if a client is eligible for a loan. This credit scoring However, after the initial loan decision is made, these data algorithm can then be used to predict the credit-worthiness points remain mostly in stacks of papers, serving little or no of new borrowers instead of relying heavily on collateral. use to the financial institution. A key to unlocking financial services which can meet the Credit scores serve as a solution to information asymmetry needs of women entrepreneurs is developing innovative between borrowers and lenders. There is a fundamental and effective loan appraisal mechanisms. In Ethiopia, MFIs inequity between borrowers and lenders because borrowers have sizeable portfolios of female group loan borrowers2, know how much they are willing and able to pay but lenders with women forming into groups to secure loans with do not know this information. A credit score serves the social collateral. These group loans typically reach a ceiling purpose of helping lenders bridge this gap of information at about $1,500 USD, however, and are not sufficient to and make a determination on someone’s probability to meet the needs of growth-oriented entrepreneurs. Lending default on a loan. Research shows that the data contained larger, individual loans to these growth-oriented clients is in application files such as gender, age, marital status, challenging, however, since the main screening tool MFIs dependents, having a telephone, educational level, and have is collateral. occupation are widely used in building scoring models.1 If all these thousands of files could be evaluated across MFIs in Ethiopia see data driven lending as a possible borrowers in order to identify trends or characteristics of solution to this collateral quandary. Effectively adopting good and bad borrowers financial institutions may be able data driven lending will allow financial institutions to lend to to lend to borrowers with less risk and lower collateral previous borrowers who could only receive groups loans, as requirements. well as new customers. Improved appraisal will enable MFIs to identify credit-worthy borrowers more accurately and forego requirements for large amounts of loan collateral. POOR GOOD LOAN FILES BORROWER’S ALGORITHM USED DIGITIZED HISTORICAL TO ASSESS NEW DATA ANALYZED BORROWERS 1 Orgler, 1971; Steenackers and Goovaerts, 1989; Lee et al., 2002; Banasik et al., 2003; Chen and Huang, 2003; Sarlija et al., 2004; Lee and Chen, 2005; Hand et al., 2005; Sustersic et al., 2009 2 As of March 2016, women represented 41% of the total number of MFI borrowers of MFIs in the Association of Ethiopian Microfinance Institutions (AEMFI). SME LAUNCHPAD PILOT READINESS ASSESSMENTS: Through the SME Launchpad Project, the team explored PROCESS & RESULTS possible partnerships with five promising fintech firms (First First Access developed a proprietary readiness assessment Access, Lenddo, FICO, DemystData and Verde) to explore tool consisting of a capacity diagnostic and a data the opportunity of working in Ethiopia on a data driven diagnostic, as a service provided to financial institutions lending assessment pilot. The team discussed with the firms that are interested in adopting data-driven lending how their technologies could work to address the collateral techniques. This tool was developed to determine a lending constraint women borrowers and financial institutions were institution’s level of preparedness for developing credit facing alike in Ethiopia. Through these discussions, the scoring algorithms and a roadmap for data driven lending team identified one fintech company that was deemed best implementation. This assessment phase is valuable to fit to work towards addressing the collateral challenge in ensure that financial institutions have a foundation in place to Ethiopia3. realize the benefits of a credit scoring platform. The team then travelled to Ethiopia to meet MFIs operating Through two types of assessments First Access is able around Addis Ababa to share the different promising to identify an institution’s strengths, weaknesses and technologies that could be implemented in their institutions opportunities. The capacity diagnostic strives to review to improve their lending to reach new and underserved capacity across business functions that are relevant for borrowers. The objective of this activity was to identify implementing credit scoring and the data diagnostic includes high-potential financial institutions interested in adopting an analysis of the institution’s data and technological technology and data-driven lending. sophistication. From these engagements a partnership was born between The assessment serves as a diagnostic tool to identify the World Bank Group, two Ethiopian MFIs and First roadblocks in the adoption of data-driven credit scoring Access, a fintech alternative credit scoring company. The solutions and provides recommendations on where to partners then worked together over the next several months allocate resources to better position institutions to adopt to conduct readiness assessments of the two MFIs and technology and fully utilize its existing data. It reviews five determine next steps towards adopting data driven lending. dimensions of readiness that have been determined as key components of financial institutions business (strategy, leadership, process, technology and information) to help financial institutions understand where they fall on the continuum of readiness for a data-driven future. PILOT PARTNERS: First Access: A NYC fintech firm, founded in 2011, offers a customizable credit scoring platform for lending institutions in emerging markets to credit score anyone. Wasasa Microfinance: A microfinance institution operating in Ethiopia with 80,000 borrowers across 34 branches and 20 rural service outlets, Wasasa is thriving to provide sustainable financial services to the poor in order to employ capital for poverty alleviation. Buusaa Gonofaa Microfinance: A microfinance institution operating in Ethiopia with 85,000 borrowers across 30 