• An alternative form of credit scoring is being used successfully in emerging markets as data on consumer consumption is rich because e-commerce is widespread.
  • The new scoring (ACS) works with artificial intelligence and social media, instead of paper-based scoring methods that depend on consumers having a bank account.
  • ACS ultimately leads to higher financial inclusion.

The concept of credit scoring first arose in the US and was initially deemed to be successful to the degree it was used as a primary, or even sole, mechanism for identifying one’s financial fitness.

However, there are differences between the US and maturing markets like Indonesia.

US credit scoring focuses primarily on banking and financial data, and mainly consists of credit cards, loans and banking usage data and alarms. To date, there are not many non-banking entities participating in this problem space in a unified way.

In areas like Indonesia, the market tends to favour one-stop-shop locations to fulfill their purchasing needs. People use e-commerce to top up their phone credit, pay bills, book domestic and international travel, commute, pay loans. They also use it to make investments, such as in gold or mutual funds.

This means that the data on consumer consumption is rich, relevant and immense. These data points are able to determine consumers' financial and consumption fitness. This concept is known as Alternative Credit Scoring (ACS).

ACS is a system that uses technology as the basis for its operations. This system is an upgrade from conventional credit scoring mechanisms that use assorted paperwork and manual credit history. ACS uses advanced technology automation (Artificial Intelligence and Machine Learning) instead.

The system also employs a set of alternative data in projecting potential borrowers. ACS no longer relies on paper-based verification here either. Instead, it uses alternative data such as social media, electronic transactions and cellular data to conduct and assess the consumer's feasibility study as they usually do not have access to banking services.

Image: PwC

ACS provides efficiency in decision-making, lower information asymmetry, and mutually exclusive systems for lenders and borrowers. The integrated data helps financial institutions make holistic and thorough risk analyses regarding potential consumers through a more comprehensive approach.

In turn, this facilitates digital and traditional lending institutions in identifying borrowers' capabilities when they propose the issuance of financial credit.

An alternative method of credit scoring through ACS would help unlock and bring more accessible funding to individuals, especially Micro, Small & Medium Enterprises (MSMEs).

These individuals predominantly fall under the unbanked and underbanked groups and tend to have no access to more formal financial institutions due to the associated higher credit risk.

Through the ACS method, financial institutions could have better visibility over potential borrowers' capacity to pay back if given any lending facility, especially with regards to the underserved segments who lack credit capacity information. Financially enabling more MSME segments will help middle-class entrepreneurs grow organically by giving them responsible economic enablement.

All of these factors, consequently and indirectly, lead to higher financial inclusion.

According to a World Bank study, a 1% increase in financial inclusion can contribute to an annual GDP growth per capita of ~0.03%. In Indonesia, the convincing economic growth and increased financial inclusion, can promote a GDP growth rate of 5% by 2020.

Here are several ways we can further accelerate financial inclusion in Indonesia, which may serve as a model for other emerging economies:

1. Create an accurate model specifically for the market

Every country has their own culture. It builds the foundation of how we think, how we spend, how we invest, how we treat our wealth and how we manage our credit. Copy-pasting models directly from matured countries like the US or China will not work.

We need smart statisticians and data leaders with experience dealing with a mature credit scoring ecosystem to apply their knowledge and build unique adaptive models for Indonesia.

2. Consolidate alternative data sets

Consumption-related data is everywhere. Unfortunately, it is non-uniform, disorganised and scattered. In the initial phase – the consolidation process – the finding, identifying and capturing of data must be conducted. Then, we use basic data cleaning, correlating and storing technology concepts to build the unified BigData concept. Only then do we have a basis for leveraging advanced technology (AI and DS modelling) and can start correlating, experimenting and building models.

Image: Experian

3. Implement adaptive learning

US credit scoring focuses mostly on banking, which is an almost static ecosystem; it does not require regular changes to the mechanism. Our ecosystem poses different exciting spaces. We focus primarily on users' consumption behaviour, a constantly changing factor. Our model needs to be adaptive to account for such regular changes.

While ACS is promising in regards to the substantial opportunity and potential impact it brings in improving financial inclusion, it requires strong collaboration and partnership from everyone in the ecosystem (data source entities, lenders and government).

To mature the system, ample time to bake and experiment is also needed.

Nevertheless, ACS will not replace traditional financial-based credit scoring. The dream is for both to live side by side and help strengthen the confidence of our lending partners, in turn enabling a world of cheaper, faster, more targeted and sustainable lending.

When performed correctly – with the right grit, passion, perseverance and governance – ACS will unlock new markets in our journey towards better financial inclusion in Indonesia.

With the continuous support of the government and infrastructure facilitation for data sharing, alternative data availability can be improved.

In addition, the constant development of innovative machine learning models is required to enhance the handling of model validation, performance, data privacy, fairness and interpretability.

Nitin Jain, VP of Engineering at Tokopedia, contributed to this article.