Machine Learning and Emerging Market Credit

Machine learning is transforming credit financing in emerging markets - opening up new opportunities for everyday investors. Read more to find out.

Machine Learning and Emerging Market Credit

Credit is vital to the functioning of the global economy, and credit risk assessment is the key process making it happen, determining who gets credit, and at what price. Historically, the process has involved scrutinizing financial information, such as payment history and any debt a potential borrower might owe, among other things.

In countries with well developed financial infrastructure, where the proper data sharing mechanisms and underwriting processes are in place, risk assessment is, generally speaking, cheap, fast and accurate (of course, there have been major exceptions to this rule).

As a result, credit is simply a part of daily financial life, whether borrowing for a mortgage, getting a business loan, or simply using a credit card. However, assessing risk and getting credit isn’t always this easy.

The struggle for credit in emerging markets

In emerging markets, access to credit is far more restricted, with SMEs in Africa citing limited access to financing as the single biggest constraint on their growth, higher than water, corruption, and even electricity in many cases. But why is that?

The answer is simple. Credit risk assessment has been both unreliable and cost-prohibitive. Without access to ‘thick’ and open financial records, primarily credit history, many lending institutions cannot reliably assess risk or offer loan products, with banks often suffering from adverse selection in less developed financial systems.

The good news is, thanks to AI-powered machine learning technology, that is all starting to change.

What is machine learning in credit risk assessment?

Machine learning is a form of data-driven artificial intelligence that allows computers to learn without having to be explicitly programmed. Algorithms leverage large datasets to ascertain patterns and relationships leading to the construction of complicated predictive models, often based on hundreds of different indicators.

In credit risk assessment, machine learning harnesses its algorithmic power to evaluate behavior and predict a party’s ability to repay loans. So huge is the potential impact of AI technologies to risk assessment in financial services, that McKinsey estimates it could deliver up to $379.2 billion of additional value each year.

Harnessing alternative data sources

The scarcity of traditional credit risk assessment data has been a problem in emerging markets. But, in recent years, in a bid to improve financial inclusivity, and with the power of machine learning behind them, financial services providers have turned their attention to other potential sources instead.

This has meant scrutinizing everything from permissioned mobile interaction data such as call activity and app usage data, to consumer habits, and even social media and messenger services. In some cases, satellite imagery and video technology has been used to estimate farming incomes and shop stock values respectively, helping to make decisions over the ability of those farmers and businesses owners to pay back loans in future.

Creating the models that determine the causal relationships between these alternative datasets and the creditworthiness of individuals and businesses would be unthinkable with a more traditional approach, which is often ponderous or arbitrary. Machine learning, however, moves away from this time-consuming and potentially inaccurate manual process towards a more inclusive and data-driven methodology.

Advantages of machine learning for credit risk assessment

Machine learning techniques have brought a raft of advantages to credit risk assessment in emerging markets that have helped extend the provision of credit efficiently and effectively.

The most important of these have included:

  • Improved ability to handle complexity. Machine learning algorithms can identify patterns in data that are not readily apparent through traditional regression models. They also deal much better with complexity — possessing the power to compute reams and reams of unstructured data that would previously have gone to waste (data that is not organized in a predetermined manner such as the video data from earlier).
  • Greater objectivity. Credit risk assessment used to be primarily based on qualitative factors, such as an individual’s employment history or personal financial situation, and was assessed by humans. With machine learning algorithms driving the process instead, analysis is entirely data-driven, making it much more objective and reducing bias.
  • Increased speed. Credit risk assessment has historically been both time and resource intensive, creating an economic barrier for the kind of low-value loans that are commonplace in emerging markets. Machine learning can unlock economies of scale for these customers by rapidly extrapolating credit risk insights from a variety of data sources, verifying data source authenticity, and ultimately, automating the credit decision-making process.
  • Improved accuracy. Machine learning algorithms provide models with greater predictive power. This decreases the chance of default, meaning lenders can be much more confident in their lending. Some lenders have seen an uplift of up to 12% in GINI (a measure of how accurate the model is in identifying “bad” borrowers, who will default in future, versus “good” borrowers, who won’t) as well as increasing the acceptance rate for new credit applications by 16% — all without any increase in bad debt. With increased confidence comes increased liquidity and more opportunity for people in emerging markets to borrow.
  • Automatically updated risk models. Machine learning performs better the more data it has to work with. This means, over time, machine learning powered risk assessment models constantly improve their ability to judge the creditworthiness of potential borrowers. The longer they are in action, the greater their potential to improve financial inclusion — and all with minimal human input.

