Digital Credit for Smallholder Farmers Lessons Learned from the Field TABLE OF CONTENTS 3 Acronyms 4 Foreword 4 Acknowledgments 5 Executive Summary 9 What Challenge We Set Out to Address & How We Approached It 16 What We Did 32 What We Learned 43 Recommendations THE WORLD BANK | Digital Credit for Smallholder Farmers: Lessons Learned 2 ACRONYMS BRAC BRAC Myanmar Microfinance Company Limited MADB Myanmar Agricultural Development Bank CGAP Consultative Group to Assist the Poor CSA Climate Smart Agriculture DFA Digital Field Application DFS Digital Financial Services EU European Union GDP Gross Domestic Product HCD Human-Centered Design KYC Know Your Customer MAP Making Access Possible MFI Micro-Finance Institution MMK Myanmar Kyat MNO Mobile Network Operator QR Code Quick Response Code SMS Short Message Service SIM Subscriber Identify Module WBG World Bank Group THE WORLD BANK | Digital Credit for Smallholder Farmers: Lessons Learned 3 FOREWORD Project activities took place between early 2017 and late 2020. With the onset of the global pandemic in the early spring of 2020, in-country operations rapidly decelerated as the government implemented a series of restrictions related to social distancing, movement, and service hours of operation. Project operations effectively ceased by the end of 2020. ACKNOWLEDGMENTS This report was prepared by Nicholas Lesher, Asuka Okumura, and Iulia Bogatu under the overall direction of Christopher Ian Brett and Julian A. Lampietti. The report was edited by Gunnar Larson and designed by Emma Schwartz. The project team wishes to thank Md. Sazaduzzaman, Wai Hin Aung, and the entire Bago region team of BRAC Myanmar Microfinance Co., Ltd; Jose Vahl and the entire L-IFT team; Telenor Myanmar; Ongo; aWhere; and Experian. The team also expresses its appreciation to Roy Parizat, Andro Koutsoudis, and Emilia Klimiuk for their contributions during the early stages of the project. This publication benefited greatly from attentive reviewing and editing support from several people. The project team would like to directly thank the following individuals for their contributions: Hira Channa (World Bank), Russell Toth (The University of Sidney), Dean Caire (independent consultant), Johan Hubert Roest (CGAP), Josh Woodard (independent consultant), Willem G. Janssen (World Bank), Jamie Barbara Anderson (CGAP), and Toshiaki Ono (World Bank). The project team also recognizes the support and financial contribution from the government of South Korea through the Korea-World Bank Group Partnership Facility, which allowed this project to proceed. THE WORLD BANK | Digital Credit for Smallholder Farmers: Lessons Learned 4 Executive Summary The ongoing digital revolution could be data-driven digital agriculture to improve food system objective was i) to evaluate the ability of digitally outcomes. enabled financial services to deliver appropriately transformative not only for the financial designed loan products to smallholder farmers, inclusion of smallholder farmers but also This report’s primary purpose is to inform and share thereby enabling them to invest in seasonal inputs for the food and agriculture sector more relevant experiences with those who have a role in to improve their productivity and ii) to assess the broadly, supporting the creation of more designing or operating digitally enabled services for commercial sustainability of financial institutions to efficient, equitable, and environmentally smallholder farmers—principally financial services provide such loans to farmers. sustainable operations. but with broader relevance for other rural-facing services. In it, we document lessons learned from an The project was implemented in a relatively narrow agri-credit pilot undertaken in Myanmar involving the time frame from 2018 to 2020 within a concentrated Digital technology use can help reduce costs, help microfinance institution BRAC Myanmar Microfinance geographic area—two townships, Daik-U and farmers make more informed and precise decisions, Co., Ltd (BRAC) and a small consortium of other Nyaung Lay Pin in the eastern Bago Region of and improve access to information, knowledge, and private sector partners. The project’s overarching southern central Myanmar. The loan product itself markets. The World Bank Group (WBG) focuses on THE WORLD BANK | Digital Credit for Smallholder Farmers: Lessons Learned 5 was launched in three iterative cycles over a one- terms of topography, ethnicity, language, and basic Product design appropriateness: This loan product and-a-half-year period. Refinements were made to public infrastructure. The project team therefore allowed a majority of borrowers to purchase some or both the product and associated processes after does not propose that the findings, lessons, or most of their seasonal inputs, with a few exceptions each cycle based on experiences and reactions of recommendations contained here are nationally in which the loan size could not cover the full key personnel and borrowers. representative. Nor is the team of the view that seasonal investment required or the funds were these experiences can be easily replicated within not available when farmers wanted to make key In total, 2,626 farmers were reached and 1,441 Myanmar or outside without careful consideration purchases. It also offered terms and conditions that loans totaling an estimated MMK 420,000,000 of key factors in specific localities where similar were flexible and responsive to cyclical agricultural (US$300,000) were disbursed. The pilot team’s initiatives might be introduced. growing patterns and an interest rate comparable to approach sought to thoroughly understand the circumstances and the needs of their target customers, using a process known as human- centered design (HCD). Based on the findings from field research, which revealed passive usage of mobile phones among smallholder farmers despite the broad device penetration nationally, the project took a “high touch” approach to implement the pilot. Key Findings While the experiences and outcomes from this project are context specific—a single market, a single loan product, offered by a single provider over the course of only three growing seasons— this report aims to distill and present high level lessons that possess broad applicability to a range of market contexts involving digitization and rural service delivery. Myanmar varies considerably in THE WORLD BANK | Digital Credit for Smallholder Farmers: Lessons Learned 6 or more favorable than the prevailing market rate for mobile money service provider Ongo to access debt. It is important to anticipate the need for financial similar formal loan products. digital loan disbursement and collection services literacy training and a tailored communication strategy. via its agent network. Commercial Sustainability: Overall, the loan portfolio Digitization is not a silver bullet for smallholder returned a profit in each of the three lending phases, Lessons Learned farmers resulting in a net profit for BRAC. Based on multiple Capturing and properly understanding the local context Smallholder farmer demand for an individual commercial decisions taken by BRAC during the final is a vital first step, one where a human centered design agri-loan product month of the pilot project period, this loan product (HCD) approach has merit. But it must also be followed Rural demand among farmers for an individual appears to represent a commercially viable offering. by a diligent application of that understanding during seasonal loan product exists and appears strong. Namely, just prior to the cessation of in-market the design and piloting phases. A well-designed These are producers who operate on less than five project activities and the onset of COVID19, BRAC product or service and an effective strategy for rural acres or who typically rent land for crop production. decided to: market segments will need to consider what degree While repayment rates for agricultural loans are quite of digitization is appropriate to pursue, at what pace, high, this loan product arrowed farmers who typically i. geographically expand access to this product and involving which intended end users. It must also would not qualify for or seek loans from traditional rural across 16 branches anticipate and address issues related to i) gender, lenders such as the Myanmar Agricultural Development ii. invest its own funds to develop or acquire the ii) building confidence and capacity in a product or Bank (MADB). That said, the repayment rate exceeded necessary digital infrastructure and build internal service and what the benefits of usage are, iii) the role 97 percent in each of the three loan cycles, with the staff capacity to deploy a digital field application of human interaction, as well as iv) trust, and v) the final cycle overlapping with the outbreak of a global (DFA) capability across all product lines personal privacy of rural customers. pandemic. Additionally, the project team observed a iii. commit to an agri-loan portfolio (the only other preference for individualized loans over group loans. Digitization from the service provider BRAC market outside Bangladesh to do so), Participation in the former can have the effect of perspective with a target customer base of 18 percent of its limiting one’s access to capital if one or more group 450,000 total customer base by 2024 While the potential exists for cost and risk reduction, members underperform, and that jeopardizes access digitization should be viewed as a gradual process iv. pursue an alternative credit scoring model as a for other members in good standing. The product must rather than a quick and seamless integration of new near term objective and fit the circumstances and activity patterns that shape a technology and systems with existing ones. It will v. broaden its commercial relationship with the rural borrower’s financial health and capacity to absorb THE WORLD BANK | Digital Credit for Smallholder Farmers: Lessons Learned 7 require adequate investments of time and resources 4. Prioritize and adequately fund an internal “digital companies, technology firms, or academic to align structures, processes, and staffing. It readiness” assessment to surface training needs institutions to obtain data relevant to smallholder also requires articulating a strategy that identifies across departments and at different staffing levels. farmer credit risk assessment. Such partnerships digitization’s potential to clearly address current pain 5. Do not exclude human-based operations (for should also clearly articulate the purpose for data points in the system as well as its limitations and the instance any part of loan application origination collection to ensure it is focused on augmenting need to maintain certain human-based operations and specific validation tasks) entirely where credit risk analysis. and interactions. limitations of basic rural infrastructure exist and 9. The public sector has a role in ensuring a healthy cannot power digital services reliably at scale. If enabling environment for rural-facing products and Recommendations digital content such as loan applications cannot be services, including digital information standards, 1. Employ an agile and iterative design approach, generated, transferred, or accessed consistently, it digital infrastructure, and customer redress such as HCD, that incorporates principles and will likely slow account activation or service delivery mechanisms. methods capable of surfacing and responding activities and restrict a provider’s ability to expand 10. Digitization of certain public sector services could to relevant practices, perceptions, and patterns geographically and grow its customer base. de-risk agri sector investment from the private of rural customers as well as key needs and 6. Satisfying account registration requirements (e.g. sector, such as citizenship identification and data challenges they currently encounter when Know Your Customer or KYC information) for rural sets related to climate, weather, and agronomy. deploying a product. customers will require creativity with the kind of 2. Where appropriate, let rural customers interact information collected and direct engagement with with digitized services through credible, trusted smallholder farmers. This may include working intermediaries, thereby allowing them to set their in well-defined tiers, in which KYC requirements own pace and usage of digitization. increase as service value increases. 3. A rural-facing offering does not need to be fully 7. Explore partnerships for financial and digital digitized when launched; rather, an incremental literacy training that leverage external networks, approach may be warranted that starts with experience, and capacity of partners (e.g. public or digitizing “low hanging fruit” to give stakeholders development-led rural extension services), rather an opportunity to absorb and adjust to the new than building out this capacity in-house. product and build trust in the service. 