The Partnership For Financial Inclusion Find The Gap Can Big Data help to increase Digital Financial Services adoption? FIELD NOTES #4 Big Data is a big topic. Rarely a day passes without news of innovative It is clear that we applications of the data we all produce through our frequent use of technology. It is also increasingly recognized that effective analysis of data are dealing with can support efforts promoting development. The Partnership for Financial a specific and Inclusion, a joint initiative of IFC and The MasterCard Foundation to expand microfinance and advance digital financial services in Sub-Saharan Africa, valuable customer is working with private financial sector clients on how to employ Big Data segment that is to promote financial inclusion. We explore how a combination of big data analytics and socio-economic research can provide a powerful tool to geographically increase adoption and usage of digital financial services (DFS). In this issue of Field Notes we share some of our findings from projects in Ghana, Uganda mobile and socially and Zambia. connected. 1 Based on the latest FinScope Tanzania data from 2013. Background IFC is committed to help create 600 million bank The goal of the Big Data projects referenced in this accounts in the developing world as part of the study was to identify which MNO customers are highly World Bank Group’s goal to bring global Universal likely to become active mobile money users, using Financial Access by 2020. A significant number quantifiable statistical predictors derived from the user of these accounts are likely to be in Sub-Saharan patterns of the current active user base. In each project, Africa, home to a majority of the 25 focus countries six months of transactional data, nearly 2 terabytes in for the initiative. IFC’s Financial Institutions Group size, was extracted, merged and examined. works with microfinance institutions, banks and mobile network operators across the continent Using mobile phone data is often touted as the panacea to support this effort, both with investment and to learn everything about consumers, but it is often advisory services. With 130 live deployments in forgotten that these data do not contain any socio- 38 markets, including mobile money pioneers economic or demographic information. To overcome such as Kenya’s M-PESA, Sub-Saharan Africa is at this limitation, IFC designed a study combining Big Data the leading edge of the digital financial services and classic surveys in order to achieve a more complete evolution in developing countries. Increasingly, customer profiling of users and non-users of financial IFC is leveraging its expanding network of digital services. The results highlight that distinctive differences financial services providers to catalyze innovative, are manifest between various user segments regarding low-cost approaches to expand financial services to gender, age, mobile call and mobile money usage. previously unbanked people. Call detail record analysis (big-data analysis) In this rapidly developing market, data analytics holds This study started with a Big Data analysis of call detail a lot of promise. It can produce necessary business records (CDR)1 covering one MNO per market, each with intelligence for service providers to fine-tune product an average of 4 million mobile subscribers. Six months development, sharpen marketing efforts, and improve of CDRs and mobile money transaction records were strategy to better reach the unbanked. extracted from the servers of the MNOs. Users were then segmented into “Voice only”, “Registered (but Most markets in Sub-Saharan Africa share some inactive) Mobile Money” and “Active2 Mobile Money” common traits. The coverage of formal financial users. The results show that these segments have services is generally low, and low-income and rural very distinctive patterns of voice calls, social network customers tend to be largely excluded. In addition, structures and geographical mobility. although mobile phone penetration is often high, Active network coverage can often be a challenge in Example Provider DFS Voice DFS Per 6 Months Registered practice. Many microfinance institutions, banks Users and mobile network operators are engaged in the Total Calls Made 734.4 995.9 1244 development and deployment of digital financial Days with Data Activity 20.38 36.4 46.33 services, although the extent and impact varies Number of Cell Towers Used 51.47 64.21 457.9 by market, often depending on local regulation. Geographic Size of Incoming 56.23 78.68 103.5 DFS providers in most markets, however, face the Network (km) challenge that although many customers have Geographic Size of Outgoing 12.12 15.65 20.01 registered for these services only a minority use them Network (km) regularly. A number of IFC’s projects are therefore Days with SMS Activity 42.34 63.64 79.99 focused on increasing usage of these digital financial Size of SMS Network (km) 10.64 20.6 29.85 services. The Big Data work undertaken under the Partnership for Financial Inclusion advances Days between Voucher 6.34 7.11 6.26 Reload providers’ understanding of existing and potential customers, and thereby helps to better drive uptake. Active DFS users, i.e. customers who use digital financial Using a company’s mobile money transactions