February 4, 2022

Gojo x Humo Webinar

On 20th January 2022, we held a webinar to share how Humo, our newest partner company, has established Humo Lab to spearhead digitalization and innovation. At Humo Lab, several products have been developed and launched, including mobile banking apps, payment apps, kiosks, and more. The webinar focused on Humo’s introduction, sharing best practices on how microfinance institutions can do better to extend financial inclusion. 

The one-hour event featured Taejun Shin (Founder & CEO, Gojo & Company), Firdavs Mayunusov (Co-founder and General Director, MDO Humo), Firdavs Nuriddinzoda (Director of HumoLab, MDO Humo), moderated by Arnaud Ventura (Managing Partner, Gojo & Company). The webinar consisted of four parts: Introduction of Gojo & Company, Introduction of Humo, Introduction of Humo Lab, and Panel Discussion.

In the first part, Taejun shared what financial inclusion is, what Gojo strives to achieve and how Gojo operates in many parts of the world with its newest footprint being Tajikistan. It was followed by a talk from Firdavs M. covering the history and evolution of Humo, the story of how Humo Lab was established, and how the collaboration between tech and business teams allows disruptive thinking. Firdavs N. continued the presentation to introduce user-friendly services of Humo Lab and strategies behind their provision. The last 20 minutes was a panel discussion answering questions from audience, including topics such as synergies between Gojo and Humo. 

The webinar was a great success, attended by over 150 people from 22 countries. The webinar was followed by an Ask Me Anything session where the audiences enjoyed direct interaction with the panelists.  

To watch or rewatch the webinar, the recording is available here:

https://www.youtube.com/watch?v=qOG8y8JIIH0

If you are interested in learning more about Gojo and receiving invitations to our future webinars, subscribe to our newsletter at the bottom of this page.

February 4, 2022

Gojo x Humo Webinar

On 20th January 2022, we held a webinar to share how Humo, our newest partner company, has established Humo Lab to spearhead digitalization and innovation. At Humo Lab, several products have been developed and launched, including mobile banking apps, payment apps, kiosks, and more. The webinar focused on Humo’s introduction, sharing best practices on how microfinance institutions can do better to extend financial inclusion. 

The one-hour event featured Taejun Shin (Founder & CEO, Gojo & Company), Firdavs Mayunusov (Co-founder and General Director, MDO Humo), Firdavs Nuriddinzoda (Director of HumoLab, MDO Humo), moderated by Arnaud Ventura (Managing Partner, Gojo & Company). The webinar consisted of four parts: Introduction of Gojo & Company, Introduction of Humo, Introduction of Humo Lab, and Panel Discussion.

In the first part, Taejun shared what financial inclusion is, what Gojo strives to achieve and how Gojo operates in many parts of the world with its newest footprint being Tajikistan. It was followed by a talk from Firdavs M. covering the history and evolution of Humo, the story of how Humo Lab was established, and how the collaboration between tech and business teams allows disruptive thinking. Firdavs N. continued the presentation to introduce user-friendly services of Humo Lab and strategies behind their provision. The last 20 minutes was a panel discussion answering questions from audience, including topics such as synergies between Gojo and Humo. 

The webinar was a great success, attended by over 150 people from 22 countries. The webinar was followed by an Ask Me Anything session where the audiences enjoyed direct interaction with the panelists.  

To watch or rewatch the webinar, the recording is available here:

https://www.youtube.com/watch?v=qOG8y8JIIH0

If you are interested in learning more about Gojo and receiving invitations to our future webinars, subscribe to our newsletter at the bottom of this page.

October 4, 2021

A Little Theorem About Credit Scoring

In order to speed up loan approval processes, many banks and microfinance institutions use computer algorithms to calculate credit scores. I'm sure many of you are familiar with seeing a numerical credit score like '700'.

Credit scoring collects various pieces of information about a loan applicant and maps them to a single number. Based on the applicant’s loan repayment history, their use of other financial services, and other factors, the algorithm calculates a score such as '670 for this applicant', '740 for that applicant', and so on. And if the score is lower than the predefined threshold, the loan will not be approved.

Here, I would like to introduce a little theorem that makes up a part of Gojo's credit scoring approach.

Whether or not you can repay a loan without issues depends to a large extent on the size of the loan. For example, if I take a loan of $1 million, and spend it all at once without thinking, I will probably have a problem repaying it later. However, if I take a loan of $1, and use it to pay for some expenses, I will probably not have a problem repaying it later. In other words, credit scoring, which measures my ability to repay the loan, should be a function of the loan size.

Now, let's extend this view just a little bit more. The larger the loan size, the higher the credit score that should be required for the borrower, and therefore the higher the hurdle. The smaller the loan size, the lower the credit score that should be required for the borrower, and the lower the hurdle.

