Introduction: Credit access, gaps, and the promise of alternative credit scoring

Traditional credit scoring systems often exclude women and underserved borrowers, particularly those without formal financial histories or collateral. Approximately 1.3 billion adults worldwide still lack a bank account, while three billion people lack sufficient credit history to access financing. Consequently, the financing gap for micro, small, and medium enterprises (MSMEs) rose by $1.3 trillion from 2015 to 2019, and the financing gap for women-led MSMEs continues to widen, reaching 1.9 trillion. In response, a new generation of credit scoring models powered by alternative data and artificial intelligence (AI) is emerging to redefine credit scoring for borrowers lacking formal histories.

What are alternative data-driven credit scoring models?

Alternative data-driven models broaden access to credit by moving beyond traditional metrics such as formal income documentation. Instead, they draw on behavioural, transactional, and digital footprint data. Common inputs include records from mobile money, digital wallets, utility payments, point-of-sale systems, and geolocation. AI plays a varied role in these models; while some fully automate score generation using machine learning, others limit AI to transaction categorisation or fraud detection, continuing to rely on rule-based or hybrid scoring approaches.

Mapping the alternative credit landscape: Market trends, growth, and case studies

The alternative credit ecosystem is young and dynamic, with over 75 percent of mapped firms founded in the last decade. Significant traction is visible across East Asia and the Pacific, Africa, and South Asia. Personal loans and MSME credit account for 47 percent and 32 percent of offerings, respectively. When deployed inclusively, these models meaningfully expand women’s access to credit. For instance, Kaleidofin in India has facilitated over seven million loans, largely to women, while Eshandi has disbursed nearly one million loans to women in Sub-Saharan Africa. Despite women’s reliable repayment habits, most models are gender-neutral by design, and only 12 percent of firms publicly reference women in their communications.

Challenges and opportunities: Risk, fairness, and inclusion in AI-driven credit scoring

While AI and alternative data can improve access, they also introduce risks of bias. Biases in training data, such as gaps in women’s digital footprints or underrepresentation in credit histories, can reinforce exclusion. Proxy variables like education level or geography may unintentionally entrench discrimination. Privacy and consent risks are particularly pronounced in environments where women rely on shared phones or have limited digital literacy. Consequently, explainability and fairness checks are critical to ensure algorithms expand access equitably.

Unlocking inclusive credit at scale: What’s next?

To enable responsible innovation, regulators and financial institutions must create regulatory sandboxes and AI-testing environments to evaluate accuracy, loss estimation, and fairness before open-market rollout. Practitioners should embed fairness testing, such as approval-rate parity, across model lifecycles and encourage sex-disaggregated and intersectional reporting to inform and improve model design. Furthermore, promoting cross-sector data partnerships and supporting open-finance frameworks will expand the visibility of underserved borrowers, whilst building consumers’ digital credit skills ensures inclusion translates into empowerment.