branches, Buusaa works to provide flexible and sustainable financial services to improve the livelihood of the resource poor households, with particular focus on women, the landless youth and smallholder farmers. World Bank SME Launchpad team: A team of World Bank Group staff working across the IFC and World Bank to investigate the opportunities data driven lending can bring to financial institutions to better reach women borrowers. The identified fintech offered the most suitable product given the current status of the Ethiopian financial sector but other fintechs offer more sophisticated solutions that can potentially be  3 brought into the Ethiopian market in the future. DIMENSIONS OF READINESS • Vision Having vision and a tactical plan for how lending STRATEGY • Clarity of priorities institutions can leverage data and adopt new • Measurement capabilities technologies is crucial to becoming truly data-driven. The right combination of project management, • Experience of project lead change management and leadership skills in an LEADERSHIP • Experience of project managers organization is key to developing a data culture. • Codification of credit policies To deploy customized algorithms and automate credit • Policy enforcement and adherence decisions, it is critical to have consistent processes PROCESS • Process standardization and strong policy compliance across the institution. • Credit risk management • Experience of technology team Core banking software, database and data • MIS and database functionality TECHNOLOGY warehouse, network and security policies are all • Data transfer capacity integral to creating a data-driven ecosystem. • Data security • Quantity of historical records Enforcing consistent and high-quality data collection, INFORMATION • Completeness of data management and storage practices is crucial • Data collection and storage for any data-driven business to succeed. Figure 1 ETHIOPIAN MFIs ASSESSMENTS STRENGTHS, WEAKNESSES AND OPPORTUNITIES TABLE: WASASA AND BUUSAA GONOFAA First Access traveled to Ethiopia to evaluate the two MFIs across the five dimensions of readiness (See Figure 1). STRENGTHS WEAKNESSES OPPORTUNITIES They held targeted interviews with senior leadership of Strong leadership Limited amount of First mover the financial institutions, observed general operations and with strategic vision data is currently advantage for data studied the loan assessment process. While in Ethiopia, First for the future of digitized (home and driven lending still their business business visit data exists in Ethiopian Access extracted a version of the MFIs’ databases to enable is not digitized) microfinance market analysis of available data. They later analyzed the quantity for institutions and quality of the institutions’ data to evaluate the feasibility willing to invest in technology of using the data for credit scoring. Rigorous Database Digitally capturing The readiness assessment reports then ranked each documentation of structures are loan application credit policies disorganized information would institutions’ capacity level across the sub dimensions of allow for increased each of the five dimensions of readiness. The two financial efficiency greater insight into credit institutions are at similar spots on the readiness spectrum. risk While they have much work to do to improve compliance Robust loan Under resourced Serving the and efficiency, they are also praised for their resourcefulness application compliance and market demand and their ability to power through challenges. Both and aspirations technology teams for individual loans to collect (reducing collateral institutions are tasked with investing in resources to build demographic, constraint) data and institutional capacity to be able to use the data in social and a way that will help them. Common strengths, weaknesses geographic data and opportunities for the financial institutions can be found Figure 2 in Figure 2. In the context of the Finance and Markets Women • Compliance enforcement Entrepreneurship Development Project (WEDP), the Digitizing the underwriting process and enforcing Launchpad team is working with the MFIs to operationalize policies automatically rather than through human to the roadmap for adopting data driven lending. Throughout human management aids in compliance efforts. This the pilot the MFIs have shown enthusiasm for the value data is particularly important for geographically dispersed driven lending could bring to their operations. Next steps will financial institutions. Operational inefficiencies and include working to codify processes and adopt electronic perceived risks can both be eliminated with better data. loan applications. This is expected to have immediate However, data cannot substitute for process. Compliance benefits because electronic applications will increase is the first step for data driven lending. Without processing speed of applications. This will also produce long compliance, an institution will not be able to generate term benefits through the collection of systematic data that consistent enough data to build effective credit scoring can be used for data driven lending down the road. algorithms. TOOLS NECESSARY FOR THE • Consistent data collection ADOPTION OF DATA DRIVEN LENDING Utilizing tools that enforce and support the collection and Preparedness for data-driven credit scoring is multifaceted. storage of loan application fields (MIS system, third party It requires a strong foundation of operational practices and software for tablet data collection, etc.) is key. Technology robust historical data. Tools and strategies that are essential can serve as a valuable enforcer of better compliance to have in place include: and allow financial institutions to realize greater efficiency. Also, it is important to make sure incentive structures • Strategic management are in place for credit officers for compliance in data It is important that leadership has experience managing collection. new strategic initiatives and technology solutions. In