If that is the case, then for any given borrower, there should be a loan size that represents a manageable hurdle. For any kind of borrower, if we keep reducing the loan size, we will eventually find an amount that matches the maximum credit score they can achieve.

Maxima’s MBela team with an MBela agent at her house after a community gathering. / Koh Terai

The above paragraphs outline a little theorem about credit scoring, which is my favorite. Of course, we can take a further step to consider how we might apply it in practice.

Why don't we just look for the (maximum) loan size we believe a borrower can handle, and use that as their credit score? Rather than producing a score like '670' or '740', it would be easier for everyone to understand that $500 is the maximum possible amount they would be allowed to borrow. If the amount is clear, borrowers can plan their investment.

 We are actually applying this approach in a loan product currently being offered by Maxima, our partner in Cambodia. Their small digital loans project (also known as MBela) uses an automated assessment process to provide a credit score in the form of the maximum amount each person can borrow. The resulting credit score is easy for both the agent and borrower to understand.

Gojo wants to provide services that are innovative in their simplicity.


Yoshinari Noguchi is a researcher at Gojo and Company. He works in Gojo’s R&D team, and is currently looking into new ways to understand and support money management for low-income people, as well as analysis of data from the Hrishipara diaries and Gojo’s own financial diary projects.

October 4, 2021

A Little Theorem About Credit Scoring

In order to speed up loan approval processes, many banks and microfinance institutions use computer algorithms to calculate credit scores. I'm sure many of you are familiar with seeing a numerical credit score like '700'.

Credit scoring collects various pieces of information about a loan applicant and maps them to a single number. Based on the applicant’s loan repayment history, their use of other financial services, and other factors, the algorithm calculates a score such as '670 for this applicant', '740 for that applicant', and so on. And if the score is lower than the predefined threshold, the loan will not be approved.

Here, I would like to introduce a little theorem that makes up a part of Gojo's credit scoring approach.

Whether or not you can repay a loan without issues depends to a large extent on the size of the loan. For example, if I take a loan of $1 million, and spend it all at once without thinking, I will probably have a problem repaying it later. However, if I take a loan of $1, and use it to pay for some expenses, I will probably not have a problem repaying it later. In other words, credit scoring, which measures my ability to repay the loan, should be a function of the loan size.

Now, let's extend this view just a little bit more. The larger the loan size, the higher the credit score that should be required for the borrower, and therefore the higher the hurdle. The smaller the loan size, the lower the credit score that should be required for the borrower, and the lower the hurdle.

If that is the case, then for any given borrower, there should be a loan size that represents a manageable hurdle. For any kind of borrower, if we keep reducing the loan size, we will eventually find an amount that matches the maximum credit score they can achieve.

Maxima’s MBela team with an MBela agent at her house after a community gathering. / Koh Terai

The above paragraphs outline a little theorem about credit scoring, which is my favorite. Of course, we can take a further step to consider how we might apply it in practice.

Why don't we just look for the (maximum) loan size we believe a borrower can handle, and use that as their credit score? Rather than producing a score like '670' or '740', it would be easier for everyone to understand that $500 is the maximum possible amount they would be allowed to borrow. If the amount is clear, borrowers can plan their investment.

 We are actually applying this approach in a loan product currently being offered by Maxima, our partner in Cambodia. Their small digital loans project (also known as MBela) uses an automated assessment process to provide a credit score in the form of the maximum amount each person can borrow. The resulting credit score is easy for both the agent and borrower to understand.

Gojo wants to provide services that are innovative in their simplicity.


Yoshinari Noguchi is a researcher at Gojo and Company. He works in Gojo’s R&D team, and is currently looking into new ways to understand and support money management for low-income people, as well as analysis of data from the Hrishipara diaries and Gojo’s own financial diary projects.

August 27, 2021

Designing technologies for financial inclusion

In today's digital world, physical cash is rapidly becoming a relic of traditional financial systems that have disadvantaged the unbanked. By combining mobile digital financial tools (such as mobile remittances and loan disbursal) with other money management tools (such as financial education), we believe unbanked people can access financial services and break out of the poverty cycle. At Gojo, we wish to include financially excluded people and enable them to achieve financial goals and self-sufficiency.

I joined Gojo as a Software Engineer in August 2020 to enable this mission and solve our mobile engineering challenges, of which there are many. I’m going to talk about two of these today.

Challenge #1:

How do you create a Digital Field Application system that works even if the cell tower is down?

Most of the time, we take internet access for granted. That’s not the case for our customers in many countries, where the internet can be patchy and a precious resource. In order to ensure that our DFA didn’t stop working when the internet did, we had to architect our technology to be offline-first.