8. Explore potential partnerships with agritech THE WORLD BANK | Digital Credit for Smallholder Farmers: Lessons Learned 8 What Challenge We Set Out to Address & How We Approached It Restricted Access to Rural/Agri-Finance climate change, which will dramatically impact food financial services. With respect to the latter, loans Among Smallholder Farmers production globally. A failure to react appropriately need to be designed with affordable interest rates will result in significant increases in food insecurity, and according to repayment terms appropriately Agricultural production is dominated by smallholder poverty, and malnutrition. The implementation tailored to their farming patterns and practices. farmers in most developing countries. There are of Climate Smart Agriculture (CSA) could greatly several reasons preventing them from adequately benefit smallholder farmers.2 However, this requires The percentage of global investing in their farms and improving production as investment in new inputs such as drought resistant demand for smallholder well as income earning potential, including a lack seed and in farmer capacity to comprehend and 70% farmer household finance of sufficient access to capital.1 These farmers are apply new farming techniques. This cannot happen that is unmet, equivalent also contending with the accelerating impacts of without access to adequate non-financial and to around US $170 billion.3 1 Other factors include: insecure land rights, prejudicial land cadaster management, lack of access to quality inputs and agri-related services such as land preparation or irrigation, weak agri-extension services, poor market linkages to buyers, weak negotiation positions during harvest periods. 2 Climate-smart agriculture (CSA) is an integrated approach to address the interlinked challenges of food security and climate change with the aims of increased productivity, enhanced resilience, and reduced emissions. 3 2019 Pathways to Prosperity report published by the Rural & Agriculture Finance Learning Lab. THE WORLD BANK | Digital Credit for Smallholder Farmers: Lessons Learned 9 Long-term financing technologies globally creates new opportunities This report documents lessons learned from a accounts for 98% of unmet to overcome some of these chronic challenges. digitally enabled agri-credit product pilot undertaken demand, but short-term The growth in digital payment could increase in Myanmar from 2017 to 2020. Although many agricultural financing needs—estimated to be the financial inclusion of smallholder farmers that of the experiences and results are unique to this US $66 billion—is under-supplied as well (equal enables pro-poor agricultural development to single market, the project has yielded observations to a 67% shortfall).4 Financing gaps inhibit the reduce poverty and increase food security.5 and findings with relevance beyond that country’s economic growth of smallholder farmers and further borders. Many of the lessons will be useful to others perpetuate existing cycles of poverty. There are The ongoing digital revolution could be seeking to address similar challenges elsewhere. myriad reasons behind this shortfall. Smallholder transformative not only for the financial inclusion This applies not only to leveraging digital technology farmers often have scant collateral, few alternative of smallholder farmers but also for the food and to support agricultural credit service access and sources of income, and limited credit history. They agriculture sector more broadly, supporting delivery but also to effectively and appropriately live in remote areas which are far away from financial the creation of more efficient, equitable, and designing services that are responsive to the activity institutions. For formal financial institutions, it is costly environmentally sustainable operations. Digital patterns, capabilities, and needs of smallholder to promote, originate, and operate financial services technology use can help reduce costs, help farmers farmer communities more broadly. in rural communities due to their remote location as make more informed and precise decisions, and well as the relatively smaller loan transaction sizes. improve access to information, knowledge, and Market Context: Significant Change in a Risk analysis is challenging due to the lack of their markets. The World Bank Group (WBG) focuses Short Period of Time credit history as well as limited understanding of on data-driven digital agriculture to improve food Agriculture accounts for approximately one third of risks associated with agricultural production. They system outcomes as shown in its strategy papers Myanmar’s GDP, making it one of the economy’s also have an insufficient understanding of the real – “Future of Food: Harnessing Digital Technologies largest sectors. Agriculture employment as a needs and circumstances of smallholder farmers. to Improve Food System Outcomes”6 and “What’s percentage of total employment is 84 percent in Cooking : Digital Transformation of the Agrifood rural areas and among poor households.9 Myanmar’s Fortunately, the growing prevalence of digital System.”7,8 agricultural productivity lags behind other countries 4 2019 Pathways to Prosperity report published by the Rural & Agriculture Finance Learning Lab. 5 https://www.cgap.org/research/publication/female-smallholders-financial-inclusion-agenda 6 https://www.worldbank.org/en/topic/agriculture/publication/future-of-food-harnessing-digital-technologies-to-improve-food-system-outcomes 7 http://hdl.handle.net/10986/35216 8 See the following also for WBG’s approach in this agenda: http://hdl.handle.net/10986/33961 9 http://documents1.worldbank.org/curated/en/806001593183687694/pdf/Myanmar-Economic-Monitor-Myanmar-in-the- Time-of-COVID-19.pdf THE WORLD BANK | Digital Credit for Smallholder Farmers: Lessons Learned 10 in the region.10 Multiple factors contribute to this FIGURE 1: Sector Overview underperformance, including inadequate supply of Access has expanded rapidly, with public goods, limited supply and low application of One of Myanmar’s largest sectors, almost 75% of farmers financially employing 84% of rural and poor included. In 2018, 48% of adults had quality inputs, low levels of on-farm mechanization, households, but productivity still lags access to an account at a formal financial and lack of access to financing for farm investments. AGRICULTURE behind other countries. FINANCIAL institution. SERVICES Women are much more likely to earn With respect to financial services, Myanmar has Mobile phone penetration is close to less than the poverty level (51% vs. 28%). 46 million unique subscribers, or about made great strides in expanding financial inclusion in Nearly half (49%) are dependent on 84% of the total population. Of those, someone else for income, recent years. The proportion of adults with accounts 80% use smartphones. ICT/MOBILES GENDER compared to 17% of men. at formal financial institutions increased from 23 to 48 percent between 2013 and 2018.11 Surprisingly particularly those without land titles, often have production. Loans are commonly structured as group though, financial inclusion amongst farmers is limited options when it comes to accessing credit. loans with joint liabilities, and full repayment required relatively high. The UNCDF-led Making Access Sources of informal lending such as lead farmers, immediately after harvest.13 Possible (MAP) Diagnostic from 2018 found that 52 input dealers (in-kind credits) or commodity off- percent of farmers have access to formal financial takers (pre-harvest cash advances to secure access Structural challenges are also associated with services mainly thanks to the Myanmar Agricultural to crops) frequently involve higher interest rates. expanding appropriate financial products to Development Bank (MADB). 62 percent of farmers Public banks, such as MADB, offer attractive interest smallholder farmers in Myanmar and other have some form of access to formal or informal rates but lending is concentrated on rice production. developing countries. These include legal, credit.12 Moreover, the timing and structure of MADB loans regulatory, and infrastructural roadblocks. The have not historically been well-designed to meet Ministry of Agriculture, Livestock and Irrigation itself Access to finance alone, however, is not sufficient if the requirements of farmers. Funds are typically identified many of these factors in its Myanmar the financial products are not designed to meet the released after planting, capped lending amounts tied Agriculture Development Strategy and Investment needs of smallholder farmers. Smallholder farmers, to plot size do not always cover the total costs of Plan 2018-2022.14,15 10 https://documents.worldbank.org/en/publication/documents-reports/documentdetail/509581468181132091/myanmar-analysis-of- farm-production-economics 11 https://www.uncdf.org/article/4500/map-myanmar-diagnostic-2018 12 Ibid 13 https://www.marketlinks.org/sites/marketlinks.org/files/resource/files/MyanmarDFSAssessment_Report_Long.pdf 14 https://www.uncdf.org/article/5612/myanmar-financial-inclusion-roadmap-20192023 15 http://extwprlegs1.fao.org/docs/pdf/mya180003.pdf THE WORLD BANK | Digital Credit for Smallholder Farmers: Lessons Learned 11 Yet, there are also opportunities to expand access Mobile Financial Services in 2016. Mobile network to formal financial products, such as a credit, to operators, banks, and fintechs have taken steps to smallholder farmers in ways that are now more cost digitize financial services and leverage the rapid effective than traditional strategies and systems in expansion of mobile technology in particular. These part thanks to the rapid expansion of the country’s factors bode well for the advancement of financial mobile infrastructure. Before 2013, a SIM card in services and are particularly relevant to smallholder Myanmar cost nearly US$3,000, leaving mobile farmers in a country where it takes on average more phones well out of reach for average citizens. than an hour for a rural inhabitant to reach a bank, However, with privatization of the country’s mobile ATM, or mobile money agent from their home, and telecom sector, competitive pressures drove the 43 minutes to reach an MFI branch.18 By comparison, price down precipitously, to just over one dollar in farmers reported traveling shorter distances to reach 2017.16 Recent data estimates unique SIM subscriber other relevant service locations, such as grocery someone else for income, such as from remittances penetration at 108 percent as many people stores (11 minutes), and public transport (29 minutes). (compared to just 17 percent of men).19 subscribe on a pre-paid basis and will swap SIMs to take advantage of lower voice/data rates when Gender Disparities in Service Access: • Though the situation varies by crop and by calling on the same carrier network. Of that 108 Persistent Gaps, Especially Within Rural region, male-headed households in general percent of total unique SIM subscribers, 80 percent Communities achieve higher incomes and profits than female- own smartphone devices.17 Female-headed households account for 24.1 percent headed households.20 of urban households and 16.2 percent in rural areas. • Women’s ownership of land title is more limited The state of digital financial services in Myanmar Women in Myanmar are significantly more likely than than that of men. 98 percent of male-headed has also expanded rapidly since the Central Bank of men to earn less than the poverty level (51% vs. 28%) households have access to agricultural land, Myanmar first released its Regulations on and nearly half of women (49%) are dependent on while 61 percent of female-headed households had access to land.21 16 https://www.telenor.com/wp-content/uploads/2018/02/Telenor-Realising-Digital-Myanmar-Report-06-February.pdf 17 https://www.gsma.com/mobilefordevelopment/wp-content/uploads/2020/02/Digital_credit_scoring_for_farmers_Opportuni- ties_for_agritech_companies_in_Myanmar.pdf 18 https://finmark.org.za/system/documents/files/000/000/187/original/Myanmar_Diagnostic_2018_CB3_repro.pdf?1601968844 19 https://documents.worldbank.org/en/publication/documents-reports/documentdetail/509581468181132091/myanmar-analy- sis-of-farm-production-economics 20 https://www.uncdf.org/article/5596/finscope-myanmar-gender-note 21 https://www.adb.org/sites/default/files/institutional-document/209596/gender-equality-womens-rights-myanmar.pdf THE WORLD BANK | Digital Credit for Smallholder Farmers: Lessons Learned 12 borrow at all, and of those who do, only a quarter (25%) do so from a formal financial institution.23,24 Moreover, while Myanmar has a relatively literate population, noticeable disparities persist between women and men. The literacy rate among women is 72 percent, compared with 80 percent among men.25 Similar disparities are seen in terms of digital literacy and access. Women are less likely to own a phone than men, whether smartphone (60% vs. 65%) or basic phone (15% vs. 21%), and the number of women reportedly using a phone trails that of men as well (76% vs. 80%).26 These gender dynamics were particularly relevant for the project team when designing its approach which resulted in 68 percent of loan recipients being women. Approach & Implementation Partners Women are also less likely than men to have formal are contributing factors to why a third of rural women This project intended to make appropriately forms of identification, such as a national registration in Myanmar (31%) were financially excluded as of designed individualized loans more accessible card, title deed, or house registration. This puts 2018 and another quarter (24%) only have access to to smallholder farmers, thereby enabling them to them at a severe disadvantage in terms of access informal financial services, which tend to be more improve their productivity by investing in seasonal to finance where such documents are required for costly than formal financial services. When it comes inputs. This entailed supporting a financial institution know your customer (KYC) verification.22 All of these to credit, more than half of women (54%) do not 22 https://www.uncdf.org/article/5596/finscope-myanmar-gender-note 23 Ibid 24 According to Making Access Possible (MAP) Diagnostic, gender disaggregated data (men vs women) for those points are: financially excluded (30% vs 31%), access to informal service only (20% vs 23%). 