services consistently at least once per month over six database and call detail records, we seek to answer months, make on average not only almost twice as questions such as what characterizes active mobile many phone calls than customers not using mobile money users? What drives inactivity? Is it possible money, but these calls also last significantly longer. The to identify behavior patterns among customers and same pattern holds regarding text messages: active to use that information to stimulate better uptake mobile money users send and receive the most SMS, of the service? And of particular importance, can followed by inactive customers, and then by non-users we better target potential new customers that are of DFS. Whilst active DFS users call and text their more likely to be active DFS users? friends, families and business partners more often, they 1 A call detail record (CDR) is a data record that documents the details of a telephone call or other communications transaction (e.g., text message, mobile internet data) that passes through a mobile phone or other device. The record contains various attributes of the call, such as time, duration, completion status, source number, destination number, or cell tower ID. 2 The team used the ambitious definition of 1 revenue making transaction per month during each of the six months in the dataset. This is because we wanted to find regular users of DFS. In our view, one transaction in 90 days does not really qualify as a new form of financial inclusion. At the same time, the model specifies which customers come close to the profile of these very active users. Using a less ambi- tious model would risk including customers who would actually not be likely to use DFS. also have a much larger social network with around 130 There is a strong correlation between high users of contacts compared to about 60-70 for other customers, telecoms services and the potential to be an active, and these contacts are geographically more spread regular DFS user. The research found that many out. In one market, for instance, the radius of the social telecoms-only customers had a demographic profile network is 170km larger. By contrast, non-users of DFS similar to these highly active DFS users. The team seem to move around much less than active users as therefore scored all telecoms subscribers according to evidenced by a lower number of cell towers picking up the extent to which users are similar to the profile of their phone signal. In one market, active users seem highly active DFS users, using a model of the 15 most to move around 10 times more than normal users. powerful variables which predicts whether a subscriber There may be other factors in play such as lifestyle and is likely to become a user of mobile money5. mobility differences between rural versus urban areas, but overall, DFS active users appear to leave their local Based on the findings, maps were compiled of current areas far more. and predicted distribution of mobile money users. The first map indicates the actual distribution of mobile Active DFS users are heavy users of all telecom services money users as of 2014, while the second one reflects and are thus the High Value Customers and early the predicted adaption, resulting in the last map adopters each MNO seeks to attract and retain. This highlighting districts with highest concentrations of reinforces the importance of DFS to the MNOs. In order likely adopters (see Figure 1). to attract and retain the most lucrative customers, MNOs need to provide a good quality mobile money service. ‘Classic’ Customer Profiling by interview (Socio- While more research is needed to better understand economic survey) In order to build a more complete picture of the profiles of different types of digital financial services customers, IFC designed and commissioned a socio-economic profiling study in Ghana6. Based on the average three- monthly volume of Voice/SMS/Data usage, the users were organized into three different segments (High/ Mid/Low users). A random selection of 500 subscribers from each segment were interviewed by phone. This brief summary focuses on initial findings regarding demographics, as well as mobile and mobile money usage. these customers and determine whether regular usage Demographic attributes of customers in Ghana: of mobile money increases usage of other services, it is Mobile phone users are more likely to be male (61 clear that we are dealing with a specific and valuable percent) and relatively young (45 percent of respondents customer segment that is geographically mobile and are under 35), with a good literacy level (only 14 percent socially well connected. For example, it is possible that had none or only primary school education7) and access some customers have small businesses for which they to financial services (66 percent have a bank account8). use their DFS accounts, which could explain both their high number of contacts and their mobility. The fact that Key findings from the socio-economic survey DFS users can afford higher bills for all mobile services might indicate that they are better off than average 1) Mobile phone usage MNO customers. This hypothesis is supported by the • 70 percent of the respondents use one mobile findings of the World Bank poverty scoring