First, we loaded our app onto an Android tablet with 32GB of storage. We were able to store all the relevant data like names, photos, and loans locally on each device. The user, typically someone like a loan agent, doesn’t even need to know whether the tablet is online or offline, because the app behaves the same either way.

In the event of the internet dropping off, as soon as the tablet is reconnected to the internet, the data automatically syncs with our servers. That enables us to maintain data quality despite patchy internet.

Field officer taking photos of new M-Lady client using Gojo's DFA / Koh Terai

Challenge #2:

How do you enable field agents to update their Customer Forms on the fly to capture desired and customisable client data?

A critical piece of our Digital Field Application is to collect client information on tablets. This information is then used to make a decision on whether a client is eligible for a loan or not. These forms can vary widely depending on the data capture requirements of the partner. Keeping the fact in mind that the KYC forms could change on a regular basis, it did not make sense for us to go for a conventional route, i.e., to hardcode forms on the tablet itself.

We solved this problem by implementing the SDUI (Server Driven User Interface) architecture for our form screens. It works in conjunction with the offline architecture I mentioned above to render the latest version of the form to the client when the internet connectivity is present, or the latest cached version in the offline scenario. 

It provided us the following benefits: 

  1. Partners no longer need to depend on mobile developers to update the app to show specific changes in forms or to change the order of the UI. An agent can now use a web portal to make any changes he/she wants in the forms and it would be reflected in the app instantly.
  2. It’s easier for loan agents to introduce new form fields, like images and map views from their office computers and have it reflect on the mobile app.
  3. It enables the engineering team to create more reusable form components and scale across partners, because we do not have to hardcode the forms for each of our partners.

As we continue to scale our Digital Field Application across Gojo partner companies, these solutions will evolve, but we're confident that our approach of immersive design and innovative development is the best way to yield technology that is as resilient and adaptive as the people in the places where it is meant to be used.


Jeet Dholakia works as part of Gojo's technology team as a software engineer, focusing particularly on mobile engineering. He is passionate about solving complex engineering problems and good mobile design, and is currently working on Gojo's Digital Field Application and Customer Mobile Application.

August 27, 2021

Designing technologies for financial inclusion

In today's digital world, physical cash is rapidly becoming a relic of traditional financial systems that have disadvantaged the unbanked. By combining mobile digital financial tools (such as mobile remittances and loan disbursal) with other money management tools (such as financial education), we believe unbanked people can access financial services and break out of the poverty cycle. At Gojo, we wish to include financially excluded people and enable them to achieve financial goals and self-sufficiency.

I joined Gojo as a Software Engineer in August 2020 to enable this mission and solve our mobile engineering challenges, of which there are many. I’m going to talk about two of these today.

Challenge #1:

How do you create a Digital Field Application system that works even if the cell tower is down?

Most of the time, we take internet access for granted. That’s not the case for our customers in many countries, where the internet can be patchy and a precious resource. In order to ensure that our DFA didn’t stop working when the internet did, we had to architect our technology to be offline-first.

First, we loaded our app onto an Android tablet with 32GB of storage. We were able to store all the relevant data like names, photos, and loans locally on each device. The user, typically someone like a loan agent, doesn’t even need to know whether the tablet is online or offline, because the app behaves the same either way.

In the event of the internet dropping off, as soon as the tablet is reconnected to the internet, the data automatically syncs with our servers. That enables us to maintain data quality despite patchy internet.

Field officer taking photos of new M-Lady client using Gojo's DFA / Koh Terai

Challenge #2:

How do you enable field agents to update their Customer Forms on the fly to capture desired and customisable client data?

A critical piece of our Digital Field Application is to collect client information on tablets. This information is then used to make a decision on whether a client is eligible for a loan or not. These forms can vary widely depending on the data capture requirements of the partner. Keeping the fact in mind that the KYC forms could change on a regular basis, it did not make sense for us to go for a conventional route, i.e., to hardcode forms on the tablet itself.

We solved this problem by implementing the SDUI (Server Driven User Interface) architecture for our form screens. It works in conjunction with the offline architecture I mentioned above to render the latest version of the form to the client when the internet connectivity is present, or the latest cached version in the offline scenario. 

It provided us the following benefits: 

  1. Partners no longer need to depend on mobile developers to update the app to show specific changes in forms or to change the order of the UI. An agent can now use a web portal to make any changes he/she wants in the forms and it would be reflected in the app instantly.
  2. It’s easier for loan agents to introduce new form fields, like images and map views from their office computers and have it reflect on the mobile app.
  3. It enables the engineering team to create more reusable form components and scale across partners, because we do not have to hardcode the forms for each of our partners.

As we continue to scale our Digital Field Application across Gojo partner companies, these solutions will evolve, but we're confident that our approach of immersive design and innovative development is the best way to yield technology that is as resilient and adaptive as the people in the places where it is meant to be used.