25 https://en.unesco.org/countries/myanmar 26 https://www.uncdf.org/article/5596/finscope-myanmar-gender-note THE WORLD BANK | Digital Credit for Smallholder Farmers: Lessons Learned 13 to test the effectiveness and commercial viability of FIGURE 2: Approach to Digitization digitally enabled financial services. Digitization was applied during the application, disbursement, and collection processes in rural Two key challenges apply in offering affordable villages as well as in offices. Multiple users were envisioned, including smallholder farmers. credit to smallholder farmers. The first is the cost DIGITIZATION APPLIED of initiating and servicing loans given the remote APPLICATION DISBURSEMENT COLLECTION locations of many of farmers relative to formal provider branches. The second is analyzing the risk this lending entails given the limited information available about the clients. The project therefore 1 2 3 4 5 6 sought to use digital technologies to reduce the cost and improve risk management to financial To source non- To collect To open accounts To review To initiate To process financial data information on and originate and process disbursements. collections. institutions in the following ways. (mobile and agri rural customers applications. applications. data). (both text and • Lower acquisition and operational costs by photos). allowing the financial institution partner to digitize key aspects of the process, such as customer risk profiles, many of whom do not have an common goal and each party’s roles and gains. onboarding, application processing, loan existing credit history. In particular, the project We developed a consortium of private service disbursement, and repayment. This would allow made use of two sources of non-financial data: providers around the key partners to build a for more efficient data collection, information 1) activity patterns linked to mobile phone platform to implement the pilot. The key partners, sharing, review, and analysis, as well as free usage,27 and 2) satellite imagery and other digital BRAC (microfinance institution) and Telenor (mobile up staff time to be allocated toward more data to gauge growing conditions as well as network operator) contributed their own resources to operational or revenue-generating tasks. production potential and risks of specific crops. the pilot to obtain lessons and build a business case. The World Bank provided i) technical assistance in • Improved accuracy in risk analysis through The first and difficult step was to identify key project design and implementation and ii) financial digital credit assessments, which has the partners with whom we could agree on a support to build the platform. Other stakeholders potential to provide greater insights into farmers’ 27 More precisely, the activities tied to a specific SIM card, which is used for mobile phone communication. THE WORLD BANK | Digital Credit for Smallholder Farmers: Lessons Learned 14 each played important roles to bring this product to FIGURE 3: Implementation Partners market (see Figure 3). Context & Impact of COVID-19 The implementation of The country confirmed its first official case of this pilot relied upon COVID-19 on March 23, 2020, which coincided with a multi-stakeholder The World Bank BRAC MFI Telenor the repayment period of the third and final lending partnership to bring The Facilitator The Lender The Phone Data cycle during this pilot project. The Government of together all of the Coordinated with project Responsible for all aspects One of the leading mobile Myanmar implemented nationwide restrictions on pieces necessary to partners, supporting research of the loan origination and network operators in physical movement, leading to broad disruptions effectively deliver this and product/process design, servicing process, as well Myanmar; provided the and capturing learnings from as contributing anonymized phone data, which was used throughout the economy, including the agriculture unique loan product to data on customer loans and to help calculate one aspect the pilot experience. and microfinance sectors. The lockdowns in farmers. repayment. of the credit scoring. neighboring countries caused Myanmar immigrants in those countries to lose jobs and prevented them from making remittances to their family in rural Myanmar. L-IFT Ongo aWhere Experian BRAC suspended financing activities in late March The Rural Promoter The Mobile Money Provider The Ag Data The Platform Provider and implemented response plans that included Supported go-to-market Provided mobile money A weather-based agricultural Provided access to a digital suspending group meetings, adhering to social activities and provide hands- agent support in the pilot intelligence company; lending platform and a credit distancing protocols, and disseminating public health on client support during the area to help with loan provided the models and scoring model that drew on loan application process disbursement and repayment agricultural data that drove non-financial data sourced information. BRAC’s country leadership decided to in the rural communities and ensure adequate levels the agricultural risk part of the from Telenor and aWhere. leverage and pivot its marketing and sales capacity selected for this pilot. of physical liquidity were credit score. present. to support public awareness and information dissemination. It also decided to issue waivers to those farmers borrowing during the third cycle because the timing of the two repayments came during a period of heightened market volatility. THE WORLD BANK | Digital Credit for Smallholder Farmers: Lessons Learned 15 What We Did Key Project Components & Pilot iterative cycles over a one-and-a-half-year period. location in rural townships where BRAC had an Overview Refinements were made to both the product and existing branch presence. Some may have already associated processes after each cycle based on been aware of BRAC and may have been former The project was implemented from 2018 to 2020 in experiences and reactions of both key personnel group loan clients—BRAC’s flagship rural product. two townships, Daik-U and Nyaung Lay Pin, in the and borrowers. In total, 2,626 farmers were provided They may also have taken out agri-loans from formal eastern Bago Region of southern central Myanmar. information about the loan by L-IFT and BRAC service providers. But there were no requirements The pilot centered around six key activities: across all three cycles, of which, 2,051 applied that borrowers during this pilot project be existing i) primary market research, ii) product strategy for loans, 1,491 were approved for loans, and 1,441 or past BRAC clients or have successfully repaid and roadmap development, iii) digital platform loans totaling an estimated MMK 420,000,000 a similar loan in the past. In the region where the development, iv) personnel training, v) smallholder (US$300,000) were disbursed in total.28 pilot was conducted, there are two distinct cropping farmer outreach, and vi) pilot implementation. seasons–monsoon and dry. Farmers who received These farmers were identified based on their loans (typically between US$100-330) cultivated on The loan product itself was launched in three 28 The number of loan applications, approvals and disbursements includes repeat borrowers who participated in more than one loan cycle. THE WORLD BANK | Digital Credit for Smallholder Farmers: Lessons Learned 16 average plot sizes of five acres. Few had access farmers could access multiple buyers (nearby millers, search of lower prices and to collect information. to irrigation during the dry season. The dominant middlemen, further markets, etc.) and many farmers They rarely rely on a single shop for input supplies. selling crop was rice—grown during the monsoon would decide to sell based on the prevailing spot They also keep livestock as an additional revenue season—followed by bean varieties, seasonal price. Loans funds were spent on a variety of inputs, stream. When cash needs spike, many farmers vegetables, and maize in the dry season, which predominantly seeds and fertilizer. In terms of input would work as daily laborers at other farms or in they often used for household consumption. These supply, many farmers prefer to shop around in some other wage-earning capacity. FIGURE 4: Project Components & Pilot Overview MAR ‘17 APR ‘18 MAY ‘18 SEPT ‘18 NOV ‘18 MAY ‘19 NOV ‘19 MAR ‘20 MAY ‘20 PARTNERSHIP BUILDING RESEARCH & DESIGN SCORING ALGORITHM GO TO MARKET FIRST LOAN REFINE SECOND LOAN REFINE THIRD LOAN REFINE ANALYSIS & LESSONS LEARNED THE WORLD BANK | Digital Credit for Smallholder Farmers: Lessons Learned 17 FIGURE 5: Overview of Initial Agri-Loan Concept BRAC’s standard loan screening criteria includes checking: Smallholder farmers WHO (male and female) • Household financial position • Family awareness of loan with or without land title • Age • Years at present address • Reputation • Early warning indicators Issued to an individual HOW • Prior loan delinquency • Personal emergency care rather than a group • Prior application falsification • Current loans To purchase seasonal inputs for a variety of WHAT different crops Rural customer engagement with a or development sector actors. These approaches lead to stronger uptake and improved outcomes. may use a baseline study, followed by a midline human-centered design (HCD) approach 1-2 years later, and then an endline study at project This project followed an HCD product design process The pilot team used a process known as human- closing. Such approaches may yield lessons when that included two rounds of field research in the pilot centered design (HCD) to develop an understanding of the window to strategize and implement adjustments location that engaged not only smallholder farmers the needs and circumstances of target customers.29 or improvements has passed. The continual iteration but also other rural market actors. These included used during HCD allows for tweaks and refinement input suppliers, local millers, mobile money agents, The iterative process that is built into HCD is a on an ongoing basis. By design, this allows for greater and BRAC’s local credit officers—actors that might distinguishing factor from traditional approaches to responsiveness to experiences testing a service or have a role in the new product or help the team better design and implementation used by some public product with actual users in a market. This in turn can understand certain needs, attitudes, challenges, or 29 The HCD approach evolved from earlier versions of design thinking dating back to the 1960’s, where it began as “participatory design” employed by anthropologists and development practitioners. See authors such as Herbert Simon and Horst Rittet. It then progressed into “user-centered design” and “meta design”, where it was primarily employed by industrial engineers and technol- ogists developing new products and services. See authors Brian Dawson and Donald Norman. In the late 1990’s early 2000’s, “service design” and “HCD” emerged as more of a mindset that emphasized collaboration, multidisciplinary approaches, holistic community development, and a focus on empathy. See authors William Rousse and Lucy Kimbell. Applying an HCD approach to develop programming is becoming more common these days. For example: https://www.unicef.org/innovation/media/5456/file THE WORLD BANK | Digital Credit for Smallholder Farmers: Lessons Learned 18 opportunities of smallholder farmers that would be FIGURE 6: Illustration of HCD Process relevant to design thinking and product development. F e a sible Human-Centered Design is a set of methods and guiding principles that help organizations discover and The initial round of research sought to document address the needs of their end users. relevant practices of smallholder farmers, specifically ble ir a agricultural and farming practices as well as mobile s GROUNDED IN De COLLABORATIVE technology and borrowing. Of 23 interviews conducted, KNOWLEDGE Complex problems require Vi 13 (57%) were with women. A loan product prototype ABOUT USERS broad and deep expertise abl was co-designed with BRAC and other consortium Before we create solutions, and cannot be solved by one e we talk to users to understand person or organization. stakeholders. It was then tested with potential female goals, needs, and attitudes. and male rural borrowers to gauge receptivity to how the product would be structured, priced, distributed, VISUAL ITERATIVE and accessed, as well as to determine whether any Visualizing helps us to see Prototyping ideas early and often, and problems, communicate testing them with users ensures that our pre-pilot launch modifications were warranted. complex ideas, and identify concepts meet our objectives before new opportunities. spending time and money creating high- fidelity designs. FIGURE 7: Round 1 Interviewees THE WORLD BANK | Digital Credit for Smallholder Farmers: Lessons Learned 19 FIGURE 8: Agricultural Practices There is not much flexibility SOIL LAND AVAILABILITY in terms of what farmers QUALITY OWNERSHIP OF LABOR can grow. Soil quality, land ownership, and availability of labor often drive farming decisions. The only thing Most farmers When labor is not that grows here here don’t own available, I don’t “In the growing season, [farmers] is paddy. their land, so grow summer need money to invest … like for they cannot do paddy. fish farming even land preparation or that sort of though it is very thing. Some people keep seeds profitable. from last year… but due to climate change, sometimes they need My land is uneven, to buy seed again. So they need so I can only grow money and then borrow it at a summer paddy on half of my land. high interest rate in the community. Now that this kind of loan is available... offering low interest rate loans to local farmers it is very supportive.” — BRAC Branch Manager, Daik-U THE WORLD BANK | Digital Credit for Smallholder Farmers: Lessons Learned 20 FIGURE 9: Mobile Phone Practices TECH Many smart phone owners don’t use SAVVY I use my phone