Our friends at MicroSave Consulting recently published a detailed, timely study on the state of digital credit in Kenya. It is chock full of great information, including lots of data, and we encourage you to dig into it in detail.

One particular theme caught our eye — the relative portfolio-level performance of the various types of digital credit providers — banks, MFIs, bank-MNO hybrids, and fintechs. We note the following:

• Non-performing loans (NPLs) for active digital credit is 27%. 16% of all digital credit given out between 2016 and 2018 are non-performing. This includes active and non-active loans, and all four types of providers. That figure is 27% for active loans only! Having worked extensively with financial service providers (FSPs) serving mass markets around the world, we know that those at the helm get quite worried when NPL inches toward 10%, and the relevant regulatory authorities start considering major interventions if it gets close to 20%. NPLs of 27%, in comparison, seems seems shockingly high.

• NPLs for active digital credit from Fintechs is 66%. That’s correct — for every digital credit given out by fintechs that is not non-performing (as opposed to “current”), there are two that are! (Hence the title of this post…) These astronomical levels of non-performance truly boggle the mind. In comparison, NPL levels at MFIs and bank-MNO hybrids are half as bad at around 30%, and banks come in at 3%.

• It’s getting worse. NPL levels were at 0% in 2016, 10% in 2017, and 29% in 2018. This trend is largely due to the fact that fintech digital credit portfolios have only started out or expanded significantly in recent years, and it takes time for NPLs to show up on the books. Nevertheless, the trend is certainly worrisome, as our experience suggests that vintage analysis would show that the “steady-state” NPL would be quite a bit north of 29%.

How could a financial institution stay in business if 66% of the loans that are expected to be repaid within a month are actually more than 90 days late? Turns out, there are some redeeming characteristics of this Wild Wild West of digital credit.

First, 81% of borrowers pay back their balance eventually, while 19% remain non-performing (slide 25). With 19 out of every 100 short-term loans having outstanding balances, one can almost start sympathizing with a 15% monthly interest rate that the likes of Tala must charge to stay in business…

Second, the balance on non-performing credit is lower than the portfolio average — about two-thirds of the 27% active NPLs have an outstanding balance of US$2 or less (slide 27). For comparison, digital credit is usually in the KES 1–500 range. This certainly dilutes the impact of the number of NPLs on the overall portfolio.

Finally, a lot of the smaller loans are taken by first-time clients who are testing out the product and don’t bother paying back if they don’t like the service (slide 27). Though it does have the effect of getting millions stuck in Credit Bureau purgatory.

While these factors mitigate some of the severity conveyed by the figures above, the fact remains that the health of the digital credit portfolios of fintechs is quite sub-par compared to those of other digital lenders.

With that in mind, we offer three thoughts.

Whither the fintech creditworthiness secret sauce? Weren’t the client’s smartphone usage patterns and associated alternative credit data supposed to provide some special juju that would allow providers to quantify uncertainty around creditworthiness for thin-file clients, and thereby provide sustainable risk-adjusted credit to them?

NPL levels an order of magnitude higher than traditional FSPs makes one wonder if the additional 10% of digital credit footprint (slide 15) is worth it. Allowing for borrowing from multiple types of digital lenders, the actual increase in footprint is probably quite smaller than 10% also.

Socializing costs at Internet speeds? With one out of three clients paying back within 90 days in the short-term, and four out of five clients paying back eventually, this is a clear instance of cross-subsidization where those who are making regular repayments on these short-long loans are making up for the non-payment by others and keeping the digital lenders afloat. A charge of up to 15% a month is not a trivial amount by any stretch of the imagination. Do the quality of life improvements associated with making immediate small-ticket credit available to the digitally connected justify the significant degree of cross-subsidization that the fintech-led digital lending business case seems to require?

And irrespective of the positive externalities, is it even fair to spread out the risk of one bad Paul on only four good Peters, when banks serving the low-income or mass-market typically have NPLs in the single digits, and rotating savings and credit associations (ROSCAs) spread out the risk of one amongst twenty to forty of his or her peers?

A pop in time saves nine? Granted, the number and size of digital loans are not large enough to be of systemic risk to the Kenyan inclusive credit market, but fintechs focused on digital lending continue to raise significant amounts of venture funds, and these hype-y things have a way of getting carried away with themselves and becoming a bona fide bubble.

The microcredit sector responded too late. The PayGo sector is going through a period of soul-searching, and will hopefully be able to “right-size” itself. Perhaps it would be wise for fintechs in particular, and all digital credit suppliers in general, to take note, pop the baby bubble that we fear is emerging, and bring discipline to digital credit operations. After all, borrowers will readily queue up to take on easy credit, and venture capital cannot but help to fuel fires in the face of potentially astronomical returns. And nothing suggests that this time will be different.

Given the increasingly digital nature of the globe, fintechs have been a major driving force behind innovation in the inclusive finance space. We hope fintechs involved with digital credit in Kenya are able to get their houses in order before there is large-scale negative impact to the consumer, the credit marketplace is significantly distorted, and regulators engage in reactive responses that mess up the sandbox for everyone.

Robbing Peter to lend to Paul was originally published in Finance for Life on Medium, where people are continuing the conversation by highlighting and responding to this story.

Machine learning (ML) provides institutions with predictive analytics that leverage large amounts of both historical and continually-generated data to obtain actionable insights. These insights, in turn, allow institutions to design and deploy interventions to increase revenue, lower costs, and compete in a dynamic marketplace by providing more customer-centric products and services.

However, few understand or have visibility into the inner workings of machine learning models so many are left wondering how good the insights actually are. For skeptics and experts alike, BFA has created the Truthalizer, a verification process that uses an out-of-sample dataset to evaluate the efficacy of a particular machine learning model as well as evaluate the intervention it triggers. As such, the Truthalizer can help leaders understand the quality of insights and how much to rely on them.

For example, the Truthalizer could help in any of the following scenarios:

• Alice is a burgeoning data scientist. She has worked through the introductory exercises offered by Kaggle, but much remains a mystery. “Which one is precision, and which one is recall,” she wonders out aloud, “and why is the Confusion Matrix itself so confusing?”
• As head of the Data Science team at his company, Bob is promoting a new model to predict customer churn. Compared to the current model, the new model has different tradeoffs between false positives and false negatives. “We have three interventions to prevent customer churn, and their costs are different, as well as how often they actually work,” his department head tells him. “Will your new model actually save us more money?”
• Carol believes that ABC Predictive Analytics Company could be game-changing for the healthcare division she oversees. Her gut, however, tells her the promised results sound too good to be true. Her Quantitative Analytics team does not have access to the proprietary algorithms that drive ABC’s predictive engine. “Are they for real, or are we being taken for a ride by the snake oil salesman of the digital age?” wonders Carol.

Each of these are variations of the question: is the machine learning model any good? At BFA, we hear this question often from our FIBR partners that use ML techniques. The question is a good one because, in our experience, the process for verifying the model is often a weak link in setting up ML, whether in-house or in partnership with an external service provider.

The Truthalizer answers this question by using an out-of-sample dataset and cost data to: 1) assess the efficacy of a particular model using standard evaluation metrics, and 2) generate return-on-investment calculations to evaluate the intervention triggered by the model.

The Truthalizer is housed at truthalizer.ai. Note that the Truthalizer does not actually run any models or undertake any computations to derive predictions; these are left to the user.

How Truthalizer can create business impact

Machine learning models can be evaluated on the frequency and costs of incorrect predictions, namely false positives and false negatives. The Truthalizer allows firms to evaluate the costs and benefits of machine learning by allowing different costs to be assigned for false positives and false negatives to accurately assess the impact of the model to the bottom line. With a better understanding of the impact on the budget, leaders can make better decisions about whether/how to deploy machine learning (ML).

Furthermore, the Truthalizer can help fledgling data science teams by introducing them to the best practices of model evaluation, and by providing a disinterested evaluation to executives and champions on the merits of a model. By combining both evaluation and return-on-investment (ROI) calculations on the same page, we hope it encourages data scientists to keep ROI considerations in mind and enables non-technical executives to understand more of the underlying technology.

We also expect the Truthalizer to act as a ward against salesmen peddling ML of the snake-oil variety. Sub-par ML implementations will provide predictions that are worse than those derived from more basic techniques, ultimately costing more than if they were not used. The tool will help users understand that sometimes bad ML is worse than none.

As such, we hope the Truthalizer will remove much of the risk associated with incorporating predictive ML models for FIBR partners and other financial service providers (FSPs).

Truthalizer in Action

The Truthalizer is designed to calculate ML evaluation metrics, as well as display and explain them. These evaluation metrics consider the rate at which the model generates false positives and false negatives. For example, many providers consider using ML to predict whether a loan will remain current or become delinquent; in this case a false positive would be a loan that will remain current but the model predicts will go delinquent and a false negative is a loan that will go delinquent but is not predicted to be so.

Consider an example:

There are 147,150 loans in the uploaded file. Of them, 4,724 are late and 142,426 are current. Continuous predictions detected. Cost of predicting incorrectly is $100, cost of preventative intervention is $10, and the effectiveness of that intervention is 60%.