Jeet Dholakia works as part of Gojo's technology team as a software engineer, focusing particularly on mobile engineering. He is passionate about solving complex engineering problems and good mobile design, and is currently working on Gojo's Digital Field Application and Customer Mobile Application.

March 30, 2021

Familiar Objects Used in Unfamiliar Ways – Smartphones in Rural Cambodia

Last March, I conducted two weeks of ethnographic field research in rural Cambodia with the local staff at our partner company Maxima.

Aside from learning about the villagers and their behaviours around money, I also tried to understand their relationship with technology.

During my research, I observed one phenomenon that surprised me.

Many of the homes I visited had mysterious numbers written on their ceilings. They were written with permanent marker, or etched into the wood. I was baffled by what they were.

Mysterious numbers written and scratched into the ceiling of a home in Banteay Meas, Cambodia / Koh Terai

Can you guess what they are?

It turns out that they are phone numbers of contacts that are important to them — doctors, police, their family members, and relatives.

The baffling part is that these people all owned feature phones, and some of them even owned smartphones.

So naturally I asked them, “why don’t you put these numbers into your phone?

Their responses made me smile.

I don’t know how to register numbers into my phone, I only know how to receive calls”.

My phone is in English and besides, I can’t read

If I lose my phone, I would lose my data.”

“If the numbers stay in the phone, they sometimes get deleted. They never move if they are on the ceiling.

My phone is in English and besides, I can’t read

I felt enlightened after hearing their responses. It became clear to me how their relationships to their mobile devices are quite different from the relationship I have with my smartphone.

For me, this leads to other interesting questions we could ask like…
- What are their relationships with their mobile devices like?
- How would that influence their relationship to mobile apps?
- Does this behaviour tell us anything about the strengths of social ties in these communities?
- Are there clues we can derive from the way people use spatial memory to organize information?

“I got this phone 2 years ago. I don’t know how to use the phone. I only receive calls. I only remember if the last two numbers are 27 it’s my first daughter, if its 20, it’s my second daughter.”

How many people do you know that store phone numbers on their ceilings at home?


Koh designs products and services for Gojo. He spends time listening to clients and potential customers to deliver well-intentioned financial and digital products for low-income households.

March 30, 2021

Familiar Objects Used in Unfamiliar Ways  –  Smartphones in Rural Cambodia

Last March, I conducted two weeks of ethnographic field research in rural Cambodia with the local staff at our partner company Maxima.

Aside from learning about the villagers and their behaviours around money, I also tried to understand their relationship with technology.

During my research, I observed one phenomenon that surprised me.

Many of the homes I visited had mysterious numbers written on their ceilings. They were written with permanent marker, or etched into the wood. I was baffled by what they were.

Mysterious numbers written and scratched into the ceiling of a home in Banteay Meas, Cambodia / Koh Terai

Can you guess what they are?

It turns out that they are phone numbers of contacts that are important to them — doctors, police, their family members, and relatives.

The baffling part is that these people all owned feature phones, and some of them even owned smartphones.

So naturally I asked them, “why don’t you put these numbers into your phone?

Their responses made me smile.

I don’t know how to register numbers into my phone, I only know how to receive calls”.

My phone is in English and besides, I can’t read

If I lose my phone, I would lose my data.”

“If the numbers stay in the phone, they sometimes get deleted. They never move if they are on the ceiling.

My phone is in English and besides, I can’t read

I felt enlightened after hearing their responses. It became clear to me how their relationships to their mobile devices are quite different from the relationship I have with my smartphone.

For me, this leads to other interesting questions we could ask like…
- What are their relationships with their mobile devices like?
- How would that influence their relationship to mobile apps?
- Does this behaviour tell us anything about the strengths of social ties in these communities?
- Are there clues we can derive from the way people use spatial memory to organize information?

“I got this phone 2 years ago. I don’t know how to use the phone. I only receive calls. I only remember if the last two numbers are 27 it’s my first daughter, if its 20, it’s my second daughter.”

How many people do you know that store phone numbers on their ceilings at home?


Koh designs products and services for Gojo. He spends time listening to clients and potential customers to deliver well-intentioned financial and digital products for low-income households.

January 14, 2021

Building a credit scorecard in Myanmar

The landscape of Hpa-an, Kayin State, Myanmar. / Taejun Shin

Myanmar’s financial services industry is nascent compared to the rest of the world, since the country only started to open up after the transition in 2011 from military rule to a civilian government. With the transition came liberalization of the financial services industry, with the Central Bank of Myanmar becoming an autonomous entity, and the enactment of the Microfinance business law in 2012. Since then, the industry has been playing catch up with the rest of the world, specifically in the area of mass market consumer lending.