tool (SWIFT) phone, while 30 percent use two or even more that showed the poverty rate of one MNO’s clients in mobile phones Ghana to be lower than the national average at 18%, • More than half of the subscribers use at least two while the rate for mobile phone owners3 is 21% and the network providers (56 percent) overall poverty rate is 24.4 percent. Unfortunately, our • The most loyal clients were among the young sample was not large enough to calculate the poverty (16-24 years) and elderly (+55 years). There is a rate of active DFS users, but one might expect the relationship between loyalty and “High” activity difference between them and the overall population to usage, especially among the young. be larger. A study in Kenya showed that DFS customers • There is an apparent gender effect which persists were twice as likely to be wealthier than non-users4. across all three levels of mobile activity: 3 Ghana Living Standard Survey 6 (2012) 4 http://siteresources.worldbank.org/AFRICAEXT/Resources/258643-1271798012256/M-PESA_Kenya.pdf and http://www.mit.edu/~tavneet/M-PESA.pdf 5 The predictions have an accuracy of between 75-86 percent. 6 A similar study is planned for Zambia. 7 The national average is above 41% according to FinScope 2014. 8 Including accounts at formal, non-bank financial institutions such as MFIs. National average is 40.9% (FinScope 2014) • Only 37 percent of the male interviewees are Why are so many GSM subscribers not using DFS in Ghana? users of only one MNO, compared to 55 percent It appears that factors related to the agent network of the female users, who seem to switch less and agent quality are not a primary cause of inactivity between providers while they tend to have in Ghana with only 2.7 percent of people claiming that lower mobile activity levels on the network agent availability prevented them from using DFS. A they use. far bigger issue is that there seems to be a qualitative • For women, the combination of using mainly one problem around awareness and products: 28 percent provider while having low activity levels means of non-users declared that they don’t have a need that there could be more room for growth for for DFS, which suggests that they may need some the MNOs that develop marketing strategies that explanation how DFS can be used to assist in their appeal to women9. financial management, and also that MNOs need to • Across all age groups, the majority of low-activity consider whether they have the right products for these users use at least two providers indicating that customers. 23 percent of customers reported that they low usage on one network does not necessarily had no money to use with DFS, which reinforces the correspond with general inactivity in mobile usage. need for customer education since even with irregular incomes, many could still benefit from DFS. 2) Usage of Digital Financial Services: 95 percent of the MNO subscribers10 are aware of DFS, Comparing findings from the socio-economic even among older users above 55 years. Most respondents survey with big data analysis became aware of DFS either through TV or radio Customers may have several DFS accounts with advertisements (more than 60 percent), while 19 percent different providers, but CDRs are usually only available became aware through word of mouth. 57 percent of from one MNO, and the commercial sensitivity of the subscribers of one MNO are actually using DFS. data makes it extremely challenging to access and combine CDRs from several MNOs. Socio-economic There is a gap between awareness and usage of mobile surveys have the advantage of potentially obtaining data money services, with many potential users who are aware about the customers’ whole DFS usage behavior with of DFS but never use them. multiple providers. This is an important consideration because nearly one out of five GSM customers use • “Low” GSM activity subscribers have the largest gap DFS from multiple providers and are likely to have a between awareness and DFS usage. In addition, preferred service that they use most often. As different there is a great disparity between male and female MNOs can have different average customer profiles, it respondents (see Figure 2). “Low-active” male users is important that the socio-economic research takes have the same level of DFS usage as “high-active” this into account. female users. • The gap between awareness and usage decreases When comparing the findings from the socio- the younger the users are. economic survey with big data analysis, the team • This data suggests that younger people and men tend found differences between “Voice Only”, registered (but to be more open to experimentation, and therefore inactive) DFS users, and active DFS users. Customers more likely to be early adopters of DFS. who are infrequent users of voice calls are also more 96.8% 96.4% 94.7% 94.7% 95.4% 93.0% 68.7% 61.9% 50.3% 49.6% 44.7% 42.3% Figure 2: Mobile money services (MMS) – Awareness & usage 9 According to the GSMA, women mainly need a lot more information, support and reassurance before they trust DFS with their money. See: http://www.gsma.com/mobilefordevelopment/unlock- ing-the-potential 10 The national average is 93 percent according to a 2013 VISA study (http://usa.visa.com/download/corporate/_media/mm-ghana.pdf) likely to