1) First, the Truthalizer presents a confusion matrix, a contingency table that compares predicted results with actual ones. This table forms the basis for evaluating the model.

The confusion matrix and evaluation metrics change their results dynamically based on the threshold chosen. A 50% threshold value is the default to differentiate between current and late loans. Since the cost of a false negative is not equal to the cost of a false positive, the Truthalizer offers two thresholds for evaluation — one that maximises accuracy, and one that minimizes cost. More on this below.

In an ideal world, the model would predict current loans as current, and late loans would be predicted as late. In the real world, we hope to minimize the number of false positives and false negatives, which are indicated above in red.

The table that accompanies the confusion matrix displays and explains three evaluation metrics: accuracy, precision and recall.

In the example above, the Truthalizer explains a recall result of 60.5%: “This means that out of the 1,000 loans that are late, the model correctly identifies 605 as being late, but misclassifies the remaining 395 as current.”

2) With this evaluation of the ML model completed, the Truthalizer uses data about the unit costs of servicing loans (entered independently) to consider the cost efficiency of an intervention to prevent default. It generates a detailed worksheet that walks users through the return-on-investment for such an intervention. The calculations reflect the fact that the cost of a false negative is not equal to the cost of a false positive.

In the example above, we assume that the cost associated with a late loan is $100. This cost could include delayed interest income, loss of principal, as well as expenses associated with chasing down repayment. It also assumes that interventions exist that can prevent a loan from going late, but that they cost $10 per loan, and are effective 60% of the time.

With these data inputted, the Truthalizer calculates savings created by the intervention as follows:

Savings = Cost with no intervention — (Cost of unpredicted late loans + Cost of loans that stayed late despite intervention + Cost of intervention)

In this case, it shows a savings of $274,450.

For continuous prediction values, the tool calculates and presents two threshold values, one that maximizes accuracy and one that maximizes savings, allowing users to snap to those values (Figure 1).

Finally, the ROI evaluation includes two graphs that display how the various costs and savings change if the threshold levels for the model are changed. They allow users to explore how the impact to the bottom line changes if the false positives and false negatives have different costs associated with them.

The FIBR Impact

BFA created the Truthalizer because we could not find a user-friendly tool for our FIBR partners that allowed for approach- and model-agnostic evaluations, and explained model performance in relatively simple terms. The Truthalizer delivers on our learning agenda by allowing FIBR partners to:

1. Adjudicate onboarding ML models to ensure that ML-based prediction models perform better than some other approach, such as expert systems
2. Compare ML models to provide a consistent baseline against which to compare competing ML approaches and corresponding models, especially over time
3. Evaluate Partnerships to independently evaluate claims made by a vendor on the efficacy of their offering

Our goal is to boost confidence in predictive models that use existing and new data. By providing a reliable evaluation of the power of these models, we hope the Truthalizer encourages businesses to leverage the data they currently generate and perhaps digitize further to generate more data.

Furthermore, the Truthalizer can help convince financial service providers (FSPs) to work with FIBR partners to “offer, fund or underwrite financial services to low-income end-users through the local businesses they are associated with”. First, it will quantify uncertainty for FSPs by reliably predicting default, churn, or any other future event of interest so FSPs can price risk of engagement with FIBR partners appropriately. Second, FSPs will be better able to predict future behavior so as to offer appropriate products and services to customers, thereby allowing them to segment customers for upsell or cross-sell.

Further, Catalyst Fund’s AI Readiness Toolkit can help startups evaluate and prepare for implementing AI more broadly.

Related Readings

For more on the work BFA does around AI, please check out our Medium content, including:

• Next Steps to an Intelligent AI Strategy: The AI Readiness Toolkit
• Racing into Machine Learning: Data Readiness and the Developing World
• Machine Learning Meets Credit Scoring: How it Can Help Reduce Loan Delinquency Costs
• Machine Learning in PAYGo: What You Need to Know Before You Jump In
• Predicting Payment Behavior in PAYGo: Machine Learning Can Power Customer Retention
• Promising AI Applications to Help Shopkeepers Manage Business

FIBR stands for Financial Inclusion on Business Runways and aims to learn how to transform emerging business data about low-income individuals and link them to inclusive financial services to deepen financial inclusion and its impact. FIBR is a project of BFA in partnership with Mastercard Foundation.

Stay connected by signing up for the FIBR mailing list and joining the Inclusive Fintech Group on LinkedIn.

Keeping Machine Learning Honest! was originally published in Finance for Life on Medium, where people are continuing the conversation by highlighting and responding to this story.

“¿Cuánto dinero cree la gente que debe tener ahorrado para casos de emergencia? ¿Es ese monto distinto al que necesitan para la educación de sus hijos? ¿Está el personal en campo teniendo estas conversaciones? ¿Pueden estas ser realmente consideradas conversaciones?”

Preguntas como estas y otras similares son bastante comunes en las instituciones financieras que buscan movilizar los ahorros de largo plazo de clientes de bajos ingresos por medio de su personal en campo. En este blog, les contaremos cómo los datos de uso de una aplicación digital le permitieron a la Fundación SAJIDA, un aliado bangladesí de OPTIX, contar con una suerte de cámara oculta durante las conversaciones entre los miembros de la fundación y el personal. Por primera vez tuvieron la oportunidad de ver cómo se desenvuelve realmente la planificación financiera, al mismo tiempo que intentaban crear una cuenta de depósitos que se ajusta a las necesidades de la gente.

En octubre de 2017 introdujimos Financial Advisory Services (FAS), una aplicación digital de consultoría para servicios financieros, diseñada para que el personal en campo de SAJIDA pudiera, primero, detectar a los miembros (así es como la fundación llama a sus clientes) con mayor potencial para el ahorro a largo plazo; después, estructurar las conversaciones con ellos en torno a sus objetivos de vida y, finalmente, simular escenarios para establecer los montos y los plazos apropiados para las cuentas de ahorro de depósito a plazo. Una de las funciones de la aplicación digital FAS es la capacidad que tiene de recolectar datos detallados de uso, incluyendo cuándo y dónde fue utilizada, qué acciones fueron ejecutadas y qué tipo de información se ingresó al sistema para hacer las simulaciones. Más de veinte miembros del personal ocuparon la aplicación FAS en 2017, y ejecutaron 11.793 acciones. Tener la posibilidad de enlazar acciones consecutivas nos permite realizar una reconstrucción aproximada de las conversaciones reales.

La aplicación digital FAS estructura las conversaciones de manera adecuada, y tiene aún más que ofrecer

Al inciar una conversación, el personal en campo utiliza la mayoría de las metas de ahorro incluidas en la aplicación FAS (figura 1). Estas metas de ahorro — ocho específicas y una genérica — fueron seleccionadas a partir de una investigación cualitativa que incluyó a miembros y al personal, y parecieran representar de manera bastante precisa las metas de ahorro de los miembros. El ícono “Casos de emergencia” (Emergency Needs) es la meta de ahorro más utilizada, con un 31 % de las conversaciones. El ícono “Para hijos” ocupa el segundo lugar, con un 25 %. Al otro lado del espectro se encuentra el ícono “Ganadería”, el cual representa solo un 1 % de las conversaciones, lo que nos lleva a cuestionar su permanencia en la pantalla de metas.

La aplicación FAS ofrece herramientas de simulación, permitiéndole al personal en campo explorar dos escenarios posibles. Los clientes tienen la opción de elegir entre un monto único que gane intereses (“Obtener ganancias”, figura 2) o intentar alcanzar una meta de ahorro específica por medio de un depósito inicial y de las ganancias por intereses acumulados a lo largo del plazo de la cuenta (“Alcanzar meta juntos”, figura 2). La opción “Obtener ganancias” es aproximadamente cinco veces más popular que la opción “Alcanzar meta juntos”. No queda claro si esto se debe a que los miembros realmente prefieren la ganancia de intereses o a que entienden una opción mejor que la otra. Nótese que la alternativa “Opción por defecto” (First Pass) es la primera pantalla en aparecer cuando el personal abre la página de simulación, antes que se haya seleccionado alguna de las dos opciones.

También notamos que, con el tiempo, el personal en campo comenzó a utilizar las páginas iniciales de la aplicación para monitorear a sus grupos. La página “Lista de miembros por centro” ofrece indicadores de calidad de crédito y ahorros, y duración de membresía, información útil para el personal que busca conversar con individuos específicos sobre temas específicos. Esto contribuyó a que un 32 % de las casi 12.000 acciones registradas se llevaran a cabo en esa página. A modo de comparación, 5 % de las acciones (~600) consistieron en establecer una meta de ahorro, y otro 21 % (~2,500) en cambiar la meta de ahorro durante la conversación.

Las conversaciones revelan los montos y los plazos preferidos para cada meta

La aplicación FAS permite seleccionar cualquier tipo de monto y plazo para las nueve metas. Esto ha llevado a los miembros a explorar los rangos de montos y plazos para cualquiera de las metas, escenarios que ellos han ingresado al sistema y que nosotros podemos extraer y analizar (figura 4). Nótese que, de considerarse montos y plazos múltiples, todos serían mapeados.