for my business. any ‘smart’ features. I’m happy to try new things. • Makes & receives calls/SMS • Facebook Smart phones are attractive to farmers • Viber, file sharing because the bigger dialer makes calling • Mobile money easier. Few buy them to use features I use my phone to communicate beyond calling. and learn so I can keep up. • Makes & receives calls/SMS • There is a low correlation between • Saves contacts • Has Facebook account the kind of phone someone owns • Uses file sharing apps and the complexity of the tasks they KEYPAD PHONE SMART PHONE are able to perform. I can’t afford a smart I share my phone with my son. I don’t • Few people use SMS. phone right now, but I know what he does on it. I just use it for could learn anything as calling. • We didn’t see significant gender long as someone shows • Makes & receives calls me how. differences. • Occasionally looks at someone else’s • Makes & receives calls Facebook for news • Saves contacts • Receives SMS • Loads airtime I don’t know how to save contacts, • Borrows airtime my daughter does it for me. I don’t need to know these things. • Makes & receives calls TECH AVERSE THE WORLD BANK | Digital Credit for Smallholder Farmers: Lessons Learned 21 Key findings include: FIGURE 10: Selection of Farmers’ Phones • Instances of informal and formal borrowing on a seasonally recurring basis for farming were reported by participants, with more men than women engaging in this type of activity. • Some participants were averse to taking on loans because of a feeling of embarrassment or a fear that they would be unable to repay the debt. • The majority of farmers interviewed own smartphones. However, some of the smartphones were low-end models with factory settings pre-set to a foreign language the users could not understand.30 • Few knew how to use smartphones for more than just making phone calls. Most farmers were not comfortable buying airtime on their own and only a few were familiar with receiving, opening, or replying to SMS text messages.31 Older children often play the role of device purchaser and instructor. Without their support, many farmers are unlikely to perform more complicated tasks. 30 So-called “fake smartphones,” which falsely carry a major manufacturer’s brand, were observed. They can be used for basic func- tions, such as making phone calls, but their ability to properly run certain apps can be unreliable and is difficult to predict in advance. 31 This is due to a number of factors, including but not limited to, competing font encoding formats and messages often sent in English. THE WORLD BANK | Digital Credit for Smallholder Farmers: Lessons Learned 22 Developing Engagement Model Options specific products (product education), apply for, access, A series of visual storyboards were developed to Equipped with a better understanding of farmer and use a given product (sales and distribution). The depict basic elements of the product according to practices, the project team then presented five models varied by the i) degree of direct engagement each model. Smallholder farmers were then asked to engagement models to the farmers. These models a farmer would have with digital technology and provide feedback so the project team could better were designed to record reactions and preferences ii) by the type of person the smallholder farmer gauge which model was perceived to be more about different ways a farmer might first hear about the would interact with to learn about, enroll, and use comfortable and convenient. service (marketing/advertising), learn more about the product (Figure 11). FIGURE 11: Rural Customer Feedback Regarding Engagement Models 1 BRAC OFFICER 2 BRAC OFFICER & 3 DIGITAL 4 MOBILE MONEY AGENT 5 SUPPLIER 8 participants preferred this MOBILE MONEY AGENT 2 participants preferred this 1 participant preferred this Everyone stated that the option. Three participants said 4 participants preferred this option. option. supplier option didn’t work or they would only be okay with the option. was not convenient for them. BRAC option (no mobile money • more efficient • continent, agent can explain the process) • more efficient for farmer and • my kids can help me with the process • I don’t use a single supplier, want BRAC process • I don’t trust the agent to have the freedom to shop anywhere • existing relationship/trust • Money on my own time, I don’t • downloading app & figuring out time to explain this well • I need more than just fertilizer or • comfort from in-person have to wait for the officer to the process would be difficult • I don’t have a relationship with seeds, I need cash for labor interactions come the agent, I don’t want him to • I don’t trust these people, they • no other parties in between • Don’t want to be seen at the know I’m borrowing money are too good at selling • more private than agent or BRAC office • What if something goes wrong, supplier • If BRAC introduces the agent to are they accountable me, this is great • Don’t like adding a middleman to the process THE WORLD BANK | Digital Credit for Smallholder Farmers: Lessons Learned 23 Following this first round of HCD research, the project • Bullet payment at the end in lieu of recurring team convened a co-creation workshop with key payments at the start – BRAC had not previously implementation partners to discuss research findings offered an agri-loan product in the country and and design the customer journey and product its repayment model for its group and SME loans prototype. It was agreed that any product prototype included a recurring bi-weekly payment over the would need to bundle the following elements: maturation of the loan. Based on farmer reactions from the HCD research round, it was evident that • “High touch” approach – Given the limited comfort these rural households had limited non-farm related levels most farmers interviewed had with using income streams. Therefore, they were likely not their phones, it would be necessary to provide a able to begin loan repayments until after they high touch approach with farmers receiving support harvested their crops. BRAC leadership agreed to from specially trained rural promoters, BRAC credit amend its standard policy to accommodate this officers, and mobile money agents. seasonal cash flow pattern. • Reduced verification procedures and expedited approval timeline – It normally takes BRAC around 1-2 weeks to make a credit decision including several site visits. With the streamlined verification process, BRAC sought to reduce the number of site visits and decrease the turnaround time from application to approval to just 3-5 days. It also dedicated a credit officer at each branch to help facilitate digital loan applications. • Mobile money optional – Recognizing that not all farmers were comfortable with mobile money, its use for disbursement and repayment would be optional. THE WORLD BANK | Digital Credit for Smallholder Farmers: Lessons Learned 24 FIGURE 12: User Journey BUILDING AWARENESS APPLICATION DISBURSEMENT REPAYMENT DIGITIZATION APPLIED ! 1 2 3 4 5 6 7 8 Farmer learns about Promoters support Farmer receives loan After approval, a credit Farmer receives Farmer travels to Farmer receives a Farmer repays the loan and receives farmer to digitally decision made by officer conducts a a disbursement collect the money repayment reminder loan via mobile money information on how to on-board, create officer; if approved, second site visit and notification (SMS + from either a mobile via a phone call. agent or at a branch. apply. an account, and officer visits, if denied, the farmer accepts loan phone call follow-up if money agent or branch fill out a digital loan officer calls them by terms and conditions. using mobile money, or at specific times. application. phone. phone call if collecting from the branch). Pilot Launch and Iteration results and key attributes from each of the three KEY FEATURES OF THE AGRI-LOAN PRODUCT The pilot was divided into three loan cycles, with Since this product targeted smallholder farmers to lending cycles. A total of 2,626 farmers were reached refinements made to each subsequent cycle based support seasonal input investments, loan amounts during the pilot. 1,441 loans were disbursed, of which on feedback from the prior cycle, instead of retaining were relatively small—between MMK 150,000 and 992 (or ~69%) were made to women. In addition to 500,000 (~US$100 to US$300). The loan duration the same processes and features throughout. These BRAC’s gender inclusive mandate, which informed was five months, with no repayments for the first three refinements did not dramatically alter the customer months, and two installments covering 50 percent of the project’s gender sensitive approach, the sizable experience or the basic product features; instead, they the principal plus interest at the end of the fourth and representation of women borrowers may also be fifth months. had more to do with adjustments to internal policies the result of male labor migration flows out of rural and processes based on how staff managed their In order to apply for a loan, farmers were required to townships.32 This trend is having the related effect meet existing BRAC qualification criteria, be between roles and responsibilities and adjusted to new digital of increasing the percentage of smallholder farming the ages of 18-65, and have had an active Telenor SIM components during the pilot. Figure 13 summarizes card for at least six months. households led by women. 32 The Myanmar Government estimates that there are 4.25 million Myanmar nationals living abroad. Regionally, drivers of migration can include higher wages in neighboring countries, conflict, and environmental migration due to natural disasters among other factors. See IOM report on Myanmar: https://www.iom.int/countries/myanmar THE WORLD BANK | Digital Credit for Smallholder Farmers: Lessons Learned 25 Because this was BRAC’s first experience with an review admissible applications. This hybrid process individualized agri-loan product, during the first was developed to overcome unreliable internet cycle the project team initially focused on ensuring a connectivity in the villages where farmers live and smooth roll-out of core, non-digital components of the frequent power outages in the township centers where offering. This included product marketing, promotion, offices were located. As a result, BRAC branch and HQ education, staffing and training as well as ensuring staff printed hard copies of the applications so they loan application and disbursement operations were could amend or complete as needed to avoid further operationally feasible and aligned with institutional delays to the review and approval process. Mobile policies and national laws and regulations. With respect and agri scores were available during the second to digital components, a digital survey tool was used cycle. However, BRAC could not always access these by field personnel to collect know your customer scores within a timeframe that aligned with the need (KYC) details and other client information during the to approve and disburse loans according to seasonal application phase. Alternative data scores (mobile farming cycles. This experience was not strong enough and agri) were not yet available and so were not to build staff confidence in these scores. experience with this trial disbursement, BRAC decided incorporated into the decision-making process. to allow all borrowers the option to use mobile money During the third and final cycle, based on the for repayment. Farmers could opt in to repaying their During the second cycle, based on experiences from experience with the software tool in the second cycle, loan installments via Ongo’s agent network. BRAC the first cycle, BRAC leadership reduced from three to BRAC equipped their branch and central staff with also pursued a dual track approach for application two the number of required in-person site visits before tablets to optimize the digitally collected application processing and approval. The first track adhered to the submitting a completed application for review. The data. Another digital component, mobile money same process employed in the second cycle, which project team’s focus centered on the deployment of a service, was also added in this cycle. Mobile money required the collection and verification of established new software tool to remotely initiate loan applications was first tested with a small group of borrowers at criteria for lending activities and supplemental criteria in the field and manage customer accounts at the the disbursement of the third cycle. Ongo (mobile specific to this agri-loan product pilot. To understand local and central offices on a digital lending platform money operator) deployed its staff to initiate bulk the effectiveness of the verification process, randomly via internet connection. BRAC then relied on a disbursements that reduced BRAC’s physical cash selected applications were given loans without credit hybrid paper and digital process to consolidate and handling requirements. Based on the successful screening. THE WORLD BANK | Digital Credit for Smallholder Farmers: Lessons Learned 26 FIGURE 13: Summary of Key Pilot Indicators – An Iterative Approach that Allowed for Refinement with Each Cycle Loan Cycle #1 NOV 2018 MAY 2019 NOV 2019 (Dry season) FIRST CYCLE SECOND CYCLE THIRD CYCLE The first cycle of loans was disbursed in November 2018, reaching a total of 785 farmers. Out of that total, MAIN FOCUS: MAIN FOCUS: MAIN FOCUS: 458 loan applications were originated, and 269 loans Loan Product Digital Platform Mobile Money & were approved. Just under two-thirds of approved loan & Enrollment & Credit Scoring System Refinement applicants (63%) were women. The repayment rate Focus on developing an Focus on developing the Focus on improving the was 