Banks in Myanmar have traditionally served the corporate sector with credit, and have only recently started to slowly expand their reach into the SME sector, with a couple of non-traditional banks dipping their toes into consumer lending. The biggest obstacle banks face is the majority of the population’s lack of credit history. This creates a catch-22 for the risk-averse banking sector, who will not lend to consumers without credit history, but cannot build credit histories for consumers without taking the risk of lending in the first place. Microfinance institutions have been left to pick up where banks fell short in providing lending services to consumers, taking high risk, and building credit histories.

Microfinance in Myanmar started with the mission of getting people out of poverty and extending financial inclusion. The gap in the provision of mainstream financial services has led to the popularity of microfinance among the un/underserved credit-hungry populace. As a result, while maintaining its social mission, the microfinance sector has also grown to be a provider of mass market retail lending, ranging from consumer lending to micro/small business lending. Such rapid expansion in the lending scene has brought the need for credit scoring to the forefront, especially among the no/thin file segment of the population. This is where the sector’s years of trial and error in building the credit history of no/thin file clients can begin to bear fruit, as the sector starts to address the need for stronger credit scoring and risk management by building credit scorecards.

A lady selling flowers to visitors of Bagan, the most popular tourist destination in Myanmar. / Taejun Shin

Credit scorecards: An introduction

So, what is a credit scorecard?

It is the heart of credit scoring. It is a checklist of data points that are collected and weighted to spit out a score that we call a credit score, and financial institutions use this score to measure the risk level of a consumer. Consumers who have high credit scores are usually considered low-risk, while consumers on the other end of the spectrum, who have low credit scores, are considered high-risk.

The credit score and its associated risk level can decide whether a consumer gets approved for a loan, the pricing on the loan (risk-weighted pricing), and in some cases, even the loan amount and term. With credit scoring playing an important role in the decision-making process, the need to understand how the credit scorecard is made becomes critical.

A credit scorecard is created by looking at data on past loans that the institution has made so that it can extrapolate its experience of past loans to future consumers. To do this, they first need to classify consumers as either “good” or “bad”, and an analysis is carried out to explore and extract a set of characteristics that makes a borrower “good” or “bad”. In this scenario, the definition of a “bad” consumer, in hindsight, is any consumer to whom the institution would choose not to offer a loan again. There are two main types of scorecards for making such an analysis: an expert scorecard and a statistical scorecard.

Let us begin with the expert scorecard. It is the most basic credit scorecard and the most commonly used scorecard. As its name suggests, it is a scorecard made with inputs from an expert. People with years of experience in lending and credit appraisal make a list of characteristics to check and score for any consumers applying for the loan. This is a very manual process that relies on the personal experience of seasoned loan officers and credit managers in the case of microfinance, and of the underwriting team, in the case of banks.

The statistical scorecard does not draw on any personal experience but instead on statistics. The scorecard is built by using regression analysis to find correlations between data points collected from consumers and the performance of their past loans. This often means that an institution has collected hundreds, if not thousands, of data points from consumers and their past loans to find the correlations.

There is a midway approach, aptly called a hybrid scorecard. This is the combination of the two scorecards where the statistical scorecard is evaluated by experts to create a final version of the scorecard.

Creating a credit scorecard

Financial institutions that are looking to build a scorecard need to evaluate whether they have sufficient data points covering:

  • Transaction history (volume and amounts of deposits, withdrawals, cash ins, cash outs, and payments)
  • Saving history (balances in individual account or across all deposit accounts)
  • Demographics (age, gender, location, etc.)
  • Loan performance (number of times a consumer is late for previous loan instalments, number of days late for previous instalments, history of delinquency)
  • Income data (individual / consolidated debt to income ratio)
  • Relationship with the institution (how long the consumer has been with the institution, other products of the institution used by the consumer)
  • Alternative sources of data such as the credit bureau, call/text data, social media usage, etc.

The more data points, the better the statistical scorecard is. If the institution does not have access to or has not accumulated sufficient relevant data points, they can create an initial scorecard by using expert team members who have the experience to make judgement calls in lending, while gradually transitioning towards a statistical scorecard. 

A restaurant owner providing buffet lunch for local people in Yangon City. / Taejun Shin

Transitioning to a statistical scorecard: The example of MIFIDA

The following is an example of one of Gojo’s partner companies, Microfinance Delta International (MIFIDA), and its journey to create a scorecard.

MIFIDA is a microfinance institution in Myanmar with around 150,000 customers and a portfolio of around $40 million. It was incorporated in 2013 but hit its stride in 2017, when it grew from a handful of branches to 60+ branches today. With such growth, the need to reevaluate its risk management policies and credit assessment became apparent. This in turn highlighted the need for a scorecard for its customers.