Machine Learning in RWA

We already work with a number of new and exciting RWA companies harnessing machine learning technology to extend credit to individuals and businesses in emerging markets that were previously cut-off from accessing these kinds of services. These include:


Aella was built to simplify credit solutions in emerging markets. To do that, it uses machine-learning tools to review customers’ spending habits and determine their creditworthiness. Not only that, but Aella Credit also uses machine learning technology to verify a would-be borrower’s identity in lieu of more traditional forms of identification.


The Rubyx platform provides AI-powered risk assessment tools and an API-driven cloud platform to provide end-to-end digital lending capabilities. It aims to create a fairer system by using machine learning techniques to help transform the way informal entrepreneurs and small businesses gain access to capital-financing.


YoFio, based in Mexico, aims to distribute credit to local corner and grocery stores. Historically, assessing the creditworthiness of these corner stores has been hard, not to mention costly. Thanks to an innovative vision-based machine learning model which analyzes a simple video of the store and its inventory to estimate its true sales and risk of closure, YoFio has transformed the underwriting process, allowing for a far more efficient distribution of credit.

The problem of siloed data

Thanks to companies like these, the credit risk assessment process in emerging markets has been revolutionized in recent years. But if credit provision in these markets is going to reach its full potential, there’s still another piece in the puzzle that needs to be addressed — the issue of siloed data.

The causes of siloed data

In emerging markets, because of underdeveloped infrastructure within financial services, credit transactions often rely on a more informal or piecemeal approach through hyperlocal intermediaries where data remains siloed off (or in some cases, not recorded at all). Alternative data sources have certainly been a massive help, but good old fashioned credit history is still the creme de la creme when it comes to credit risk assessment.

Even within that more traditional framework, though, a joined up approach between providers is lacking. So, say someone did get access to credit, they wouldn’t necessarily be able to carry that credit score with them across to another lender — the information stays with the lending intermediary. This means that despite the massive improvements delivered by machine learning, credit markets still remain inefficient.

Proper credit information sharing across institutions is the solution. This even reduces the risk of banking crises, especially in developing countries, due to its moderating effect on adverse selection and bad/risky loans. And that doesn’t just apply to emerging markets, either. Better infrastructure for credit information sharing can help parties all around the globe assess risk.

Getting to grips with this problem requires using another different technology — blockchain — leveraging its inherent advantages of data immutability, security and transparency to create what is known as an on-chain credit scoring system.

What is on-chain credit scoring?

On-chain credit scoring (OCCS) is a new way of assessing creditworthiness that uses a blockchain to help provide a more accurate and transparent assessment of risk than traditional methods. On-chain credit scoring works by combining on-chain loan transaction data (such a DeFi loan repayment) with off-chain data, such as repayment history through real-world lending intermediaries, on an immutable and permissionless ledger (a blockchain).

The end result is effectively a public credit bureau that allows for trustworthy on-chain credit scores. This also ensures no single party has ownership over credit score data, and that any relevant party can access the credit scoring information to assess creditworthiness should they have legitimate cause to do so.

The Creditcoin approach

At Creditcoin, we’ve harnessed the power of blockchain and machine learning to provide a better structure to distribute credit in emerging markets. To do that, Creditcoin integrates with local lending intermediaries (many of which rely on machine learning methodology to help extend credit to previously excluded individuals and companies, such as those mentioned above) while matching borrowers and lenders according to their loan term preferences and recording the resulting loan transactions on a blockchain.

This level of transparency aids financial auditing, meaning any would-be lender can assess the creditworthiness of a potential borrower, as well as making credit markets more efficient and contestable.

How can I get involved?

Not only does this help improve financial inclusion, but it creates exciting new investment opportunities too. Through Creditcoin’s primary technology partner, Gluwa, investors can invest in finance providers in emerging markets all across the globe, generating exceptional returns of up to 20% APY while simultaneously helping to inject credit into markets that need it most.

What’s more, you can view the credit performance of select fintech partners live on the Creditcoin Block Explorer, helping you audit the performance of the companies asking for your investment.

If you want to get involved, simply download the Gluwa App to start investing today!

🍏 Apple | 🤖 Android

Website | Twitter | Discord | Medium | Youtube | Telegram(ANN) | Whitepaper(ENG) | Opensea

About Creditcoin

Creditcoin is a foundational L1 blockchain designed to match and record credit transactions, creating a public ledger of credit history and loan performance and paving the way for a new generation of interoperable cross-chain credit markets.

By working with technology partners, fintech lenders such as Aella, and other financial institutions across global emerging markets, Creditcoin is securing capital financing, building credit history and facilitating trust for millions of underserved financial customers and businesses based on the principles of RWA.