100 percent. This cycle revealed several insights appropriate agricultural loan, digital platform and the credit credit scoring algorithm and prototyping the go to market scoring algorithm. Refine loan streamlining the application that were used to inform refinements made prior to the activities, enrollment process, and enrollment based on and the back-end process for second cycle in May 2019. and disbursement. findings from first cycle. BRAC. 957 884 • Need for product training refinement. Given the Farmers 785 Reached New: 730 (76%) New: 458 (52%) non-traditional nature of the repayment terms, there Returning: 227 (24%) Returning: 426 (48%) were initial challenges explaining to farmers how Loans Applied For 458 709 884 interest was calculated as it differed considerably from other loans they had taken where amounts Loans 269 (59%) 453 (64%) 811 (92%) Approved Women: 169 (63%) Women: 278 (61%) Women: 610 (75%) were uniform and distributed uniformly along the Men: 100 (37%) Men: 175 (39%) Men: 201 (25%) period of the loan, often using a simple straight line Loans 761 (94%) rate calculation. Disbursed 227 (100%) 453 (100%) Women: 573 (75%) Men: 188 (25%) • Need for improved data collection standards Repaid 707 (93%) and training for credit officers. There was a lack On Time 227 (100%) 453 (100%) Women: 441 (75%) Men: 226 (25%) of consistency in the data inputs as well as in the Disbursed via 160 (21%) quality of photos taken for digital loan applications. Women: 88 (55%) Mobile Money Improved standards and instruction were required Men: 72 (45%) to ensure that credit officers could receive proper Note: the percentages after each number (except for gender breakdowns) shows the conversion information through the system. rate between stages. THE WORLD BANK | Digital Credit for Smallholder Farmers: Lessons Learned 27 • Need for printed materials because many farmers will not or cannot decide on the spot. Having clear, printed materials that farmers, especially women farmers, could take away with them before making a decision was critical in helping them to determine on their own or with family members whether they wanted to apply for a loan.33 • Group meetings reigned supreme as the best channel for promoting and mobilizing rural communities. As identified in the HCD research phase, group meetings remained the best way to promote the product in a community, particularly when supported by local leaders. • Applying traditional vetting methods for individual • Despite encouraging topline indicators regarding • Ensuring customer privacy in a rural group setting rural applicants can be complicated and costly mobile telecom coverage and smartphone can be problematic. Helping farmers complete given the difficulty in locating residences or penetration, deploying a mobile app proved their applications at the group meeting meant that farms and the distances between the two. The much less viable. Encouraging topline indicators others could overhear their personal information. At process of conducting in-person site visits to verify regarding mobile telecom network coverage the same time, this was seen as a potential positive elements of a loan application for each individual and rural smartphone ownership suggested since it meant farmers would be less likely to farmer was complicated because some farms were opportunities to deploy a mobile app that would provide false information. Also, finding applicants’ often remote and not close to the farmer’s home, allow farmers to participate in the digitization houses for home visits was sometimes challenging requiring a significant amount of time to reach.34 process by applying for loans from their as there are no standard addresses and farmers do smartphones. However, there was a much weaker not always answer their phones. digital/mobile comprehension and usage. 33 In a country like Myanmar, where literacy is relatively high, printed materials were useful. In countries with lower levels of literacy, other methods of disseminating information, such as using audio or visuals, might be necessary to consider. 34 This was further complicated by the fact that most BRAC and L-IFT field staff are female, so they preferred to travel in pairs for safety, which reduced the number of farms that could be visited in a day. THE WORLD BANK | Digital Credit for Smallholder Farmers: Lessons Learned 28 Loan Cycle #2 • Telenor SIM requirement led some farmers to lower than expected. Relative to the farmers that (Monsoon season) mistakenly assume the loan was from Telenor were interviewed during the HCD research phase, instead of BRAC. The mobile scoring required the pilot phase attracted poorer clients.35 This The team found that the first cycle was helpful at that farmers have a Telenor SIM. This led to some resulted in smallholder farmers with lower access to generating demand for the product. Digital platform farmers mistakenly assuming that the loan was from functional smartphones and less experience using and scoring models were launched in the second Telenor. those phones than had been anticipated. cycle. During this cycle, 957 farmers were reached, 709 applied for loans, out of which 453 were approved • Given the relatively small loan sizes and no land • Digital systems have dependencies and and received loans. 187 of them were returning title requirement for applicants, technology requirements that often cannot be met or borrowers from the first cycle. As with the first cycle, access and skills of actual borrowers were much found in rural operating environments, which the repayment rate was 100 percent. The second cycle revealed some additional insights that were useful as the team prepared for the final round of iteration with UNIQUE CHALLENGES POSED BY THE MYANMAR LANGUAGE SYMBOLS & FONTS the third cycle. Those included the following: Two competing and incompatible encoding standards are currently in use Myanmar: Unicode, the international standard, and Zawgyi, only used in Myanmar. The government has tried to stimulate greater adoption of • Loan structure and three-month grace period Unicode but its efforts have not yielded broad results to date in rural area. This complicates a number of was well received, but room for improvement digitization activities. In addition, characters and symbols on a Latin-based script keyboard requires specific remained as many farmers requested a longer knowledge that is not widely disseminated in Myanmar. These issues can lead to delays or errors when repayment period. The loan grace period was digitally transcribing and sharing information, which impacts not only administrative processes (e.g. application attractive to farmers, although there was interest in validation and review) but also communications and marketing activities where text must be printed or shared having a longer repayment period to better match electronically across commonly used platforms and software (e.g. Google docs). Continued expansion of quality brand smartphones, however, increases the access to the emergence of user-friendly text conversion software the harvest cycle and a larger loan to cover all plug-ins that can be utilized by a large segment of the population, regardless of technical skill level. seasonal input investments. 35 Loan amounts from MADB are proportional to the borrower’s land size. As a pilot, we introduced a loan amount cap of ~US$350 irrespective of the size of the borrower’s land. This cap was viewed as too small for covering the inputs needs for large farms (such as 10 acres). Farmers with relatively large land who can access larger loan amounts from MADB loan did not find the loan product attractive. Conversely, farmers with small plots of less than 2 acres found this product attractive. Farmers with large land plots tended to be more well off than farmers with small land. THE WORLD BANK | Digital Credit for Smallholder Farmers: Lessons Learned 29 complicated system access and hindered reason for the difference in approval and disbursement administrative operations. The presence and numbers was the migration of mostly male farmers, reliability of basic infrastructure (i.e. power, mobile often to Thailand or Malaysia, between when they telecom signal) remains a considerable challenge applied and when disbursements were ready. As a in the pilot area. Rural promoters and BRAC result, more loans were approved than were actually credit officers encountered power outages and disbursed. In total, 160 farmers successfully participated intermittent internet connectivity either at their in all three lending cycles. branch office locations or during visits to farming communities. This complicated system access as Just over one in five farmers received disbursements well as hindered basic administrative operations. via mobile money. This was the first cycle that the option was made available. All farmers were expected • Digitization such as digital intake forms may to repay their loans via mobile money, with most doing increase certain process efficiencies and help so through an over the counter (OTC) payment made at increase collection consistency and quality, but an agent point. Like the previous cycles, the third cycle it also introduces greater operational rigidity yielded some worthwhile insights. relative to analog systems. Analog processes offer users greater flexibility and can be more • Dedicated mobile money agents may offer easily amended. Requiring certain fields in a digital Loan Cycle #3 greater service and support for farmers because form can help to increase consistency and quality (Dry season) customer engagement is built directly into the control. At the same time, if not designed with The third cycle, which launched in November 2019, model and full-time agents are incentivized proper flexibility in mind, those systems can be less focused on refining digital systems and processes, as to support customers in ways non-exclusive, user-friendly than paper-based forms.36 In addition, well as on launching the use of mobile money for loan commission-based agent are not. The mobile the inherent flexibility of analog systems used for disbursement and repayment. During this cycle, farmer money model used during this pilot was particularly managing operations in a rural environment can outreach had become much more refined. Of the 884 interesting. The project partner, Ongo, uses often be attractive to staff to revert to in areas farmers reached, all of them submitted an application, full-time staff rather than shopkeepers or other where digital infrastructure is non-existent or with 811 approvals and 761 disbursements. The primary entrepreneurs who agree to provide agent services volatile. 36 For example, by not allowing a user to easily correct data. THE WORLD BANK | Digital Credit for Smallholder Farmers: Lessons Learned 30 in addition to their core business activities. This is frequently, often to receive better promotional often the case with primarily person-to-person (P2P) deals. SIM card sharing is also common. This mobile money models.37 made it more difficult for farmers to repay their loans via mobile money agents since a matching • Customer interface with mobile money—or any phone number is required. BRAC recognized this digital service—can be kept analog such as pattern and implemented a periodic updating issuing printed cards with a QR code to farmers, of its customer’s mobile phone numbers to which removed any requirement to interface ensure minimal complications with the digital directly with technology and reduced an initial disbursement and repayment process. It also barrier to entry. Ongo provided farmers with became increasingly relevant with the outbreak of business cards with a QR code that agents could COVID-19, as the dominant communication channel quickly scan to process the disbursement or link BRAC to its customers. repayment. This method did not require farmers to have to interface with the technology at all, thereby • COVID-19 likely impacted the ability of some reducing an initial barrier to entry that may have farmers to repay on time. As it turned out, otherwise existed. this cycle overlapped with the outbreak of the COVID-19 pandemic.38 This had a noticeable • Frequent changing of SIM cards made use of economic impact on many participating farmers. mobile money for repayment more challenging, Out of 761 total loans disbursed, delayed which led BRAC to implement periodic update of repayment occurred in only 54 cases. And of these its customers’ mobile phone numbers to minimize 54 cases, only 9 loans remain outstanding as of the errors. Farmers tend to change their SIM cards drafting of this report in September 2020. 37 In addition, Ongo primary works with business customers to help digitize supply chains. As a result, their agents travel to local busi- nesses to collect cash. During those visits, they were also able to schedule times to meet farmers close to their homes and to hand deliver them cash. This type of cash delivery model, rather than one that puts the onus on farmers to travel to an agent to cash out reduces the burden on farmers to access their loans. This aspect of the model may not work in all geographies, particularly in areas where crime is more prevalent and/or where a male agent meeting a female borrower would not be considered culturally appropriate. 