MIFIDA already had a scorecard for its MSME customers, but it was a basic expert scorecard that covered the usual characteristics such as: debt coverage ratio, the ratio of repayment amount to income, number of outstanding loans, age, years in the business receiving the loan, etc. But it did not have a scorecard for its mass market lending products, such as its group loans.

MIFIDA therefore set out to reevaluate its current MSME scorecard and to create a new scorecard from scratch for its group loans. Below, I will cover the re-evaluation and update of the MSME scorecard, and the challenges we encountered in the process. I hope to cover our journey toward creating a new scorecard for the group loans in a later post.

Relevant data is paramount for making a statistical scorecard, and this is exactly what MIFIDA did not have. It had only implemented its core banking system in recent months and even then, it only had transactional data going back as far as the data that had been migrated into the system. Despite being around seven years old, MIFIDA did not have digitised historical data on clients. There was also no guarantee that the digitised data was reliable.

This ruled out immediate creation of the statistical scorecard for MIFIDA, but as they had experts who have been making loan decisions for years now, they decided to create an expert scorecard based on the experience of their staff. They listed down everything that made a consumer “good” and “bad”. From that listing, the team trimmed it down to 14 specific characteristics that would be most telling of the customer’s behavior and provided the weightings on each characteristic to be scored. A new application form was then drafted so that the data needed for scoring could be captured.

Market-wide challenges in credit scoring

MIFIDA is using this new expert scorecard and application form as stepping stones toward a future statistical scorecard of its own. Apart from the lack of data points mentioned above, the current challenges that MIFIDA is facing in creating the statistical scorecard are:

  1. A lack of data analysts and data scientists in Myanmar. Even if you have the data, there are few people in Myanmar with the skills to do the necessary analytics to build and produce the scorecard. It would require a person well versed in R or Python to handle large datasets, do exploratory data analysis, find correlations using regression or one of a few other methods, and then make a production-level scorecard that could be used in the field.
  2. The lack of a credit bureau. Anyone who wants to double check a customer’s self-reported credit history will simply have to trust the consumer as there is no centralized database to check against. In recent years, MCIX (Myanmar Credit Info Exchange) has started to provide such a service to the microfinance sector, but it is still a nascent endeavour, as it currently only shows some of the loans that the customer has taken from other microfinance institutions, and sharing of delinquency data is still a work in progress. Until MCIX or the national credit bureau are fully-fledged,  with the majority of financial institutions onboard, MIFIDA will have to check credit histories either by building these histories itself, or through traditional means such as asking family, relatives or local authorities.
  3. Tying into the institutional lack of data is that most customers are no/thin file customers who are only just beginning to be financially included. This means that they are at the start of their journey to build a credit history with a formal financial institution. Building such histories takes time. On the other hand, it also presents an opportunity to financial service providers to get the data they want to collect from customers right, so that it can be processed and used for scoring in the future.

Financial institutions in Myanmar, MIFIDA included, are currently working on overcoming those challenges of building a statistical scorecard and transitioning from expert scorecards, as there is a whole world of new opportunities if the transition is successful. 

An artisan who makes umbrellas in Pathein City. The town is known for umbrellas. / Taejun Shin

The rewards of better credit scoring

Myanmar has seen one example of an institution that is inching closer to a full statistical scorecard, and the opportunity this has provided to that institution.

The institution is Yoma Bank. Their digital lending product, called SMART Credit, is made for the mass market with a hybrid scorecard in the backend that is recalibrated every year with the help of Experian, one of the biggest providers of credit scoring and analytics in the world. This has helped Yoma Bank to expand its lending portfolio to everyday consumers and to a new market segment that it would not normally lend to due to the associated risk.

MIFIDA hopes to replicate that success by building its own customers’ credit history, while using an expert scorecard to mitigate the current risks until sufficient data is collected for a statistical scorecard. MIFIDA will also look to move onto digital lending and digitizing much of its operations so that its loan officers can focus more on building relationships with customers instead of focusing on application forms and transactions. Such digitization would allow for the collection of well-structured data points that could be used to move onto a statistical model, enabling MIFIDA to expand more easily to new customer segments with reduced risk in future by providing a comparable baseline for the new segment’s credit scoring.


Kaung Set Lin is Gojo's Country Officer for Myanmar, and has over 6 years of experience in Myanmar's financial sector, primarily focusing on developing and implementing digital financial products. His work includes managing the rollout of Gojo's digital products, including our Digital Field Application (DFA).

January 14, 2021

Building a credit scorecard in Myanmar

The landscape of Hpa-an, Kayin State, Myanmar. / Taejun Shin

Myanmar’s financial services industry is nascent compared to the rest of the world, since the country only started to open up after the transition in 2011 from military rule to a civilian government. With the transition came liberalization of the financial services industry, with the Central Bank of Myanmar becoming an autonomous entity, and the enactment of the Microfinance business law in 2012. Since then, the industry has been playing catch up with the rest of the world, specifically in the area of mass market consumer lending.