Estos son algunos de los ejemplos de plazos y montos asociados a metas de ahorro que se pueden extraer de estos datos:

• Los plazos de ahorro más cortos se presentan en las categorías “Casos de emergencia” y “Migración al extranjero” con alrededor de 2,5 años, mientras que los más largos se registran en “Para hijos” y “Ganadería” con plazos de alrededor de 3,5 años.
• Los ahorros registrados en la categoría “Para hijos” presentan el monto promedio más bajo, cerca de Tk. 53.000 (USD $630), mientras que el monto más alto está asociado a “Negocios”, y alcanza cerca de Tk. 470.000 (USD $5.600).

Naturalmente, hay un rango asociado al monto y el plazo de cada una de estas metas, y las medianas normalmente son más bajas que los valores promedio. Por ejemplo, para la meta “Para hijos”, la mediana es de Tk. 30.000, mientras que el monto promedio es de Tk. 53.417. Y si bien un 5 % de las conversaciones exploró un monto de Tk. 6.000 (USD $73), un 20 % también exploró un monto de Tk. 100.000 (USD $1.220) (figura 5). Estos datos son tomados en consideración al momento de incorporar estos insights a la comunicación del producto o, incluso, al diseño de productos para casos específicos de uso.

El monitoreo de usos mantiene al personal en campo enfocado

Dado que la responsabilidad principal del personal en campo es generar y administrar prestamos , resulta útil tener la posibilidad de ver si los estímulos que tienen para monitorear los ahorros de depósito a plazo están dando frutos. Las 11.793 acciones registradas se tradujeron en 1.614 conversaciones, lo que indica que el personal le dio bastante uso a la aplicación.

Sin embargo, el nivel de uso entre el personal en campo no es uniforme. Los cinco primeros del personal son responsables del 70 % de las transacciones. Este sesgo refleja, por un lado, la rotación que hubo en las dos sucursales en las que se entregaron aparatos con la aplicación FAS, y por otro, un interés variado por la aplicación de parte del personal.

El área piloto de Manda (figura 6) muestra la actividad del personal en campo y las conversaciones que ha sostenido. Lo íconos de colores representan las ventanas de tiempo. Las ubicaciones han sido un poco manipuladas para hacerlas menos precisas. La extensión de esas ubicaciones y la diversidad de los tiempos son alentadores.

Los datos de uso también le han permitido a SAJIDA obtener información sobre lo que el personal debe reforzar. Usualmente, esta necesidad ha surgido cuando el personal no ha invertido suficiente tiempo en las conversaciones, o cuando se ha registrado alguna deficiencia en su modo de uso de la aplicación. Solo en escasas ocasiones esto se ha debido a un comportamiento errante. En una oportunidad, uno de los miembros del personal pasó cuarenta y cinco minutos durante un fin de semana apretando botones para alcanzar récords de acciones: creía que este nivel de uso de la aplicación estaba alineado a incentivos.

Impulsando las ventas cruzadas por medio de la planificación financiera

La aplicación FAS le permitió a SAJIDA lograr un avance significativo en la venta cruzada de ahorros a sus clientes de crédito, un objetivo clave al momento de sumarse al programa OPTIX. Al promover una planificación financiera a medida, la aplicación FAS está diseñada para contribuir de manera positiva a la salud financiera de sus miembros, enfocándose particularmente en brindar la posibilidad de planificar el futuro en términos financieros y de prevenir shocks financieros. Los agentes de campo estiman que la tasa actual de conversación es de 20 % aproximadamente. Contar con una herramienta CRM le permitiría a SAJIDA llevar un mejor registro de las tasas de conversión de conversaciones que involucran la aplicación FAS.

Los siguientes pasos

Una vez concluida la implementación piloto de la aplicación FAS, SAJIDA decidió fusionar la aplicación con su aplicación Microenterprise Loan Origination (Generación de Crédito para Microempresas), entregándoles a los agentes de crédito una sola herramienta para manejar las relaciones de préstamo y ahorros del cliente. CEP, el aliado de OPTIX en Vietnam, lanzó su propia versión de la aplicación FAS con metas de ahorro creadas a medida. Mientras estos aliados sigan buscando la mejor herramienta para ayudar a sus clientes a satisfacer sus necesidades y lograr un estado financiero saludable, seguiremos manteniéndolos informados a través de blogs, webinars y otras publicaciones en la medida en que estás historias se vayan desarrollando.

Ashirul Amin, PhD, es un consultor asociado senior en BFA.

BFA y MetLife Foundation crearon OPTIX para apoyar a las instituciones financieras en la construcción de mejores carteras de productos para clientes de bajos ingresos. A partir de 2016, OPTIX se ha asociado con cuatro instituciones financieras en México, Colombia, Vietnam y Bangladesh para entender mejor los beneficios de las estrategias de venta cruzada para las empresas y sus clientes.

Una aplicación digital, una conversación y una nueva cuenta: un novedoso enfoque desde el cual… was originally published in Finance for Life on Medium, where people are continuing the conversation by highlighting and responding to this story.

Reverse engineering the loan officer-client conversation provides key insights

“How much money do people typically imagine they should have when saving for emergency needs? Is that different from what they need for the children’s education? What kinds of conversations are field staff having?”

These and similar questions are quite familiar to financial institutions trying to mobilize long-term savings from low-income clients through their field staff. In this blog post, we share the story of how mobile application usage data allowed SAJIDA Foundation, an OPTIX partner from Bangladesh to gain insights into member-staff conversations without actually being there: a kind of virtual “fly on the wall”. They gained an unprecedented look into how financial planning really unfolds as they sought to tailor a term-deposit product to people’s needs.

In October 2017, we introduced the Financial Advisory Services (FAS) mobile app designed to allow field officers at SAJIDA Foundation to identify members (as SAJIDA calls its clients) with longer-term savings potential, structure conversations around life goals, and simulate scenarios to arrive at appropriate amounts and terms for the term-deposit savings account.

Built into the FAS app is the capability to collect detailed usage data, including when and where the app was used, what actions were undertaken, and what information was entered to simulate scenarios. Any interaction on the device, such as the click of a button, filling in a text field, or submitting a form, is recorded. More than 20 officers used the FAS in 2017 and undertook 11,793 actions on the app. By threading together consecutive actions, we can approximately reconstruct actual conversations.

FAS app frames conversations as intended…and then some

When starting a conversation, field officers mention most of the savings goals the FAS offers (Figure 1). Users were able to choose from a list of eight specific goals and one “other”. These goals are chosen based on qualitative research with members and staff, and appear to be a fairly accurate representation of member savings goals. “Emergency Needs” is the most common, featuring in 31 percent of the conversations. The icon “For Children” comes second, at 25 percent. “Livestock” features in only 1 percent of the conversations, leading us to question its permanence on the goal screen.

The FAS app offers two simulation tools, allowing field officers to work through two possible scenarios around how much to save now, and what amount to withdraw at the end of the term. Members can choose to put aside a lump sum that earns interest, and the app tells them how much they can expect to withdraw at the end of the term (“Earn Income” in Figure 2). Or, clients can enter a future desired withdrawal amount — a target — and the app tells them the value of the deposit needed to get to that target in the desired period of time (“Reach Target Together” in Figure 2). The “Earn Income” option is about five times more popular than the “Reach Target Together” option. Whether this is a function of actual preference for earning interest or a result of understanding one option better than the other is not clear. Note that “First Pass” captures the first time the field officer lands on the simulation page, when neither option has been chosen.

We also noticed that, over time, field officers started using the initial pages of the FAS app for general monitoring of their groups. The “Member List by Center” page provides indicators for quality of loan, savings and length of membership, which is useful information for field officers looking to speak to specific individuals about their specific issues. This page constituted about one third of all the 12,000 actions. For comparison, 5 percent of the actions (~600) were around setting a savings goal, and another 21 percent (~2,500) around changing the savings goals during the conversation.

Conversations Reveal Preferred Amounts and Terms for Goals

The FAS app allows amount and term to be chosen for any of the nine goals. This has resulted in the members discussing a range of amounts and terms for any given goal, scenarios which they entered into the system and that we are able to extract and analyze (Figure 4). Note that if multiple amounts and terms were considered, all of them were mapped.

Examples of terms and amounts associated with goals we can see from this data include:

• Saving for “Emergency Needs” and “Migration Abroad” involve the shortest time frames of around 2.5 years, while “For Children” and to “Livestock” have the longest time frames of around 3.5 years.
• Saving “For Children” involves the smallest amount, around Tk. 53,000 (US$630), while the largest amounts are associated with “Business”, at around Tk. 470,000 (US$5,600).

Median amounts varied by goal. For example, for the goal “For Children,” the median amount is Tk. 30,000 and the mean is Tk. 53,000. And the range is wide: while 5 percent of conversations explored an amount of Tk. 6,000 (US$73), 20 percent explored an amount of Tk. 100,000 (US$1,220) (Figure 5).