38 The government extended its nation-wide curfew until May 15, 2020. BRAC required that all staff, including field staff, work from home. The government also place travel restrictions on movements during the day and group meetings, which further inhibited BRAC’s ability to physically meet with customers. THE WORLD BANK | Digital Credit for Smallholder Farmers: Lessons Learned 31 What We Learned 1. Smallholder farmer demand for an High demand and strong ability to repay was among smallholder farmers who owned plots of less individual agri-loan product observed among participating smallholder than two acres.39 Further, a 100 percent repayment farmers. The demand and repayment rates also rate was recorded during the first and second Rural demand for this type of loan product exists led to a willingness by BRAC to scale up beyond lending cycles. In general, BRAC and other MFI’s in and appears strong. The product must fit the the pilot and expand the geographic availability of Myanmar have very low delinquency rates in their circumstances and activity patterns that shape this individualized agri-loan product from 2 to 55 traditional, non-agricultural operations.40 While the a rural borrower’s financial health and capacity townships targeting ~2,600 to ~20,000 farmers. precise reasons behind the high repayment rates to absorb debt. It is important to anticipate the BRAC also adapted its newly launched agri-loan of this product for stallholder farmers were not fully need for appropriate financial literacy training program to include greater repayment flexibility, captured during the pilot, two potential reasons were and a tailored communication strategy. including an option to receive and repay loan funds posited anecdotally from interactions between rural with mobile money. Demand was particularly strong borrowers and field staff. One was BRAC’s careful 39 They tend to have limited alternatives for borrowing money than farmers who own large plots. Farmers with larger plots expressed a desire to borrow larger amounts, to cover their entire production costs. 40 http://hdl.handle.net/10986/27557 THE WORLD BANK | Digital Credit for Smallholder Farmers: Lessons Learned 32 and perhaps more cautious screening process given While demand was strong, many smallholder farmers. BRAC was initially hesitant to introduce that this was a new loan product, and therefore farmers struggled to comprehend how interest such a long grace period given that it primarily lends warranted extra scrutiny. Another possible factor for this new type of loan was calculated. Based to rural groups with a two-week repayment cycle, behind high repayment rates was peer pressure on prior borrowing experience or understanding, in which repayment is linked to multiple sources from fellow borrowers in the community who did not farmers were more accustomed to a scenario of income. Despite the more flexible design of this want to jeopardize future access to this type of loan. whereby they are told how much they owe as a product, the loan disbursement was late due to single lump sum. In this pilot, communication was technical reasons with the digital loan application To accommodate this level of demand for less direct as staff attempted to explain how the process. In addition, although the length of growing financing, providers require commercially viable three-month grace period impacted the calculations periods differs between monsoon and dry seasons, models with acceptable operating costs and for the two payments that would have to come in the schedules and terms of the loans remained the lending risks. In general, microfinance loans without month four and five. Rural residents may simply fear same. The loans were therefore not well aligned to collateral have a more frequent repayment schedule, borrowing that leads to over-indebtedness or have the actual planting cycle for many farmers, which while agricultural production demands seasonal an aversion to it because it could lead to negative resulted in many farmers having to repay the loan loans with a flexible repayment scheme. Diversifying perceptions from community members. These before harvest. In some cases, they had to borrow into lending products with greater repayment dynamics indicate a need for dedicated financial money from elsewhere or sell livestock to do so. flexibility are often assumed to impose additional literacy training that is designed and delivered Farmers expressed interest in extending the loan risks among already risk-sensitive MFIs, making them for smallholder farmers so that individuals and period by another month for paddy to avoid this more reluctant to make larger loans available.41 It households are better equipped with the knowledge premature repayment due date. The generally may therefore be necessary to introduce screening and skills to manage a healthy level of finance and positive feedback from farmers and 100 percent protocols and de-risking mechanisms properly what types of products best suit their needs and repayment rate prior to COVID-19 has made BRAC tailored to the realities of operating in remote capabilities. receptive to offering more user-friendly terms and areas involving the acquisition of individual rural flexible products to farmers. The pilot demonstrated customers, particularly when a provider is launching The ability to adapt loan tenure, grace period, that viable market opportunities exist for commercial a product or seeking to expand into a new market and repayment schedule to seasonal cash flow lenders that can balance capital and operational segment. cycles of dominant crops also played a key role risks associated with deploying rural lending in strong repayment rates among smallholder operations against farmer needs for unsecured 41 https://openknowledge.worldbank.org/handle/10986/27557 THE WORLD BANK | Digital Credit for Smallholder Farmers: Lessons Learned 33 lower value, longer duration loans, in which recovery is concentrated at the end of the lending period. PRELIMINARY FINDINGS FROM IMPACT EVALUATION OF THE AGRICULTURAL MICROCREDIT PRODUCT IN MYANMAR 2. Digitizing operations from the To better understand the impact of the loan product, BRAC commissioned a study to compare the impact on farmers who received the loan to a control group who did not receive the loan. The study looks to examine smallholder farmer perspective the impact of the loan on farmer input use and income for a sample of farmers who received the loan in the Digitization is not a silver bullet. Capturing and November 2019-April 2020 season (Loan cycle #3). The sample used in the study consists of three different types properly understanding the local context is a of groups; those randomly selected to receive a loan (188), those selected to receive a loan by BRAC loan officers, vital first step in which an HCD approach has (168) and a control group of farmers who did not receive the loan (97 farmers). considerable merit. But it must also be followed In the first set of analyses, the team compared both treatment groups to the control group in two separate sets of by a diligent application of that understanding regressions. Since the control group was not selected randomly, the team used a statistical matching technique called Propensity Score Matching, which uses baseline data to improve the comparison between the treatment during the design and piloting phases. A well- and the control group. designed product or service and an effective delivery strategy for rural market segments will Evidence of the impact of the loan on aggregate outcomes such as total input expenditure and income is noisy and not statistically significant for either treatment group. However, looking at disaggregated expenditure need to consider what degree of digitization outcomes the team finds that those who were randomly selected to receive loans, spent MMK 108,000 (~US$77) is appropriate to pursue, at what pace, and more on fertilizer expenditure, while the treatment group selected by BRAC officers spent MMK 101,559 (US$73) involving which intended end users. It must more on pesticides, on average in the November 2019-April 2020 season when compared to the control group. also anticipate and address issues related to i) These impacts are against an average loan size of MMK 290,000 (~US$186). The team therefore finds limited gender, ii) building confidence and capacity in evidence that receiving the loan is correlated with higher input expenditures. a product or service and what the benefits of The team also compared outcomes between the two treatment groups, providing insights into whether BRAC usage are, iii) the role of human interaction, as officers did a “better” job of selecting farmers who would benefit from the loan, compared to those who were well as iv) trust, and v) privacy. randomly selected. “Better” in the context here would be if farmers selected by BRAC had higher levels on the key outcomes studied. The team finds that those who were selected by BRAC, spent MMK 318,866 (US$228) more on agricultural inputs and had higher levels of agriculture profits. Digital solutions need to be calibrated to the In a second set of analyses the team studied the impact of the credit score on the treatment effect of the loan realities of the rural operating environment and product. However, the team did not find clear evidence in this preliminary evidence of the impact of credit score target customer segment. Service providers will on the loan outcomes, and the results appear to be in the opposite directions for the two treatment groups. want to counter-balance topline figures and trends Further analysis is needed here to draw any conclusions. from secondary sources about the potential to THE WORLD BANK | Digital Credit for Smallholder Farmers: Lessons Learned 34 leverage digital technology by undertaking primary penetration appeared strong within the target will be costly to administer and oversee at scale market research to achieve a more granular level communities, many farmers used their smartphone (e.g. maintaining an accurate database of active of detail. A firm grasp of local market conditions passively. They were unfamiliar with how to customer SIM numbers). This gives challenges to lends itself to the ability to effectively communicate use SMS, never download apps, or upload any operations which use mobile phone number as with customers, and to increase their willingness information through their phones. Having to use their one of key customer ID, as well as mobile big data- to engage by improving their understanding of phones to initiate an application, approve terms and based scoring model, which requires behavior data how digital technology can benefit them. In the conditions, or receive and repay loans was therefore attached with a given SIM card over a certain time case of Myanmar, recent trends and developments new to most farmers, and outside of the comfort period. In addition, Mobile devices are frequently regarding mobile telecom service coverage zone of many. This suggests that in addition to used in a communal manner, with some smallholder and mobile device penetration suggested that financial literacy, digital literacy training and capacity farmers having only SIM cards and relying on digital solutions may be deployed in a farmer- building may be required when introducing digital someone else’s mobile device to make or receive led engagement model. However, the realities of financial services to smallholder farmers. It also calls. This impacts the ability of smallholder farmers unreliable power supply and mobile connectivity, the underscores the importance of undertaking primary to independently access the service. prevalence of “fake” smartphones, the challenges market research that employs an HCD approach so to accessing mobile devices, the frequent loss or that these deeper insights into customer needs are Understanding gender differences is critical to sharing of SIM cards, and poor mobile literacy each identified and properly addressed. proper product design and deployment. Local played a role in shaping the digitization process from gender dynamics will likely have implications for how a rural smallholder farmer’s perspective. In some areas, residents may own multiple SIM technology and financing is accessed differently by cards and SIM “swapping” is common, driven by women and men. Gender should also inform product Careful consideration should be given to existing promotional campaigns, and may impact the ability design and delivery. mobile usage patterns and preferences of to develop individual scoring models based on a smallholder farmers. Smallholders may exhibit single mobile network operator’s billing and data In many rural cultures and societies, women and a very narrow pattern and so lack familiarity with records system. For service providers seeking to men congregate in different places, and follow important actions they are expected to perform leverage a single SIM number as a primary means daily routines that afford males considerably to access and use the service. For example, the to authenticate a customer’s identity, this practice more mobility and freedom. Expectations about team discovered that even though smartphone may create a number of operational challenges that appropriate forms of behavior and comportment THE WORLD BANK | Digital Credit for Smallholder Farmers: Lessons Learned 35 three lending cycles. For example, BRAC’s senior leadership maintains a high percentage of female loan officers to make outreach activities with rural women more comfortable. The project team also distributed paper brochures during information meetings with prospective borrowers because they recognized that women may not be able to unilaterally decide at the meeting whether to apply for a loan. They could refer to the paper brochure later in conversations with their husband or other male family members and apply once they had time to make a decision. Building a human bridge for rural last-mile service delivery is often necessary to stimulate acceptance and usage – though it may be costly. It is critical to identify the role of key local intermediaries in the design process as they can provide smallholder also distinguish gender roles and norms, and these Products and services that fail to adequately account farmers with exposure to digitally