Banks in Myanmar have traditionally served the corporate sector with credit, and have only recently started to slowly expand their reach into the SME sector, with a couple of non-traditional banks dipping their toes into consumer lending. The biggest obstacle banks face is the majority of the population’s lack of credit history. This creates a catch-22 for the risk-averse banking sector, who will not lend to consumers without credit history, but cannot build credit histories for consumers without taking the risk of lending in the first place. Microfinance institutions have been left to pick up where banks fell short in providing lending services to consumers, taking high risk, and building credit histories.

Microfinance in Myanmar started with the mission of getting people out of poverty and extending financial inclusion. The gap in the provision of mainstream financial services has led to the popularity of microfinance among the un/underserved credit-hungry populace. As a result, while maintaining its social mission, the microfinance sector has also grown to be a provider of mass market retail lending, ranging from consumer lending to micro/small business lending. Such rapid expansion in the lending scene has brought the need for credit scoring to the forefront, especially among the no/thin file segment of the population. This is where the sector’s years of trial and error in building the credit history of no/thin file clients can begin to bear fruit, as the sector starts to address the need for stronger credit scoring and risk management by building credit scorecards.

A lady selling flowers to visitors of Bagan, the most popular tourist destination in Myanmar. / Taejun Shin

Credit scorecards: An introduction

So, what is a credit scorecard?

It is the heart of credit scoring. It is a checklist of data points that are collected and weighted to spit out a score that we call a credit score, and financial institutions use this score to measure the risk level of a consumer. Consumers who have high credit scores are usually considered low-risk, while consumers on the other end of the spectrum, who have low credit scores, are considered high-risk.

The credit score and its associated risk level can decide whether a consumer gets approved for a loan, the pricing on the loan (risk-weighted pricing), and in some cases, even the loan amount and term. With credit scoring playing an important role in the decision-making process, the need to understand how the credit scorecard is made becomes critical.

A credit scorecard is created by looking at data on past loans that the institution has made so that it can extrapolate its experience of past loans to future consumers. To do this, they first need to classify consumers as either “good” or “bad”, and an analysis is carried out to explore and extract a set of characteristics that makes a borrower “good” or “bad”. In this scenario, the definition of a “bad” consumer, in hindsight, is any consumer to whom the institution would choose not to offer a loan again. There are two main types of scorecards for making such an analysis: an expert scorecard and a statistical scorecard.

Let us begin with the expert scorecard. It is the most basic credit scorecard and the most commonly used scorecard. As its name suggests, it is a scorecard made with inputs from an expert. People with years of experience in lending and credit appraisal make a list of characteristics to check and score for any consumers applying for the loan. This is a very manual process that relies on the personal experience of seasoned loan officers and credit managers in the case of microfinance, and of the underwriting team, in the case of banks.

The statistical scorecard does not draw on any personal experience but instead on statistics. The scorecard is built by using regression analysis to find correlations between data points collected from consumers and the performance of their past loans. This often means that an institution has collected hundreds, if not thousands, of data points from consumers and their past loans to find the correlations.

There is a midway approach, aptly called a hybrid scorecard. This is the combination of the two scorecards where the statistical scorecard is evaluated by experts to create a final version of the scorecard.

Creating a credit scorecard

Financial institutions that are looking to build a scorecard need to evaluate whether they have sufficient data points covering:

  • Transaction history (volume and amounts of deposits, withdrawals, cash ins, cash outs, and payments)
  • Saving history (balances in individual account or across all deposit accounts)
  • Demographics (age, gender, location, etc.)
  • Loan performance (number of times a consumer is late for previous loan instalments, number of days late for previous instalments, history of delinquency)
  • Income data (individual / consolidated debt to income ratio)
  • Relationship with the institution (how long the consumer has been with the institution, other products of the institution used by the consumer)
  • Alternative sources of data such as the credit bureau, call/text data, social media usage, etc.

The more data points, the better the statistical scorecard is. If the institution does not have access to or has not accumulated sufficient relevant data points, they can create an initial scorecard by using expert team members who have the experience to make judgement calls in lending, while gradually transitioning towards a statistical scorecard. 

A restaurant owner providing buffet lunch for local people in Yangon City. / Taejun Shin

Transitioning to a statistical scorecard: The example of MIFIDA

The following is an example of one of Gojo’s partner companies, Microfinance Delta International (MIFIDA), and its journey to create a scorecard.