Monitoring Interactions Keeps Field Officers on Task

Given that a field officer’s primary responsibility is originating and servicing credit, it helps to be able to see whether the regular encouragements to attend to term-deposit savings is yielding fruit. “The 11,793 actions translated to 1,614 conversations undertaken by about 20 field officers in 2017,” a figure that suggests field officers used the app a fair bit.

Usage is not uniform across all field officers, though. The top five officers were responsible for 70 percent of the transactions. This skew partly reflects turnover in the two branches where the 10 devices were provided with the FAS installed, as well as varied interest in the FAS on the part of the staff.

Field officers do seem to move about in their area of operations and have these conversations in the field. The spread of those locations and diversity of times are encouraging as it shows that staff made the effort to go to members’ houses at times other than when center meetings happen, allowing for more personal conversations.

Usage data has also been useful to retrain staff when needed. Usually, this has been because they were not spending enough time on conversations or there was some other shortcoming with their mode of engagement. On rare occasions, it has been to address the errant behavior. In one case, one staff member spent 45 minutes on a weekend punching away at buttons on the screen to rack up action records because he thought that this type of engagement was incentivized.

Furthering Cross-Sell Through Financial Planning

The FAS app allowed SAJIDA to progress significantly in cross-selling savings to their credit clients — a key goal in joining the OPTIX program. By supporting customized financial planning, the FAS is designed to contribute positively to the financial health of its members, particularly with its focus on providing the ability to financially plan for the future and to stave off financial shocks. Field officers estimate that the current rate of conversation is about 20 percent. A customer relationship management (CRM) tool would allow SAJIDA to better track conversion rates from the conversations involving the FAS app.

Next Steps

As the pilot implementation of the FAS app concluded, SAJIDA decided to merge the app into its Microenterprise Loan Origination app, providing the loan officer with a single tool to manage the borrowing and savings relationship of the client. CEP, the OPTIX partner in Vietnam, has rolled out its own version of the FAS with customized in savings goals. As these partners continue to seek the best tool to help clients meet their financial health needs, we will keep you posted through blogs, webinars and other publications as these stories unfold.

BFA and MetLife Foundation created OPTIX to support financial institutions in building better portfolios of products for their low-income customers. Starting in 2016, OPTIX has partnered with four financial institutions in Mexico, Colombia, Vietnam and Bangladesh to understand and advise on the benefits of cross-sell strategies for the businesses and their clients.

Ashirul Amin is a senior associate at BFA.

Originally posted on CFI

An App, a Conversation and a New Account: A Novel Approach to Talking About and Tailoring Term… was originally published in Finance for Life on Medium, where people are continuing the conversation by highlighting and responding to this story.

[This post has been updated with a postscript on July 27, 2018.]

In the pursuit of solutions to persistent poverty and inequality in our world, we often charge ahead with well-meaning programs without ever stopping to step back, see if the problem has changed in nature, or whether our actions still represent the best solutions. For example, “everyone” knows that handing out bank accounts should improve financial well-being, right? But does it, given that a vast majority of accounts opened are either inactive or have very low balances, or both?

Well, we can now actually start to reality-check some of those assumptions. Gallup’s new Global Financial Health Study of 15,000 individuals worldwide tells us that:

• Those who use formal financial services are much more likely to report greater levels of financial security and control.
• Some demographic characteristics, particularly employment, are strongly correlated with financial security and control, but gender, not so much.
• Employment and youth showed interaction effects while gender and marital status do not, meaning that there are more-than-additive effects when the employed and youth use formal financial instruments.
• There are strong country effects, encouraging us to continue to get to know low-income customers — whether we visit by matatus, tuk tuks or rickshaws — and learn more about their financial lives.

The MetLife Foundation-backed initiative set out to understand worldwide perceptions of financial control and security, which in turn are key components of financial health. Between January and March this year, Gallup conducted its survey across 10 countries — Bangladesh, Chile, Colombia, Greece, Japan, Kenya, South Korea, the US, the UK and Vietnam. Metlife Foundation and Gallup have generously made the data available and have invited the curious to explore.

This was not an invitation one could resist; what follows are highlights of findings from a more detailed treatment of the data here.

Three Reasons Why This Dataset is Great

I find datasets like this quite exciting. The primary reason is because it allows us to explore causal mechanisms that are taken as self-evident truths in public but that, in private, elicit some amount of soul-searching from practitioners. While these findings should surprise no one working in the field of financial inclusion, what this survey allows us to do is to get a sense of the magnitude of the correlation between, on the one hand, demographics, the extent of financial inclusion, relevant interaction effects and country effects, on the other, financial control and security measures.

The second reason is the additional levels of clarity this kind of analysis provides on top of what the survey firm reports. In Gallup’s summary document, they make the case that there is no clear relationship between average account ownership figures at the country level on one hand, and the reported degree of financial control and security (also at the country level) on the other. The analysis presented here adds color to that conclusion by noting that, while this finding is true, accounting for country-level diversity allows us to also see all kinds of patterns underneath where there is evidence of a clear connection between usage of formal financial services, and financial control and security.

The third reason is that it gives us cause to pause and ask if, and to what extent, it is time to move along the microcredit-to-microfinance-to-financial-inclusion progression that represents an evolution of approach around improving financial services and its impact for low-income individuals and communities. Financial health, as a measure of resilience to shocks and the ability to pursue one’s financial dreams, has come to be seen as a worthy successor to inclusion, for example. No doubt, many financial institutions could do a better job at tailoring products to needs so that so many accounts do not remain fallow. Some will even point out that the considering inclusion as a holy grail or panacea is the problem, and that it excludes other semi-formal or informal instruments.

Nevertheless, having had the opportunity to have gone through how individuals have used the tens of millions of accounts from about two dozen financial institutions over the years as part our work, I continue to be awed by the infinite creativity people bring to bear to even the most boring, vanilla bank account. While we should absolutely continue to pursue deliberate interventions to improve the financial wellbeing of low-income communities, this survey reminds us that even in the absence of that kind of effort, many are still able to achieve higher levels of financial control and health when simply presented with the tools to do so. In other words, if you build it, many will come.

Now, let’s dive into the details.

Demographic and Financial Inclusion Effects

There are certain expectations we have in terms of who would do better financially. We would expect older adults to be more in financial control, and those with bank accounts to feel greater financial security, right? Let’s look at how reported financial security and control varies by demographics, and extent of financial inclusion (Table 1).

This table, as well as the ones that follow, converts the odds ratios obtained from logistic regressions and presents them as percentages that allow us to compare statistically significant, different responses across several categories within the same variable. Only covariates that are significant (at least 0.05 level) are shown.

As an example of what this table tells us, let’s look at the gender effect in detail. If they lost their income and had to rely solely on savings or things they could sell, women reported being 21 percent less likely to be able to cover expenses associated with basic needs for a month, and 18 percent less likely to do so if the period was a week. Women did not report owing money to a financial institution or person at a statistically significant different levels than men, nor did they find difficulty in servicing that debt. Women were also 28 percent more likely to regret the decision to spend money on something they didn’t need.

No statistically significant difference between men and women was seen in the other nine out of 10 financial control measures, such as satisfaction with the level of input into a household’s financial decisions, ability to pay for an emergency, ability to save in the past, and level of enjoyment in financial planning. Thus, there was some gender effect, in that women report negative results for three out of the fifteen measures, compared to men.

Employment had the most number of discernible effects. Compared to the unemployed, those who were employed were 33 percent more likely to survive a month on savings alone, 66 percent more likely to owe money to a financial institution, 94 percent more likely to be satisfied with their level of input in household decisions, 88 percent more likely to feel that they can pay back the money they owe, 44 percent more likely to be able to pay for an emergency, 51 percent more likely to enjoy financial planning, and 22 percent less likely to feel that their future will not change for the better.

Note that this included those who have a job or who are self-employed and those who report being unemployed. Because gainful employment is seen as one of the key drivers of financial wellbeing and upward economic mobility, being able to disentangle the impact of different types and different degrees of employment would be informative.

Sometimes, the absence of evidence is also instructive. For example, there was no difference between urban and rural respondents except for a 37 percent greater likelihood for rural respondents to be able to last for a week on savings. If the reported levels of financial security and control are similar, does this suggest that the more informal financial instruments prevalent in rural areas are enough to support similar level of financial well-being, or is that “just” the perception and not reality?

We leave it to the reader to explore the other demographic effects for now, with the note that there is a lot more to unpack here.

Now, let’s look at the financial inclusion metrics, which are the rightmost three columns in Table 1. There was a strong correlation between the use of formal financial services (and the confidence in them) and reported levels of financial control and security. Positive effects included 80 percent greater likelihood to survive for at least a month on savings, 77 percent greater likelihood to repay all debts, 70 percent greater likelihood to be able to pay for an emergency and 56 percent higher likelihood of enjoying financial planning. Negative effects included 27 percent greater likelihood to regret a decision to spend money on things they don’t need and 31 percent greater likelihood of having tried to save in the past, but failing.