particular, financial institutions need a project manager • Effective data storage who can clearly communicate the value proposition of Storing data in a standardized structure with an easy to credit scoring, who has access to senior leadership, understand code book is imperative for analysis. decision-making power, knowledge of branch-level • Value of Tech Mindset: operations and a background in technology. Think of technology as a driver of growth not an • Codified business rules afterthought. It is important that the leaders of financial Strong credit manuals and policies and staff training on institutions develop a clear plan around messaging about these credit policies are key for effective compliance with the value proposition of data driven lending to ensure credit policies. there is buy-in from branch managers and credit officers • Monitoring and reporting LESSONS LEARNED Calculating dimensions such as cost per loan and For Lenders: Ethiopia is an environment with very low underwriting time per loan is essential to be able to levels of digitization. If financial institutions are interested track improvements from adopting data driven lending in adopting data driven credit scoring mechanisms, they techniques. Also, keeping timely, granular data on PAR need to be dedicated to investing in capturing and storing and arrears strengthens the predictive power of credit accurate data. Using the current manual process for data algorithms. capture and storage is not an optimal solution, as human touch will more likely lead to incomplete and possibly • Standardized processes inaccurate data capture. Moving to digitized data capture Digitizing disbursement data and loan application data is an important first step for the lenders to consider. and setting up systems to allow for storing multiple Lenders also must consider that good data is not enough records for repeat customers is important for algorithm to successfully adopt data driven lending. They need to be development. This allows financial institutions to capture dedicated to building institutional capacity so they are able data on approved and rejected loan applications and to use the data in a way that can help them improve the way changing personal details in the core banking software. their business is run. institutions often need to be convinced of the value of integrating technology in their operations. The World Bank has the opportunity to play a valuable role in the pairing of fintechs and financial institutions to increase access to finance for underserved populations by providing technical assistance and cost sharing support for financial institutions to pilot innovative interventions. For all: Conducting thorough feasibility assessments before implementing a fintech pilot is crucial to effectively integrate a fintech initiative in a financial institution’s operations. Understanding where a financial institution stands on the Before working with the World Bank on this project, spectrum of readiness including the state of their data, their First Access had not considered expanding into institutional capacity to effectively integrate a credit scoring Ethiopia. Not only have we now determined that initiative and their commitment to adopting technology is Ethiopia is a viable market for our platform, but we key. The World Bank Group could add value by establishing also received valuable market feedback to validate its own credit scoring assessment tool kit that can be used our current and future products. We hope to have a across projects and countries to evaluate where financial significant presence in the country in 2018 and to find institutions stand on the readiness spectrum of adopting more opportunities to collaborate with the World Bank credit scoring technology. The creation of such a toolkit in Ethiopia and beyond. would enable project teams to spend fewer resources on hiring firms to conduct proprietary assessments and provide Rohit Acharya, valuable information to understand the status quo before First Access Chief Data Scientist and Co-Founder embarking on data driven lending initiatives. THE FUTURE OF DATA DRIVEN LENDING Technology has a key place in increasing access to finance for underserved populations, including women owned SMEs. The WBG can play a game changing role in demystifying the process of how technology can transform lending practices in developing markets. Adopting data FINANCIAL AFRICAN driven lending must be done incrementally and must begin TECHNOLOGY FINANCIAL COMPANIES INSTITUTIONS by first evaluating where financial institutions currently stand. Technology adoption is attainable but it has to be done in small steps. If financial institutions are agile, go for the small For World Bank Group: There is space for the World wins and continuously push incremental change they will see Bank Group to serve as a mediator of fintech and financial valuable and sustainable changes in their operations and be institutions to establish effective working relationships. able to better serve borrowers. Through this pilot project, it became evident that fintechs often need support in operating in developing countries Exposing financial institutions to the most up to date and adapting their technology and procedures. They also breakthrough technologies encourages them to invest often have small teams with limited capacity. Many fintechs in innovations that support sustainable financial growth. are just starting to enter Sub-Saharan Africa and trying to Brokering relationships between financial institutions and pick new markets for their operations. First Access had the most innovative global financial technology companies not considered the Ethiopian market before this project but will allow the World Bank Group to spark pioneering change now they are developing a new product that was informed to increase access to finance for undeserved borrowers, by their work in Ethiopia and feel it is a promising market particularly women entrepreneurs. for their work moving forward. On the other hand, financial