enabled products too tend to inhibit women’s options – including what for and address these differences may, in practice, or services in a rural context. Rural last mile delivery is considered acceptable in how they engage in exclude women. An HCD process can be particularly will often require the use of individuals, micro- or financial markets. Crop production differs leading to helpful in drawing out some of these different needs small enterprises, or larger businesses to act as different farm-based cash flow patterns, and levels and expectations. During this project, the project authorized agents or representatives. If findings of technology access and digital literacy may vary team actively solicited the thoughts and preferences during the design and prototyping phases indicate considerably.42 of women at multiple stages of the design process that digitization would exceed the current readiness, as well as their experience during each of the capacity, or interest of smallholder farmers, key 42 See the following links to CGAP publications on the role of social norms and how women access and use financial services: https://www.cgap.org/blog/how-social-norms-men-restrict-womens-financial-inclusion THE WORLD BANK | Digital Credit for Smallholder Farmers: Lessons Learned 36 intermediaries could play a constructive role passed through them are properly delivered. For collection—a smallholder farmer may have several and facilitate access to the product or service example, some initial conversations with local elders reasons for such a visit. These services also open without requiring the smallholder farmer’s direct lacked specificity in terms which farmers were up new avenues for fraud that rural low-income involvement. While this may increase operating likely most appropriate for this product. This led customers are unfamiliar with and ill-equipped to costs in the short term, once adequate volumes to a number of community mobilizations that drew mitigate or avoid.44 When considering broader of customers and transactions are reached, this older, more affluent, or larger scale farmers instead issues of data privacy and protection, the use of cost structure would shift (ideally decline) making it of younger farmers with smaller plots or who rented non-financial data sets to produce alternative credit commercially viable for rural-facing service providers land. scores often requires accessing or generating to sustain. information about household activity patterns and Digitization may help alleviate mental barriers to farming practices. And while service providers may Developing trust among rural populations in borrowing within certain rural communities by be able to leverage this information to provide digital any new service, digitally enabled or otherwise, addressing a specific aspect of privacy, but it also products that are more affordable, more accessible, requires the ability to connect with customers and raises new concerns about how to treat, manage, and better tailored to the needs of smallholder communicate its value proposition. A thoughtful, and protect information about smallholder farmers, it raises important questions about how user-centric design process and a sound launch farmers that previously was not digitized. The this information should be treated. Particularly strategy are necessary, but rural customers will likely ability to remotely access formal lending services with respect to digitized agricultural information not confer their trust simply because the service affords rural borrowers a certain degree of privacy. attributable to a specific farming household or plot, exists on the market. They need to be properly A smallholder farmer may not want to be seen there is considerable regulatory and legal ambiguity introduced to the service and have sufficient time walking into an MFI or bank branch out of fear at present regarding who owns this information, using it to understand how it practically functions that others will speculate it was to borrow money who can access it and how it should be stored and and experience its benefits. In this project, local because she or he is desperate or a poor manager shared. Some governments, including the United elders provided valuable entry points into rural of household finances.43 However, when visiting an States, EU, and New Zealand, recently published communities. But these sources of rural trust also authorized agent that provides a range of digital working guidelines and proposed best practices to require a certain degree of cultivation and outreach financial services—such as money transfer, cash-in, provide structure to ongoing conversations about to ensure that whatever message or request is cash-out, bill payment and loan disbursement or this nascent and rapidly evolving topic.45 43 In some cultural norms, borrowing money is a source of shame. 44 See CGAP publication for details: https://www.cgap.org/research/publication/fraud-mobile-financial-services 45 The EU has released a Code of Conduct on Agriculture Data Sharing by Contractual Agreement, The US American Farm Bureau Federations have released a Privacy and Security Principles for Farm Data, and New Zealand has released a Farm Data Code. THE WORLD BANK | Digital Credit for Smallholder Farmers: Lessons Learned 37 3. Digitizing operations from the service provider perspective While the potential exists for cost and risk reduction, digitization should be viewed as a gradual process rather than a quick and seamless integration of new technology or systems with existing ones. It will require a broadly established and well-understood internal rationale, adjustments to organizational culture, and adequate investments of time and resources to align structures, processes, and staffing. It also requires a strategy that identifies digitization’s potential to address current pain points in the system as well as its limitations and the need to maintain certain human-based operations and interactions. Not all processes and systems are equal when doing so allowed them to outsource a less efficient success. Digitization of those elements required it comes to digitization. Digitizing selectively in a process of branch-based cash disbursements significant reforms to existing systems and familiar context in which a small number of inefficient, analog to a mobile money service provider with a rural processes as well as work internally to build staff elements are replaced may be more attractive and agent network. However, digitizing other elements capacity and comfort levels to trust and use these feasible than a wholesale digitization effort that of the envisioned product design—such as the new digital tools. Attempts to fully carry out the replaces entire systems—cutting across multiple credit score, customer account creation, and loan proposed digitization effort would have resulted processes, departments, and personnel. For application, review, and approval—progressed in further time and investment within the current example, BRAC found it relatively easy to digitize its at different speeds and to varying degrees of consortium configuration. Ultimately, BRAC elected loan disbursement and collections operations since THE WORLD BANK | Digital Credit for Smallholder Farmers: Lessons Learned 38 a more targeted approach that alleviated a near and operational implications of last-mile delivery term operational challenge—funds disbursements into smallholder farming communities. Even though STREAMLINING APPLICATION VERIFICATION PROTOCOLS CAN MAKE SERVICING and collections. This brought about gains in terms of a service offering leverages digital technology, INDIVIDUAL RURAL LOANS both efficiency and cost in its loan disbursement and the provider will likely require a rural network of MORE EFFICIENT AND LESS COSTLY collection activities. BRAC leadership also observed personnel or agent affiliates in the short to medium At the outset of the pilot, BRAC chose to retain during an exit interview that digitizing information term to support marketing and customer acquisition its standard practice of three site visits during associated with rural customer application as well as post-acquisition customer education and an individual loan application review. Since origination and onboarding contributed to staff support. These networks also play an important role the prospective borrowers are often in remote places with no bank branch nearby, these productivity increases in the range of 40-60 percent. in establishing trust with farmers. They also generate visits substantially contribute to the lender’s This effect on productivity was realized in the field digital information on smallholder farmers needed costs for facilitating these loans, reducing their and in the office among personnel tasked with to onboard new clients and validate credit models attractiveness. As the pilot progressed, BRAC different loan origination and approval duties. that do not yet exist in the service providers’ or other agreed to simplify the verification process. systems. Following the first loan cycle, the number of Digital technology reduces—but does not house visits was reduced to two—an initial visit eliminate—the need for an on-the-ground Mobile money can offer attractive benefits for by the loan promoter during loan application presence. It is important to keep in mind that digital both lenders and borrowers. While only around origination, followed by a visit by the credit officer. solutions need to be calibrated to the realities of the 20 percent of borrowers in the third cycle were This reduced verification process eliminated the usual branch manager approval. During the local, rural environment. Often, connectivity in these given the option to receive their loan via mobile second cycle, house visits were randomly done areas is at best a recurring challenge and at times money, the positive experience with even this limited for repeat borrowers, with only 70% receiving non-existent. If the service requires a stable internet number of clients helped BRAC to see the benefits house visits. This was reduced to 50% for old and connection and does not support certain offline of this channel. Not only did it reduce the amount new borrowers alike in the third cycle. features such as account origination and data entry, of staff time and cash handling risk needed to this will impact the ability of smallholder farmers physically disburse cash, it also benefited borrowers to register for or access products and services. who were able to receive their loans much faster the option of mobile money as a disbursement/ Without an existing service presence in rural areas, than the previous cash-based method. As a result, collection channel for other products as well to providers should not overlook the strategic, financial, BRAC decided to offer all customers of one branch obtain further experience. THE WORLD BANK | Digital Credit for Smallholder Farmers: Lessons Learned 39 Awareness of unique e-KYC challenges is key This is an important lesson learned. The introduction when entering new markets. In Myanmar, rural of non-financial data and alternative credit scoring addresses often do not exist. Birthdays are generally models must be done incrementally in a way that documented according to the Myanmar lunar clearly demonstrates to credit officers and senior calendar and names may not have standardized management the model’s predictive potential. The English spellings. These types of local dynamics may technological feasibility of new scoring models prove problematic when procuring an international is a necessary but not sufficient condition. This platform provider that is not set up to accommodate will require deploying the model long enough to deviations from international standards as part of its achieve a certain customer base, portfolio size, and existing software application suite or customization default ratio to better identify risk factors as well as offering. to observe performance during periods of acute economic stress or downturn (e.g. major weather Alternative data sources for credit scoring models event). complicated. If multiple people are topping up and requires further research and experimentation. using the SIM card of an applicant, it is difficult to The pilot showed that it is technically possible to Ultimately, financial institutions will only begin know what to attribute to that individual. Moreover, introduce alternative sources of digital data, such to adopt such models when they are convinced of it is impossible to know if an individual has been as mobile and agricultural data, to inform credit their accuracy and benefit. sharing their SIM card unless they self-report that. In decisions. At the same time, this is a process that countries where SIM/phone sharing is less prevalent, cannot be rushed. There is potentially significant Robust credit scoring models require strong data, such a model might be much more effective. In some risk to any financial institution that institutes a new ideally attributable to a single individual. While the contexts, where the SIM card is shared amongst model for credit decisions, particularly one that pilot’s duration was not long enough to aggressively family members and where agricultural income is has not been tried elsewhere. While BRAC had stress test the model, it did generate lessons considered household income, SIM sharing might be access to these data sources in the second and relevant to other service providers and stakeholders less of a concern as a proxy for income. However, it third cycles, the timeliness and robustness of the considering similar designs for a rural financing is still impossible for the mobile network operator to data was not always aligned to their needs.46 As a product. Given the degree of phone sharing that know