This is where we must make the “correlation does not equal causation” disclaimer. While one could make the case that greater levels of financial inclusion contributes to greater levels of financial control and security, there is an existing recognition of inclusion being a pseudo-necessary but not-nearly-sufficient requisite for financial health.

Some might also argue that people are tapping into the formal financial system because they have sufficient funds to spare; after all, not having enough money was listed as one of the key reasons for not opening a bank account by two-thirds of Findex respondents.

Interaction Effects Between Demographic and Financial Inclusion

Interaction effects are of great interest because they help us determine if two or more covariates operate together in such a way that, together, they have some effect that is above and beyond what their individual effects are. For example, we saw earlier how employment and usage of formal services is associated with higher levels of financial control and security. Exploring interaction effects allows us to ask if the effect of someone who is employed and uses formal financial services is more than the effect of just being employed, irrespective of whether they use formal financial services or not, plus just using formal financial services, irrespective of whether they are employed or not.

Turns out, there is a strong interaction effect between employment and financial inclusion (Table 2). Those who were employed and had used mobile financial services were 41 percent more likely to survive for a month on savings, compared to those who were not employed or who did not use mobile financial services, or both. This 41 percent is over and above the 33 percent greater likelihood from employment alone, suggesting a rather strong employment-based interaction effect.

We also saw widespread interaction effects around youth, defined as those who are less than 30 years of age, and rural residents. There was little interaction effect around gender and marital status.

Country Effects

If there was any doubt in anyone’s mind, the country effects from the Gallup date point out the obvious; countries are different. The logistic regression runs in Stata assigned Vietnam as the reference, which means that all values were relative to the responses from Vietnamese respondents. Even though the absolute numbers were relative to Vietnam, we could still explore other country pairings by comparing their relative effect levels.

Here is a selection of some interesting country level effects:;

• Bangladesh, Greece and Vietnam were at similar levels when considering a month’s worth of expenses, while Vietnam had the lowest levels when looking at one week. (Note that lack of percentages suggests effects statistically indistinguishable from Vietnam.)
• The Japanese had the lowest likelihood of owing money to both an institution and an individual. Greece also had quite low levels of reported indebtedness, but were twenty times more likely to find it difficult to make debt payments, compared to Vietnam. The US had the highest levels of indebtedness amongst the 10 countries.
• Kenyans were least likely to be satisfied with the level of input to household financial decisions, to be able to pay for a financial emergency today, and to have people who can help financially, while Bangladeshis were most likely to expect a positive change in their financial futures.
• Overall, the advanced economies of Japan, South Korea, the UK and the US have similar profiles for financial control and security measures.

While the Gallup data does not have enough country level metrics to explore what might explain some of these rather large variations, there are many datasets out there which one could stitch to attempt to identify some national and sub-nationally relevant indicators.

Overall, this MetLife Foundation-funded Gallup survey illustrates the correlation between financial control and security on one hand, and financial inclusion on the other, and how it varies depending on demographic characteristics. This post is just scratching the surface of the wealth of information that is contained in this dataset though, and we plan on taking deeper dives into particular variables and particular countries to explore specific themes of interest. Watch this space for more updates in the near future!

Special thanks to Damian Milverton and Rasima Swarup for their help in pulling this piece together.

Postscript

July 27, 2018

We would like to clarify that we have used a different definition of the “financial security,” “financial control” and “financial inclusion” variables compared to how Gallup has defined them in their results. These differences are noted below; for details, please refer to the Gallup Global Financial Health Survey webpage. Any definitions attributed to Gallup below are quoted from the document titled Gallup_Global_Financial_Health_Study_Methodology_Report.pdf.

Financial Security

How Gallup Defines It

Financial Security can take one of three categories: “Financially Insecure,” “Financially Stretched” and “Financially Secure” based on five questions. The three levels are obtained thus:

• Financially Insecure: Respondents who said they had less than a month of savings or assets OR said their debt made it “very difficult” to pay for other things.
• Financially Stretched: Neither Insecure nor Secure.
• Financially Secure: Respondents who said they had more than six months of savings and, if they have debt, paying back that debt is “not at all difficult”.

How BFA used this data in our analysis

We took each of the five questions separately, looked at them as variables of interest and labelled each of them as financial security variables. We did not use the three-level Financial Security variable or create another composite variable.

Financial Control

How Gallup Defines It

Respondents are marked as having a high level of financial control if they respond positively to 8–10 of the questions, and marked as not having a high level of financial control if they respond to fewer than 8 questions positively.

How BFA used this data in our analysis

BFA took the results of each of the ten questions separately, looked at them separately and labelled each of them as financial control variables. BFA did not use the same notion of high levels of financial control as defined by Gallup or create another composite variable.

Financial Inclusion

How Gallup Defines It

Gallup uses the same definition used in the World Bank’s Global Findex for financial inclusion and includes it as a secondary data indicator.

How BFA used this variable in our analysis

BFA used three questions from the survey that were indicative of inclusion (Q23, Q25 and Q19).

Gallup Survey Shows Financial Security and Control Rises with Financial Inclusion, But Not Equally… was originally published in Finance for Life on Medium, where people are continuing the conversation by highlighting and responding to this story.

Field officers at Sajida Foundation, an NGO and microfinance provider in Bangladesh, are constantly on the go. They each have over 300 customers to manage: They are responsible for disbursing loan payments, ensuring timely repayment, and in addition, they make up to half a dozen personal visits each afternoon. These officers also act as financial advisors to many of their clients, who they get to know well over time. But the reality of their workload doesn’t allow them to spend more time coaching clients, which is an activity many enjoy and that clients benefit from.

Field officers understand their clients have rich financial lives, with needs, aspirations and goals for the short and long term. Some clients hold sizable savings balances as they seek to save for these goals. Officers know their clients could earn more by converting those savings to a term deposit account. Yet, in many cases, advice-giving is complex: It is difficult for officers to answer questions around tiered interest rates and to identify the right customer profile to offer differentiated products.

AN OPPORTUNITY FOR EFFICIENCY: THE FAS APP

As an OPTIX partner, the SAJIDA Foundation in Bangladesh worked with BFA to find a way for field officers to put their knowledge and intuition about clients to better use and to increase their efficiency. The solution is an Android-based Financial Advisory Services (FAS) app for field officers that identifies “super-savers” and provides officers with useful simulation tools. The FAS app was piloted in two branches and is now being deployed to other branches.

SAJIDA Foundation’s decision to develop the FAS app grew out of two realizations:

• Identifying “super savers”: “Super savers” are individuals who save diligently while servicing a microcredit loan, thereby accumulating a large amount in regular savings without making any withdrawals. Yet information on “super savers” is not readily available to field officers during field visits, which focus on loan servicing and information collection.
• Product simulations: Target-based narratives are a good start to discussing term deposits, but field officers had a difficult time answering questions about expected income based on tiered interest rates, particularly if members were aiming for a specific amount or comparing returns with competing offerings from other institutions.

FAS App in Action: Identifying Super Savers

The app’s landing page gives field officers a summary of clients’ loans and savings balances, as well as delinquent loans. Like the SAJIDA Foundation’s collection sheets, data is pulled directly from the SAJIDA databases in English, while the labels are in Bangla, making the interface intuitive for loan officers.

Figure 1. Portfolio Summary (Displays: Outstanding loan balance, balances of three savings accounts, PAR rate, and number of clients).

Figure 2. Center Summary (Displays: Member-wise list with indicators for quality of loan and savings, and length of membership).

The Center Summary, like a dashboard in the app, displays a list of all the members in that Center, along with the status of any current loans and savings. (Fig 2). The app categorizes clients into 1, 2, or 3 stars.

• Three stars: Members who have more than Tk. 10,000 (USD 125) in their voluntary savings account, and have not withdrawn for more than a year. These “super savers” are the primary targets for term accounts.
• Two stars: Those members who have between Tk. 5,000 (USD 62.50) and Tk. 10,000 (USD 125) in voluntary savings, and have held that balance for more than a year. They are secondary targets, who may be engaged in a conversation about long-term goals and term deposit accounts.
• One star: Those members who have between Tk. 4,000 (USD 50) and Tk. 4,999 (USD 62.50), and have not withdrawn in more than a year. These clients are clearly savers, but have not yet achieved the minimum balance for term deposits. Field officers coach them in considering their long-term goals and encourage them to save the minimum level.

Figure 3 Target Savings Page

The Center Summary also notes the status of all outstanding loans in the group. Loans have three states: good, overdue (i.e. somewhat behind on payments), and bad (i.e. recovery efforts have failed, and will be written off soon).

The Summary allows field officers to quickly identify members with three stars and “good” loans, to efficiently zero in on those who may be interested in flipping from savings to term accounts.

From the Summary page, clicking on any member takes the field officer to a page with details on: current and previous loans; balances of compulsory, voluntary, and term deposit savings; and the option to see more details such as choosing a new savings goal to simulate.

These client histories help officers as well as members, who appreciate being able to see their entire portfolio at a glance. In addition, the stars seemed to have a motivating effect for clients, an unexpected but welcome side-effect.