never have used DFS. Whilst younger people In summary, big data analysis is a new and developing are the most active users of voice, they also have the tool, offering huge potential to support financial largest share of registered yet inactive DFS accounts. inclusion by precise targeting. However, classic research This suggests that there is room for improvement of methods of reaching out and talking to customers the services offered by DFS since these mobile savvy will also remain crucial for product development and younger customers have not been engaged by the improvement of services. The research undertaken as current offerings despite using their phones regularly part of this study underlines the importance of both. for other reasons. It is an important finding that the study confirmed the loyalty and increased usage effect of DFS: customers who were active as DFS clients of one MNO were also more likely to have high usage on the same GSM network. Conclusion Big data is a powerful tool, and by enhancing it with consumer profiling research it is possible to target groups of customers with precision. In this study, big data was used to discover the profile of those MNO voice customers most likely to become regular users of digital financial services; research was then used to identify socio-economic groups that fit these profiles but were not using DFS. The research then identified geographic locations with populations having a high propensity to use DFS but that were currently underserved by MNOs. It is reasonable to expect that a combination of targeted marketing and the provision of DFS use cases of relevance to these profiles should result in significantly increased active usage of DFS. The research focused on two main sources of information. Firstly, from the mobile call detail records we learned that mobile subscribers, who use digital financial services consistently (at least once per month over several months) make for instance more and longer phone calls than subscribers that do not use DFS. Information like this generated through big data analysis helps MNOs to effectively and efficiently expand market reach and promote the use of both DFS and their telecoms business. In Ghana, the use of the findings in this study have already led to the financial inclusion of more than 70,000 additional people. At the same time, going out and undertaking direct customer surveys - the second source of information used - also has significant value for MNOs. The socio- economic research surveys showed that there is high potential for growth in the mobile money markets examined, given that nearly half of the voice subscribers have never used DFS. In particular, the youth segment and infrequent female voice users are high potential target groups who could be approached with tailored products, services and communication strategies that could lead to increased use of digital financial services and ultimately to greater financial inclusion. Authors SVEN HARTEN is the IFC Results Measurements Specialist leading the knowledge and learning agenda of the Partnership for Financial Inclusion. His current research focuses on developing successful business models for microfinance and mobile financial services in Sub-Saharan Africa. JOSHUA BLUMENSTOCK is an Assistant Professor at the Information School, an Adjunct Assistant Professor of Computer Science and Engineering, and founder and co-Director of the Data Science and Analytics Lab at the University of Washington. His research develops theory and methods for the analysis of large-scale behavioral data, with a focus on how such data can be used to better understand poverty and economic development. MUHAMMAD RAZA KHAN is a research assistant at University of Washington, experienced in both Research and Development and Commercial Software Development. Currently working on different problems related to Big Data and Computational Social Sciences like Product Adoption and Churn. JOHANNES KINZINGER is a Monitoring and Evaluation Consultant working for the knowledge and learning agenda of the Partnership for Financial Inclusion. His research focuses on characteristics of microfinance clients as well as the factors that influence the take-up and usage of financial services in Sub-Saharan Africa. Contact the Publisher: December 2014 Edition Anna Koblanck AKoblanck@ifc.org +27(0) 11-731-3000 IFC, Sub-Saharan Africa 14 Fricker Road, Illovo, Johannesburg The Partnership for Financial Inclusion aims to expand commercial microfinance and advance digital financial services to bring financial services to 5.3 million previously unbanked people in Sub-Saharan Africa by 2017. It is a $37.4 million initiative of The MasterCard Foundation and IFC that brings together the intellectual and financial capital of the Foundation with IFC’s market knowledge, expertise and client base. The partnership is also joined by The Development Bank of Austria, OeEB, and collaborates with knowledge partners such as the World Bank and CGAP. An important objective of the partnership is to contribute to the global community of practice on financial inclusion, and to share research and lessons learned. This publication is part of a series of reports published by the program. To find out more, please visit www.ifc.org/financialinclusionafrica