who is using the phone, and whether it is only result, credit decisions were often made using more exists in rural Myanmar, using mobile usage data members of the immediate household or otherwise. traditional data and decision-making processes. as a proxy for disposable income becomes more 46 The fact that farmers were not loyal users of a single SIM card introduced an additional layer of challenge. THE WORLD BANK | Digital Credit for Smallholder Farmers: Lessons Learned 40 MAKING USE OF ALTERNATIVE INFORMATION TO BUILD CREDIT PROFILES The pilot explored incorporating two alternative, non-financial sources of data to help build credit profiles for smallholder farmers, many of whom did not have formal credit profiles: mobile phone and agricultural risk data. The mobile phone data looked at factors such as how long the farmer had a SIM, usage and top-up history, and current airtime balance over a historical period, which required at least six months of SIM activity. This was used as a proxy for traditional financial data, such as cash flow. The agricultural data relied on crop risk scores that were created for four crops: rice, maize, green gram, and black gram. The crop risk scores relied on historical weather data from satellite imagery and local weather stations. Scores ranged in value from 0 to 200. Higher scores indicate lower agroclimatic risk and deemed to have greater production potential, a higher likelihood of successful harvest, and more likely to be able to repay their loans. The main practical challenge in using alternative data for credit scoring is in first pairing the data with loan repayment data for a given target population (such as farmers) so that the relationships between alternative data points and loan repayments can be quantified. While mobile network operator data has been proven to predict repayment risk well for very small, short-term (30 day) consumption or income-smoothing (nano) loans (*), the extent to which these relationships hold for smallholder farmers planting crops with a 6 or more month growing cycle can only be known by collecting the mobile network operator data at the time of loan disbursement, lending to the farmers, waiting for loans to mature, and observing the relationships between the alternative data points and loan repayment. The less a lender screens borrowers in such a ‘knowledge building’ pilot, the more clearly the relationships between the alternative data characteristics and loan repayment can be seen. Obtaining at least 500 delinquent accounts (a notional minimum for statistical scorecard development (**)) entails a cost that most MFIs will find preclusive, but paying it could help obtain a ‘golden sample’(***) that allows it to develop a model that is scalable (through digital platforms) to clients it otherwise would never had risked lending to. In this pilot, despite 18 months and three lending cycles, there were fewer than 60 delinquent loans and no way to know with any certainty if either the mobile or agricultural data based scores were at all effective. This suggest future pilots in similar markets might be designed to increase the risk taken already in a second cycle, once the viability of the farmer loan product itself has been proven. * https://www.ifc.org/wps/wcm/connect/region__ext_content/ifc_external_corporate_site/sub-saharan+africa/resources/dfs-data-analytics ** CGAP “How to use advanced analytics to build credit-scoring models that increase access” page 19 *** David Hand “Dark Data” page 239 THE WORLD BANK | Digital Credit for Smallholder Farmers: Lessons Learned 41 Mobile credit scoring can inadvertently exclude applied, soil quality, pest and disease prevalence, decision-making, tracking, and supervision, with some potential customers, especially women. and eventual crop yield. These data points implications for how office-based personnel at Since the mobile credit score involved data from a would have strengthened the model’s predictive different management levels operate. But activities single mobile network operator, eligible customers capabilities, linked them more explicitly to a farmer’s and processes associated with loan origination, had to have active SIM cards with that provider for plot, and allowed for seasonal updates for further funds disbursement, or funds collection may not a minimum period of six months. This resulted in recalibration and improvement. The team also be impacted to the same degree. Digital systems, some farmers being excluded, as they may live in had to address the issue of how to treat the same because of their reliance on other infrastructure, an area with poor network coverage, use a SIM card score assigned to two different crops and whether lack a level of flexibility relative to human-based registered in a family member’s name, or simply to map risk according to both score and crop or operations. If systems are down, due to connectivity, prefer another provider. At the same time, partnering only according to score.47 To link crop scores to power, device, or software issues, access is with multiple mobile network operators may not quantitative risk of delinquency as well as compare restricted and performance may suffer if non- always be feasible. In markets where a single crop scores among different crops, adequate farmer- digital work-arounds do not exist. Additionally, operator is more dominant, this may be easier to level data need to be collected for farmers who digital systems are typically are efficient and overcome. Being aware of this challenge and taking receive loans and repay either on time or late. effective where adequate digitized data exists. As steps to reduce its potential impact is advisable from a result, service providers will want to consider the offset. Service providers will want to balance the whether digitization offers greater value in better potential efficiencies of digitization with the serving the needs of an existing client base as Crop risk models may require multiple data inputs existing flexibility of human-based operations. it matures and more digital data accrues versus to calibrate, and score interpretation across Potential efficiency gains from digitization may not acquiring new customers who will likely have a different crops needs to be properly understood. be rapidly realized or broadly distributed. Within much small digital information footprint. This could During the pilot, not all desired data inputs could service providers, digitization’s efficiencies are also have implications for when, how, and for what be collected. Many data sources required direct commonly associated with improved standardization, purpose customers are exposed to digitization and engagement with farmers or plot visits to collect, aggregation, processing, and management of encouraged to be active or passive participants. such as the specific breed of a given crop, the date information. This will often correspond to more of crop planting, when and how much fertilizer was streamlined, cost-effective methods for loan 47 The two most common crops grown by farmers during this pilot were rice (during monsoon season) and beans (during dry season). If the crop risk score generated for both rice and beans is equal, it was not immediately clear whether the production risk was the same or not and to what extent that should impact the lending decision. THE WORLD BANK | Digital Credit for Smallholder Farmers: Lessons Learned 42 Recommendations 3. A rural-facing offering does not need to be fully digitized when launched; DESIGN VARIOUS USERS rather, an incremental approach may be warranted that starts with digitizing “low hanging fruit” to give stakeholders 1. Employ a research and design approach 2. Where appropriate, let rural customers an opportunity to absorb and adjust to such as HCD that incorporates principles remain at the periphery of digitization the new product or service. While BRAC and methods capable of surfacing but able to interact with the product or was open to a more comprehensive digitization relevant practices, perceptions, and service through credible intermediaries. strategy, they quickly realized that easiest patterns of rural customers as well Rural digitization should not extend beyond what aspect of individual rural lending to digitize, as key needs and challenges they the local environment and observed practices and the one where they saw the most likely encounter. Digitization is not a silver bullet and of likely users will allow, especially at the outset. financial and operational benefit, had to do with service providers must assess the degree of While smallholder farmers are typically the target managing physical cash to process either loan readiness of smallholder farmers to understand, end user, they may not be able to immediately disbursements or repayments. adopt, and actively use a product or service. take up new digitally enabled products. THE WORLD BANK | Digital Credit for Smallholder Farmers: Lessons Learned 43 4. Prioritize and adequately fund an preferences regarding where, when, and how operationally feasible, and financially viable. internal “digital readiness” assessment this communication should occur. In the near to surface training needs across term, human touch points, whether in-person or 7. Explore partnerships for financial and departments and at different staffing via call center, can provide additional service digital literacy training. The need for levels. Since the responsibility for educating backstops that are necessary when targeting financial and digital literacy training for rural customers often falls on branch or field rural customers. populations will only increase as the availability personnel, it is essential to understand their and reliability of digitally enabled products capacity as well and to then design accordingly. broadens geographically. While highly market They must be able to establish a level of and context specific, partnerships among public, comprehension and confidence with the private, and development sector actors are likely SERVICE PROVIDERS selected digital solution so they can explain what the most effective way to deploy and scale such it is and how it works. Digitizing paper-based trainings. Service providers should not seek systems for use in rural areas with limited digital 6. Service providers may need to adjust to build out this capacity in-house and instead infrastructure is not a “quick fix” and instead or expand the type of information leverage networks, experience and capacity of requires longer term thinking, internal capacity collected from smallholder farmers and external partners (e.g. public or development-led building, and champions. rely more heavily on direct engagement rural extension services). to satisfying account registration 5. Do not exclude human-based operations requirements (e.g. Know Your Customer 8. Explore potential partnerships with entirely given the limitations of basic or KYC information). In anticipation of agritech companies, technology firms, rural infrastructure required to power information gaps and collection challenges, and academic institutions to obtain data digital services reliably at scale. Rural service providers should proactively assess and relevant to smallholder farmer credit customers will likely require ways to access their specify critical gaps, consult relevant laws and risk. Digital footprints of farmers captured by accounts, ask questions, make comments, or regulations—especially those related to account agritech companies have significant potential seek redress for specific issues and concerns. issuance and identity verification and validation— to immediately help lenders understand how They may expect an option to communicate with and determine what adjustments can be made smallholder farmers differ from one another a service representative directly and have strong to implement a process that is compliant, and identify those most ready to take and THE WORLD BANK | Digital Credit for Smallholder Farmers: Lessons Learned 44 able to repay a formal loan.48 The next step of turning this data into a robust statistical model with accurate quantitative estimates of the likelihood to repay a loan requires first lending to a selection of the farmers and observing repayment. In addition to credit, crop risk scores have the potential to facilitate the provision of crop and/or location-specific information to farmers. There should also be consideration given to why these data sets are being generated to avoid data collection simply to have as much information as possible on existing customers. PUBLIC SECTOR regarding how digital data and information 10. Digitization of certain public sector 9. The public sector has a role in ensuring is generated, owned, accessed, used, and services could reduce private sector risk a healthy enabling environment for rural- protected. They should also prioritize digital in agricultural investment. For instance, connectivity in a way that promotes accessibility, digitized citizenship identity could help account facing products and services, including reliability, and affordability for rural communities. registration and e-KYC. Easy to access digital digital information standards and digital Further, they may want to consider mechanisms data sets containing climate, weather, and infrastructure. National governments should for customers, and rural customers in particular, agronomic information would help analyze consider a range of initiatives—policy, regulatory, to complain and seek redress. agricultural risks. investment—that clarify rights and obligations https://www.gsma.com/mobilefordevelopment/wp-content/uploads/2020/02/Digital_credit_scoring_for_farmers_ 48 Opportunities_for_agritech_companies_in_Myanmar.pdf THE WORLD BANK | Digital Credit for Smallholder Farmers: Lessons Learned 45