“Wow, Ma, you have three stars!” crowed Nilufer Begum’s daughter, as she and her mother pored over a handheld screen showing her loan and savings records with SAJIDA Foundation. Next to them, the field officer, Arif, explained that three stars were reserved for the best of SAJIDA Foundation’s savers. Using the 3-star system, he advises her on transferring some of her savings to a term deposit account to earn more interest.

SAVINGS SIMULATIONS

The FAS app offers two simulation tools: (1) one based on more general income generation and (2) a target-based one. These allow officers to work with their clients and show them possible future scenarios, telling them either how much they would save after a certain amount of time, or how long it would take them to save up for something specific.

The Term Deposit Income Calculator:

Figure 4. Simulation: Income Generation (Displays: Maturity amount and income received, given initial deposit and term).

This tool allows field officers to show someone who has a certain amount of idle cash how much they could earn with a term deposit account. Before the app, officers used a grid to show returns based on a few initial deposit amounts and terms, but the app allows them to use a specific amount that is relevant to the client, to produce potential income figures in less than a minute.

The typical term deposit value solution offers a higher rate of interest income than current accounts, with increasingly higher rates for longer committed terms. Thus, someone saving Tk. 10,000 (USD 125) for two years will receive 7 percent per year, resulting in an income of Tk. 1,400 (USD 17.50), and a total withdrawable amount of Tk. 11,400 (USD 142.50).

The Target-Based Savings Calculator:

During discussions with clients, they asked questions like, “If I needed Tk. X in three years, how much do I have to save now?”

Target-oriented savings pitches are more tangible for users and Sajida used focus groups to develop nine goal categories: education, home improvement, children, marriage, business, migration, livestock, emergencies, and other. When considering target-based savings, the officer inputs how much the client would like to withdraw at the end of the period, and the simulator calculates how much has to be deposited today to generate that amount.

For example, to withdraw TK. 10,000 (USD 125) in two years, one would have to deposit Tk. 8,772 today (USD 109.65), which would earn Tk. 1,228 (USD 15.35) in interest over the two years, allowing the member to meet her target.

Figure 5. Simulation: Target-based Saving (Displays: Initial deposit required and income received, given target maturity amount and term).

REFLECTIONS ON CROSS-SELL AND NEXT STEPS

The FAS App was designed to provide field officers with client information in a user-friendly manner to prioritize outreach, and facilitate conversations around various saving and income-generation scenarios. The three-star rating system for savers allows field officers to tailor their advisory sessions, and members seem to appreciate achieving the highest tier with three stars. The eight savings goals represent a simple but powerful distillation of on-the-ground qualitative work, UI (user interface) design and UX (user experience) feedback. The dual simulation modes of Income Generation and Target-based Saving cater to the two predominant lenses through which members view term deposits.

SAJIDA Foundation accomplished what it set out to do: give its staff the right tools to ease their jobs, to have better conversations with clients and a way to more effectively serve the needs of existing clients with tailored term deposits. Field officers are now empowered to have knowledgeable conversations as trusted financial advisors. The FAS app is a great example of cross-sell done well and the right balance between technology and human touch.

NEXT STEPS

The SAJIDA Foundation and BFA will refine the simulation narratives based on the occurrence of target-based term deposits vs. income generation ones. We will also determine the ROI of the FAS app by looking at the number of new term deposit accounts and increased efficiency among field officers. Stay tuned as we keep iterating and give you more updates!

(This post was originally published on NextBillion)

BFA with the support of MetLife Foundation (note: both are NextBillion partners) and fiscal sponsor Rockefeller Philanthropy Advisors created OPTIX to support financial institutions in building better portfolios of products for their low-income customers. Starting in 2016, OPTIX has partnered with four financial institutions in Mexico, Colombia, Vietnam and Bangladesh to understand and advise on the benefits of cross-sell strategies for the businesses and their clients.

‘Cross-sell Done Well’: How One Finance App Found a Balance Between Digital and Human Touch was originally published in Finance for Life on Medium, where people are continuing the conversation by highlighting and responding to this story.

In the summer of 2017, SAJIDA Foundation, an NGO and microfinance provider in Bangladesh, made a bold decision — it was going to make an entire branch of about 1,600 clients (or members, as SAJIDA calls them) cashless. Rather than attending group meetings to make microloan repayments in person, these members would show up at a mobile money (MM) agent and make their payments through a bill-pay option at a time and place convenient for them. Going cashless meant that in one fell swoop, digital financial services (DFS) would make group meetings, the locus of traditional microcredit for decades, redundant.

What follows is the story of redefining group meetings in a world of increasingly digitized microfinance, largely based on a renewed appreciation of the non-pecuniary roles these meetings perform. This is the first of a series of pieces in which BFA will track SAJIDA’s journey over the coming months as it digitizes its operations and goes cashless.

CASHLESS TRANSACTIONS ARE THE FUTURE …

In February 2017, SAJIDA started introducing digital loan disbursement and collection of installments at 15 branches through bKash, one of the major providers of MM in Bangladesh. This effort was driven by the reality that fewer and fewer members were coming to group meetings regularly because they were too busy. In 2016, we received this feedback from many SAJIDA field officers and confirmed this through a survey in which 87 percent of respondents reported that less than half the group attended meetings regularly. The same survey also found that only 7.7 percent of members looked forward to group meetings to talk to others, and only 3.9 percent socialized at these meetings, calling into question the widely assumed role of group meetings in generating social capital. The social role of groups seemed to have eroded. Simultaneously, MM had become ubiquitous as a convenient and time-saving payments solution. The time seemed ripe to try replacing cash handling at group meetings with MM-based transactions.

Soon after rolling out MM in early 2017, SAJIDA encountered several challenges. First, despite many improvements, members who were non-habitual MM users continued to point out pain points around opening accounts and needing to cash-in before making installment payments. Second, changes to MM regulations drastically lowered the balance and transactional caps of mobile wallets, making it impossible to disburse loans through wallets. Uptake of MM among its members was modest but not stellar, and it was evident to SAJIDA that more work was needed to smooth out the customer experience if MM was to work for the institution.

SAJIDA’s solution came through a partnership with Rocket, the digital financial services offering from Dutch-Bangla Bank Limited (DBBL), one of the major private banks in Bangladesh. DBBL had vastly increased its agent footprint and undertaken an extensive marketing campaign, and most importantly, it offered a bill-pay option. Unlike with bKash, SAJIDA members could show up at a Rocket agent and make installment payments without having to open an account.

While some have noted the downsides to such over-the-counter transactions, in this instance, the partnership with Rocket made a tangible difference in reducing friction in the MM customer experience. SAJIDA management also believed the partnership could deliver efficiency gains in the form of automated bookkeeping, reduced cash handling costs, and more time for loan officers to recruit and service loan recipients.

Based on these observations, SAJIDA decided to convert an entire branch to Rocket-based MM in June 2017. They bet that customers would find the MM experience to be so convenient, that after getting over the minimal initial learning curve, anyone would want to pay through Rocket rather than cash. They identified one of their older branches, Abdullahbag, as the pilot location for going cashless. It had an abundance of agents in close proximity to centers, and had other branches nearby should some members refuse to make the switch.

In many ways, SAJIDA’s bet seems to have paid off: In the eight months since going cashless, roughly 1,400 members have transacted digitally to the tune of US$100,000 per month. While there has been a slight reduction in the number of members, portfolio metrics (number of loans, outstanding amount, delinquency, etc.) have been indistinguishable from the typical month-to-month variations natural to microfinance.

More importantly, members love the ability to pay whenever and wherever — no more rushing to the group meetings for 30 minutes while food is cooking on the stove. There is also an element of pride in transacting with a bank (i.e., DBBL). Loan officers and accountants enjoy working with MM because they no longer have to endure the risk of carrying cash around. In addition, they no longer have to spend hours a day counting cash, so they can do other productive things. As an institution, SAJIDA values MM because digitization made their operations more efficient; overhead expenses associated with cash-handling, bookkeeping, monitoring and auditing can now be drastically reduced, and collection data is available in real time.

We at BFA are also pleased with the outcomes so far because, on top of all of these benefits, going cashless furthered one of the objectives of the OPTIX project — optimizing the way financial institutions offer products and services to their members.

… BUT IT’LL STILL TAKE A GROUP TO MAKE IT THERE

Nevertheless, there have been some downsides to the move. Group meetings (without joint liability) have been a key feature of traditional microfinance in Bangladesh, and from the start, both SAJIDA and BFA shared a certain uneasiness about doing away with them. While we were aware of academic and anecdotal evidence regarding potential negative impacts of eliminating group meetings, we were willing to do away with these weekly meetings in the spirit of experimentation, since their raison d’etre as a cash collection point no longer existed.