MIFIDA is a microfinance institution in Myanmar with around 150,000 customers and a portfolio of around $40 million. It was incorporated in 2013 but hit its stride in 2017, when it grew from a handful of branches to 60+ branches today. With such growth, the need to reevaluate its risk management policies and credit assessment became apparent. This in turn highlighted the need for a scorecard for its customers.

MIFIDA already had a scorecard for its MSME customers, but it was a basic expert scorecard that covered the usual characteristics such as: debt coverage ratio, the ratio of repayment amount to income, number of outstanding loans, age, years in the business receiving the loan, etc. But it did not have a scorecard for its mass market lending products, such as its group loans.

MIFIDA therefore set out to reevaluate its current MSME scorecard and to create a new scorecard from scratch for its group loans. Below, I will cover the re-evaluation and update of the MSME scorecard, and the challenges we encountered in the process. I hope to cover our journey toward creating a new scorecard for the group loans in a later post.

Relevant data is paramount for making a statistical scorecard, and this is exactly what MIFIDA did not have. It had only implemented its core banking system in recent months and even then, it only had transactional data going back as far as the data that had been migrated into the system. Despite being around seven years old, MIFIDA did not have digitised historical data on clients. There was also no guarantee that the digitised data was reliable.

This ruled out immediate creation of the statistical scorecard for MIFIDA, but as they had experts who have been making loan decisions for years now, they decided to create an expert scorecard based on the experience of their staff. They listed down everything that made a consumer “good” and “bad”. From that listing, the team trimmed it down to 14 specific characteristics that would be most telling of the customer’s behavior and provided the weightings on each characteristic to be scored. A new application form was then drafted so that the data needed for scoring could be captured.

Market-wide challenges in credit scoring

MIFIDA is using this new expert scorecard and application form as stepping stones toward a future statistical scorecard of its own. Apart from the lack of data points mentioned above, the current challenges that MIFIDA is facing in creating the statistical scorecard are:

  1. A lack of data analysts and data scientists in Myanmar. Even if you have the data, there are few people in Myanmar with the skills to do the necessary analytics to build and produce the scorecard. It would require a person well versed in R or Python to handle large datasets, do exploratory data analysis, find correlations using regression or one of a few other methods, and then make a production-level scorecard that could be used in the field.
  2. The lack of a credit bureau. Anyone who wants to double check a customer’s self-reported credit history will simply have to trust the consumer as there is no centralized database to check against. In recent years, MCIX (Myanmar Credit Info Exchange) has started to provide such a service to the microfinance sector, but it is still a nascent endeavour, as it currently only shows some of the loans that the customer has taken from other microfinance institutions, and sharing of delinquency data is still a work in progress. Until MCIX or the national credit bureau are fully-fledged,  with the majority of financial institutions onboard, MIFIDA will have to check credit histories either by building these histories itself, or through traditional means such as asking family, relatives or local authorities.
  3. Tying into the institutional lack of data is that most customers are no/thin file customers who are only just beginning to be financially included. This means that they are at the start of their journey to build a credit history with a formal financial institution. Building such histories takes time. On the other hand, it also presents an opportunity to financial service providers to get the data they want to collect from customers right, so that it can be processed and used for scoring in the future.

Financial institutions in Myanmar, MIFIDA included, are currently working on overcoming those challenges of building a statistical scorecard and transitioning from expert scorecards, as there is a whole world of new opportunities if the transition is successful. 

An artisan who makes umbrellas in Pathein City. The town is known for umbrellas. / Taejun Shin

The rewards of better credit scoring

Myanmar has seen one example of an institution that is inching closer to a full statistical scorecard, and the opportunity this has provided to that institution.

The institution is Yoma Bank. Their digital lending product, called SMART Credit, is made for the mass market with a hybrid scorecard in the backend that is recalibrated every year with the help of Experian, one of the biggest providers of credit scoring and analytics in the world. This has helped Yoma Bank to expand its lending portfolio to everyday consumers and to a new market segment that it would not normally lend to due to the associated risk.

MIFIDA hopes to replicate that success by building its own customers’ credit history, while using an expert scorecard to mitigate the current risks until sufficient data is collected for a statistical scorecard. MIFIDA will also look to move onto digital lending and digitizing much of its operations so that its loan officers can focus more on building relationships with customers instead of focusing on application forms and transactions. Such digitization would allow for the collection of well-structured data points that could be used to move onto a statistical model, enabling MIFIDA to expand more easily to new customer segments with reduced risk in future by providing a comparable baseline for the new segment’s credit scoring.


Kaung Set Lin is Gojo's Country Officer for Myanmar, and has over 6 years of experience in Myanmar's financial sector, primarily focusing on developing and implementing digital financial products. His work includes managing the rollout of Gojo's digital products, including our Digital Field Application (DFA).

Newsletter

Sign up to receive news from Gojo here.