The value of having a data centric approach to set up a high performing DFS agent network

Introduction

Africa is at a turning point. It has been growing at a significant rate over the past years, but the economic growth has not transformed into the same level of financial inclusion. Economic growth has resulted in more opportunities for low income households, but with this comes a need for more affordable and accessible financial services. DFS providers have been playing a vital role in servicing the market segment traditionally excluded from the formal financial sector.

Mobile money is a crucial service for the previously unbanked. It improves the drive for a cashless environment and provides communities with a platform to transact in a similar way that you would have at a bank. Setting up a network of mobile money operators, termed “agents”, plays an essential part in the success of the ecosystem that provides basic financial services.   But setting up this agent network can also be a daunting task if not done in a supervised manner.  Data analytics providers can benefit, achieving a high performing DFS agent network.  This blog post looks at three key metrics that should be considered for optimization of resources allocation and the profitability and efficacy of the agent network.

As part of embedding responsible finance practices into an organisation, and enabling financial inclusion in underbanked markets, it is important to have a high performing, motivated and well distributed agent network.  DFS providers have the responsibility to serve clients with respect, ensure that their information is securely stored, provide an infrastructure that is liquid, well trained, available during trading hours, well located and educated in the available product offerings.  Although there is a human element involved in creating the mentioned environment, good data analytics and alternative data sources assists providers to achieve agent networks that are both high performing and aligned with the principles of responsible finance.

Population density

The physical location of any part of the network is very important, and understanding where people live, where they work and how they move between the two locations can be a proxy for a good location. Traditional government statistics (usually outdated, and only captured on province/district/suburb level) and alternative data sources (like http://www.afripop.org/) can be used more successfully together. For instance, Afripop uses satellite imagery from NASA, and found there to be a direct correlation between the night lights that can be seen from space, and population density. Using these datasets, DFS providers can get localised population density in 1x1km square arrays, that provides granular, geospatial information about the potential customer base. It allows providers to make informed data driven decisions for developing agent networks based on timeous population density statistics. The below image is an example of available Afripop data with estimated localised population density for Kenya, demonstrating how population is concentrated in certain areas of the country.

[1]

 Money Corridors

A money corridor is a high value/volume route between money senders and receivers. The traditional model for sending and receiving money consisted success of public transport and relied on a network of taxis and buses. Understanding these traditional ways increases the value from the insights as to where traders, parents, farmers and others need to send money, and where to expand an agent network. In turn, it improves the optimal agent placement and liquidity management. The below image demonstrates how public data sources can be used to provide insights into international money corridors, and the same principles can be applied to intra country corridors.[2]

Liquidity

Liquidity management is probably the most difficult factor in any network. As an agent/partner needs to balance the need between having enough e-money as well as cash, the rebalancing of this can be quite tricky and nuanced. Various factors complicate this, like proximity to ATMs, banks and other financial access points that enable the replenishment of either e-money or cash. In the rural areas and typically unbanked areas rebalancing can be a big problem as formal financial institutions and banks may not have a presence. Knowing where formal and informal financial institutions are to aid in managing liquidity, is a tremendous benefit.

Proper liquidity management and financial planning allows agents to balance cash flow more smoothly. As a result, agents can process more transactions and serve customers better and more responsibly, if they ensure that they have enough cash to satisfy the demand. If not, liquidity gaps will appear, affecting agent performance and customer experience. Institutions like FSD (Financial Sector Deepening) have done some remarkable research on this, producing insightful datasets that can be used to estimate proximity to available formal and informal financial institutions. The below image shows a summarised geospatial plot of financial access points in Zambia. Using this data, DFS providers can see where there are support networks that may add in potential partnerships for rebalancing float and managing liquidity of an agent network.

[3]

Historically the set-up of a network was a daunting task, requiring significant field research and man hours. Finding the right location that serves the community, creates money corridors, and is also profitable to the agent and the DFS provider is vital. Some companies have been very successful in applying a data centric approach to agent roll-out strategies as well as to manage the performance and impact of the network. Data Analytics and Digital Financial Services Handbook[4] summarizes various DFS cases studies, providing practical insights and lessons learned.

Among them is the case of Zoona, the leading DFS provider in Zambia, offering OTC transactions through a network of dedicated agents. Zoona has a data-driven company culture and tasks a centralized team of data analysts to constantly refine the sophistication and effectiveness of its services and operations. Zoona has developed an in-house simulator to determine the optimum location for agent kiosks. The approach uses Monte Carlo simulations[5] to test millions of possible agent location scenarios to identify which configurations maximize business growth. These simulations provide scenarios that introduce complexity and uncertainty in the roll-out strategy, and can provide confidence intervals around the expected outcomes. To ensure reliability, modelled scenarios are cross-referenced with input from the field sales team, which has local knowledge of the area and the outlets under the most pressure. Together this establishes a customer centric approach to selecting the most impactful location. Zoona keeps a close eye on the agent lifecycle, and have a robust way of setting targets to agents in order to measure performance. To set targets in a way that it is both motivating and realistic, it acknowledges the fact that agent location and age can add nuance to performance expectation. Zoona analyses the agent data to project the performance expectations for agent segments, such as urban and rural, producing ‘performance over time’ cohort curves for each agent, down to the suburb level. These KPIs support a model for good, robust agent management methods. As a result, the Zoona’s network consists out of 1,300 agents across Zambia, and have a 60 days active consumer base of more than 1.5 million consumers, with a product suite consisting of domestic money transfers, a low-cost overdraft facility for agents, an e-wallet and more. Such results demonstrated DFS that act as a responsible financial service provider that serves the community and works tirelessly for financial inclusion.

Conclusion

Using organizational data and alternative data sources proves to be valuable in setting up a high performing DFS network. Although some of the alternative data sources might not be as up to date as organizational data, it still adds value by blending it with existing DFS data sets. Applying responsible finance practices to DFS providers means setting up liquid, high performing networks to improve financial inclusion. Using a combination of alternative and traditional datasets proves to be extremely effective in reducing the unknown variables during the management and expansion of networks, and the impact it has on financial inclusion is enormous.

 

Adapted from a case study and additional content presented in the Data Analytics and Digital Financial Services Handbook (June, 2017), this post was authored by Morne van der Westhuizen, IFC-Mastercard Foundation Partnership for Financial Inclusion, for the Responsible Finance Forum.

 

[1] Figure 1. Alpha version 2010 and 2015 estimates of numbers of people per grid square, with national totals adjusted to match UN population division estimates (http://esa.un.org/wpp/) and remaining unadjusted. Source: http://www.worldpop.org.uk

[2] Figure 2: Estimated Cross border remittances from Ghana during 2016. Source: http://www.worldbank.org/en/topic/migrationremittancesdiasporaissues/brief/migration-remittances-data

[3] Figure 3. Extract from the FSD Geospatial Project showcasing the geospatial plots of MFI’s and SACCO’s. Source: http://www.fsdzambia.org/updated-geospatial-map-for-financial-service-access-points-in-zambia/

[4] https://responsiblefinanceforum.org/wp-content/uploads/2017/06/IFC-Data-HandBook.pdf

[5] Monte Carlo simulations take samples from a probability distribution for each variable to produce thousands of possible outcomes. The results are analysed to get probabilities of different outcomes occurring.