SAJIDA has monitored the impact of going cashless and group-less on its portfolio quality and staff-client dynamics to date. The feedback from the field officers (FOs) through conversation and surveys is as follows:

• FOs feel they are losing touch with members because of the lack of face-to-face contact.
• FOs find it harder to recruit members since one of the key mechanisms for recruiting new members was existing members bringing interested individuals to group meetings.
• It has become harder to solicit character references from group members, particularly long-standing ones, for larger loans.
• It is time-consuming for FOs to distribute account statements house-to-house, in lieu of writing in passbooks at meetings.
• The opportunity to inform members of changes in policy or new services from SAJIDA, and to share other important information has been drastically reduced.
• Attendance at group meetings (or lack thereof) and reports from other members would often serve as an early warning sign of a member intending to leave the area without settling their obligations — this channel of communication is gone.
• Meetings enabled a new FO to meet all group members at once, but now new FOs need individual introductions instead.

This increasing disconnect between FOs and members — and amongst members themselves — has been of great concern, and it has validated some of the initial apprehensions of SAJIDA’s leadership. Yet given the benefits noted earlier, especially for members who absolutely love the flexibility of being able to pay close to their home at any time of the day, it is clear that it makes no sense to turn back the clock and go back to cash.

So SAJIDA now faces a new question: Is there a way to preserve the utility of groups, without losing the efficiency gains from going cashless?

THE RISE OF CASHLESS GROUPS

After much soul-searching, internal discussions and weighing of all the evidence, SAJIDA has decided to reinstate group meetings on a monthly basis, to carry on its previous roles. But the following adjustments are being made to the cashless microcredit program:

• Group meetings will be reinstated once a month, instead of once a week, at the central location in the community, to preserve the human touch points that FOs feel are so critical to making microcredit work. During meetings, FOs will review SAJIDA’s products and services, and conduct various due diligence. And members can share their experiences with each other, particularly new members.
• Members will attend these meetings once a month, and receive their account statement directly from the FO. FOs will show up at the center location every week at a given time, just as before. These sessions will not be for organized meetings with the entire group, but will serve more as office hours for the FO. As a result:
• Members will know when and where to find the FO if they need him/her.
• FOs will be available to take care of all admission and loan disbursement-related paperwork.
• Any member who misses the monthly meeting will be able to collect their account statement during one of the subsequent three FO sessions.
• The location where center meetings are held will continue to be the focal point for recruiting members, and for introducing a new FO to existing members.
• Anchoring FOs to regular, weekly “office hour” sessions also makes sense from an organizational point of view.
• Preparing documentation and answering members’ questions are expected to take just 15–30 minutes, leaving plenty of time for outreach in the neighborhood.
• Because centers are spread out, these sessions will ensure the FO covers the entire area they are assigned, and not limit themselves to easy-to-reach areas.

Four additional branches went cashless on Dec. 1 of last year, with this altered group-based approach. The original Abdullahbag branch will also start phasing in this approach after sufficient outreach with members has been undertaken to convince them of the need to change their mode of interaction with SAJIDA again.

THE KEY TAKEAWAY

SAJIDA’s experience demonstrates the value of stress-testing to expose the weaknesses of an approach. This testing forced SAJIDA to innovate and find a space for cashless groups to provide the social interactions that make their group-based microfinance model work — and it did so before members were significantly inconvenienced, or significant damage happened to the branch’s portfolio.

This experience also provides another data point in the debate on the implications of digitizing microfinance. We offer this as the key takeaway: The value proposition for going cashless is clear for both financial institutions and their members, but there are risks to assuming that groups are disposable once their primary purpose of cash collection no longer exists. In situations like SAJIDA’s, the optimum combination involves both cashless transactions and cashless groups working in concert, complementing each other.

Originally posted on Next Billion

BFA and MetLife Foundation created OPTIX to support financial institutions in building better portfolios of products for their low-income customers. Starting in 2016, OPTIX has partnered with four financial institutions in Mexico, Colombia, Vietnam and Bangladesh to understand and advise on the benefits of cross-sell strategies for the businesses and their clients.

Is Mobile Money Killing Off the Group Microfinance Model (And Would That Be Such a Bad Thing)? was originally published in Finance for Life on Medium, where people are continuing the conversation by highlighting and responding to this story.

In a previous analysis, we laid out the machine learning process we conducted for Destacame.cl, a company of the Catalyst Fund, which provides alternative credit scoring to underbanked consumers in Latin America. For a firm like Destacame.cl, the investment in predictive algorithms for credit scoring is a no-brainer. It is an inherent part of their digital product and data-driven business model, and can be seamlessly incorporated. But what about for a brick-and-mortar financial institution that is just beginning to explore digital or more sophisticated technology applications? How should the financial service provider and its leadership weigh the potential benefits against the initial investment in high-tech credit scoring?

Let’s look at this question from the perspective of a mid-size financial institution that provides low-income clients with non-collateralized loans. We’ll demonstrate that credit scoring optimizes three business layers that improve the overall bottom line for financial institutions.

BUT FIRST, HOW SHOULD LENDERS THINK ABOUT THE COST OF DELINQUENCY?

Before we can start modeling the impact of credit scoring, we start first with understanding the financial institution and the workings of its business. A business case typically “bakes in” all manner of revenues and expenses associated with a particular service that a financial institution provides. These inputs and outputs are distributed over various units, such as accounts and transactions, that can be combined with the various incomes and expenses to arrive at the profitability associated with the service’s activities. In the following case, we focus on the elements directly affected by credit scoring.

There are three primary sources of financial “costs” associated with delinquent loans:

1. Write-offs from bad loans. After a certain period of time and effort chasing down delinquent loans, the financial institution must make the call and recognize that an outstanding loan amount is irrecoverable. The result is loss of capital and loss of income from that capital.
2. Delayed interest income. When a client misses a payment, expected cash flow from both principal and interest is delayed. While some financial institutions have the ability to charge interest for each day of delinquency, those with less sophisticated portfolio management systems can only expect repayment of the original amount. Interest is temporarily forgone as trapped principal is not actively earning income.
3. Delinquency servicing expenses. Compared to maintaining a current loan, a delinquent loan requires significant amounts of time and effort. To recover the loan, field officers and staff from the financial institution often must make multiple visits to the client — and sometimes even travel to other parts of the country to track down borrowers who have literally run away with the money. Any legal proceedings add even more expenses.

Financial institutions must negotiate these tradeoffs in managing the costs. Some lenders are persistent and try to work with delinquent clients for up to two years. They might eventually recover loans others would have written off, but they incur significant expenses in the recovery process. Less persistent lenders run the risk of signaling to potential clients that it’s “easy” to get away with non-repayment.

Below is a simulation of a portion of the business model that BFA built to analyze the impact of credit scoring during the Destacame.cl engagement. The figures are representative of a mid-size financial institution that provides microloans, with a not insignificant delinquency problem.

WHAT IS THE IMPACT OF DELINQUENCY?

Once we model out the delinquency and write-off specifications noted above, we find the impact on the business case as follows:

For this simulation, the base case demonstrates that the impact of delinquency is of the same magnitude (from 1.4 to 3 percent) for all three “cost” levers. Losses from write-offs are greater than from lost interest income. Interest repayments are front-loaded in P+I installments at the start of the loan life cycle and lenders really are recovering mostly principal at the end of the cycle. Loans also tend to become delinquent later in their life. The prospect of funds lost through write-offs should be considered as significant as funds lost through foregone income. This phenomenon of loss of principal is often underappreciated in practice, as focus is often on the income forgone from delinquency and the costs to make those loans current. Finally, delinquency servicing costs take away a full 3 percent from the bottom line of the portfolio and not just the delinquent loans.

HOW DOES CREDIT SCORING IMPACT THE BUSINESS CASE OF A CREDIT PRODUCT?

To quantify creditworthiness, the financial institution’s credit scoring process looks at previous credit history, income stability and related financial behaviors, among other factors. It uses techniques that range from logistic regressions to random forests, and each new or existing client is ultimately assigned a probability of default. This score is typically combined with on-the-ground due diligence — and only then does the financial institution decide whether to extend credit.

Intuitively, we see how credit scoring helps the financial institution reduce all three “costs” by bringing down its overall level of delinquent loans. Credit scoring should identify a better pool of loan customers at the start of the loan cycle, resulting in better or less risky business for the lender.

THE LOAN PORTFOLIO WITH DEFAULT AND CREDIT SCORING

Now, we expand the base case with a credit scoring algorithm from the model. We assume a true positive rate (TPP), or sensitivity, of 50 percent. That is, 50 percent of loans that eventually become delinquent are identified as delinquent. This number falls in the range of around 35 to 65 percent for TPPs of credit scoring models deployed.

The impact of proper credit scoring crosses all three “cost” levers since it has the effect of depressing delinquency levels overall and results in reduction of all costs associated with delinquency, as well as a reduction in the number of loans that become delinquent.

CREDIT SCORING — SPECIFIC IMPACT ON ‘COSTS’

As expected, most expenses halve themselves in the credit scoring model. Given the parameters, this also has the impact of turning the net present value (NPV) positive, suggesting that the discounted cash flows from the product have turned positive — a clear and attractive incentive for proper credit scoring.

Note that the reduction is not exactly 50 percent across the board as the month-to-month impact cascades, principal payments are back-loaded and delinquency requires repeated contacts.

NEXT STEPS FOR ORGANIZATIONS DEPLOYING CREDIT SCORING ALGORITHMS

By targeting delinquency-prone loans at acquisition and eliminating a significant proportion of them, credit scoring helps an organization address all three “cost” levers associated with bad loans — deferred income, recovery costs and write-offs. However, we often see discussions of the impact of machine learning credit scoring tools on the first, sometimes on the second, and seldom on the third. Financial service providers would do well to keep all three in mind, as the impact is often in similar orders of magnitude.

Thus, lending organizations can extend this exercise in two ways:

• Apply a predictive algorithm on loans to identify those that will eventually be written off. It is sometimes the case that clients whose loans are written off have a different profile than those who simply fall behind for some of the time. Surgically targeting these loans will allow write-off mitigation on top of what comes proportionately through delinquency mitigation.
• Reset the inflection point where delinquent loans are earmarked to be written off. This inflection point is typically when institutions decide it is no longer “worth it” to chase a delinquent loan. As noted earlier, there are two major considerations to this — how much it costs to chase a loan, and what signal it sends to borrowers if recovery efforts are relaxed. By being able to generally reduce but also exactly quantify the cost of recovery versus the cost of write-off, as well as the potential impact of signaling, organizations can decide to double down or ease off on recovery, because the overall reduced delinquency gives it the breathing room to do so.

(Reposted from NextBillion)

Catalyst Fund accelerates promising inclusive fintech startups and spurs innovation in financial inclusion in emerging markets with the support of theGates Foundation, JPMorgan Chase & Co and the collaboration of Accion Venture Lab, Omidyar Network, Gray Ghost Ventures, Quona Capital and Rockefeller Philanthropy Associates.

Stay connected by signing up for the Catalyst Fund mailing list and joining theInclusive Fintech Group on LinkedIn.

If you liked this post, please recommend by clicking ❤.

Machine Learning Meets Credit Scoring: How it Can Help Reduce Loan Delinquency Costs was originally published in Finance for Life on Medium, where people are continuing the conversation by highlighting and responding to this story.

Afsana works at a garment factory in Dhaka and has two young daughters. She recently opened an Astha term deposit account with SAJIDA Foundation for 10,000 Bangladeshi Taka ($120) after learning about its benefits from a loan officer. For Afsana, Astha has allowed her to lock away a lump sum which is difficult to do on her limited salary and pressing needs. She explains, “My children are studying. I want to invest money for their higher education. I would also like to save up to build a house in the village one day.”

Astha, a term deposit account, allows SAJIDA Foundation members to save toward a purchase or financial goal. Members can deposit small amounts for a specified length of time while earning at-market interest. Astha, meaning “reliance and dependability” in Bengali, is a product of SAJIDA Foundation, an NGO which offers a microfinance program and various health and other development initiatives throughout Bangladesh, including long-term savings options for unbanked or underbanked individuals like Afsana.

Uptake of the Astha savings product was good, but through the OPTIX project, BFA conducted data analytics and client research revealing that most of the accounts had lower balances than anticipated, with a median of around BDT 2,000 ($24).

By redesigning and rolling out new features to increase adoption of the Astha accounts, SAJIDA Foundation saw a 97 percent increase in accounts opened and 59 percent more mobilization of savings, comparing the Astha savings profile from July to November 2016 to the same period the year before.

The Astha term deposit illustrates how financial institutions can optimize cross-sell to better serve the needs of their low-income customers. By focusing on clients’ needs and improving the existing suite of product and services, cross-sell enables an institution to further its impact by offering more relevant services to underserved customers. The different cross-sell strategies explored through SAJIDA Foundation serve to help other OPTIX partners and financial institutions to learn from this work. Since 2016, OPTIX has also partnered with three other financial institutions in Mexico, Colombia and Vietnam to advance cross-sell and germinate these strategies around the world.

THE REDESIGN: NEW FLOOR AND FLEXIBILITY IN ASTHA

With the guidance of BFA’s business and customer insights experts, SAJIDA Foundation made the following changes to the design of the term deposit:

1. Minimum balance set to BDT 5,000: Costing this product, we realized that the break-even balance was closer to BDT 10,000 ($120). Existing users were saving less than half that amount. As a compromise, we chose the minimum balance of BDT 5,000 ($60).
2. Multiple term lengths allowed: The previous product had a single term of six and a half years based on the assumption that members would find the notion of “doubling their money” enticing. The offer worked well as a selling point, but when a pressing need arose, the prospect of future versus present value was not enticing enough to prevent liquidation. Now, eight terms between six months and five years are offered to accommodate the varying needs for flexibility.
3. Tiered interest rates introduced: Interest rates were recalibrated into tiers based on investment period, market rates and business case impact.
4. Conversations oriented to goals: Field officers engaged in structured conversations around a specific goal that the member would save toward, instead of a generic income generation sales pitch.
5. No account openings within one week of loan disbursement: Originally, many Astha accounts were opened when members visited the office during loan disbursements, as a matter of convenience. The simultaneous offering resulted in members opening term deposit accounts with low starting balances (i.e., by reassignment of some funds from the disbursed loans or from funds out of pocket). By separating the two transactions, the act of saving toward a goal by opening a term deposit account became more deliberate.

FROM EVIDENCED-BASED REDESIGN TO IMPACT

1. Increase in number of Astha accounts opened: Five hundred and eighty-three accounts with a balance of at least BDT 5,000 ($60) were opened in the five-month period in 2015, while 1,148 were opened during the corresponding period in 2016, resulting in a 97 percent increase. Staff offered a couple of reasons for this increase.
2. By focusing on quality instead of quantity (i.e. fewer larger-balance accounts), staff could dedicate more time educating members on the utility of Astha accounts.
3. Members with more than BDT 10,000 ($120) in their Voluntary Savings Account and that had not drawn funds over one year were prompted to consider transferring that balance to an Astha account. This resulted in more interest income for the member and helped SAJIDA Foundation in asset-liability matching due to Astha’s longer term.
4. A distribution shift toward a minimum balance of BDT 5,000 ($60). With a new account balance minimum of BDT 5,000 ($60) from no minimum previously, the average balance of the total number of accounts went up, as expected (Figure 1). However, if we considered accounts with balances of minimum BDT 5,000 ($60), the average balance went down (also Figure 1). Data also showed that more accounts were opened at the minimum balance levels in 2016 (66 percent) compared to 2015 (46 percent). Why? Although no minimum balance was usually indicated to customers, field officers would arbitrarily mention BDT 5,000 ($60) when describing the Astha account. It is very possible that this tendency “anchored” the staff, and thus, the member’s mind to save the minimum balance and depressed the average balance for the Astha portfolio. SAJIDA Foundation responded by instead encouraging field officers to use BDT 10,000 ($120) or BDT 20,000 ($240) when giving examples in discussions with customers.

1. Increase in total savings mobilized through Astha accounts. Total balances across all Astha accounts increased from BDT 12.9 million ($161,000) in 2015 to BDT 17.4 million ($218,000) in 2016. When only accounts with BDT 5,000 ($60) or more are considered, the 2015 balance levels were BDT 11.0 million ($137,000), resulting in a 59 percent increase of savings mobilized.

The total amount mobilized increased for all months except for November and doubled in value in July. There was a slight dip from 3.0 million in 2016 to 2.9 million in 2017 (Figure 2). We need to keep a close eye on data after November to see if this decrease is an anomaly or the beginning of a slowdown in savings mobilization in Astha accounts.

REFLECTIONS ON CROSS-SELL AND NEXT STEPS

The redesigns resulted in a discernible positive impact on account openings and savings mobilized. However, the shift in the average balance of accounts with minimums of BDT 5,000 ($60) was less obvious. The decrease is potentially the result of another possible cause — savers that typically maintain balances below BDT 5,000 ($60) are now saving at the threshold. If this is the case, then the outcome would be a positive one for SAJIDA Foundation. By monitoring the data after the staff started using different language on account balance minimums, it will reveal whether anchoring was the real cause of the shift.

In the spirit of peer-learning and cross-collaboration across OPTIX institutions, SAJIDA Foundation’s experiences will be very valuable to another project partner, the Capital Aid Fund for Employment of the Poor (CEP), in Vietnam, when deploying their own term deposit product. CEP will not only be able to leverage what we learned at SAJIDA Foundation but also benefit from lessons around packaging and marketing a long-term savings product, goal-oriented narratives for field officers to promote term deposits and best practices for appropriate analytics to track product performance.

BFA and MetLife Foundation created OPTIX to support financial institutions in building better portfolios of products for their low-income customers. Starting in 2016, OPTIX has partnered with four financial institutions in Mexico, Colombia, Vietnam and Bangladesh to understand and advise on the benefits of cross-sell strategies for the businesses and their clients.

Ashirul Amin is a senior associate at BFA.

Originally posted on Next Billion

How SAJIDA Foundation Practically Doubled Customer Uptake of Savings Accounts Using Evidence-Based… was originally published in Finance for Life on Medium, where people are continuing the conversation by highlighting and responding to this story.