How alternative and cash flow data are enhancing consumer underwriting worldwide

Traditional credit models perform well for borrowers with conventional credit histories, but they leave many creditworthy consumers behind whose income or employment doesn’t fit the standard mold. As work becomes more flexible and the gig economy continues to develop, a growing share of consumers are falling into this category. This is a challenge that is shared globally.

Alternative data and cash flow underwriting are practical tools to address this gap. While distinct, both approaches can be used to improve credit access and risk management.

Alternative data captures behavioural and digital signals, like platform activity and device characteristics. It can significantly improve risk differentiation for thin or non-traditional credit profiles, supporting higher approval rates without materially increasing losses.

Cash flow underwriting focuses on real-time analysis of income, balances, and spending patterns to directly measure repayment capacity. It is particularly effective at improving early-stage delinquency detection and line sizing. Cash flow analysis has long been central to corporate banking and is now gaining momentum in consumer lending as access to real-time financial data becomes more widely available.

The pace of consumer cash flow underwriting adoption varies meaningfully by region. In markets such as the US and UK, Open Banking frameworks and mature data infrastructure have made consumer cash flow underwriting increasingly accessible for fintechs and banks alike. In Canada, progress has been more gradual as Open Banking standards continue to evolve. In parts of Latin America, where bureau coverage can be limited and bank account penetration has historically been lower, lenders initially relied more heavily on alternative data embedded within digital platforms.

This regional dynamic is illustrated in Mexico through ecosystems like RappiCard, where behavioural signals from its digital platform have been used to expand access in the absence of traditional credit histories.

Alternative data case study: RappiCard

In 2024, researchers from the National Bureau of Economic Research examined whether data from a delivery platform could predict creditworthiness for borrowers with zero formal credit history.

The study focused on RappiCard Mexico, which issues credit cards through the Rappi mobile app, a platform widely used across Latin America for on-demand delivery of everyday goods.

Using monthly account-level revenue and cost data, the researchers built a machine-learning model to predict the probability of default (defined as 60 days past due). Across a sample of 146K first-time borrowers, the model achieved an area under the receiver operating characteristic curve (AUC-ROC) of 0.796, indicating strong discriminatory power.

The highest contributing inputs to model performance were the digital footprint characteristics, such as device type, operating system, and email provider. The transaction-level data ranked second, including order frequency, payment methods, and time-of-day patterns. Predictive accuracy improved for borrowers with a longer transaction history on the platform. By contrast, geographic and socioeconomic variables provided only marginal incremental value.

While this example focuses on platform-derived digital behaviour rather than bank-account cash flow, it demonstrates how non-traditional signals can materially improve risk assessment for thin-file borrowers.

Measuring ability to pay in real time

At its core, lending relies on two pillars: willingness to pay and ability to pay. Credit bureau data primarily captures willingness through historical use of credit products, while cash flow underwriting focuses on a customer’s ability to pay.

Historically, assessing ability to pay has been manual and document driven. Borrowers submit pay stubs or bank statements, and analysts review documents and calculate ratios. In many markets, particularly where digital infrastructure is still developing, this approach remains common today. The process does not scale easily, which is one reason automated consumer underwriting came to rely so heavily on credit scores.

Analyzing cash flow data

As Open Banking frameworks and fintech data platforms have matured, lenders are now able to analyze transaction-level financial data directly from bank accounts with borrower consent. This allows automated extraction of practical financial signals such as:

• Income level and consistency
• Average and minimum balances
• Fixed obligations and recurring expenses
• Spending volatility
• Overdraft and negative balance events

These signals move beyond a historic point-in-time income assessment, which, depending on the product, may be self-declared and unverified. By incorporating multiple behavioural dimensions, they provide a more complete view of repayment capacity and reflect changes to a borrower’s current financial condition faster than they would appear at the credit bureaus. Credit scores only react after stress shows up as derogatory statuses. Cash flow often shows stress earlier by detecting when income or balances drop or spending accelerates. In many cases, these changes precede delinquency by weeks.

In practice, for fintechs and mid-market lenders, consumer cash flow underwriting does not require highly sophisticated features to be effective. Many lenders find that a small set of variables, (i) income, (ii) balances, and (iii) overdraft behaviour, explain a large share of performance differences. These signals can be used to:

• Approve borrowers who would otherwise be declined
• Right-size credit limits based on real affordability
• Adjust pricing or terms for near-prime customers
• Monitor portfolios and intervene earlier when risk rises
• Practical Implementation

Cash flow underwriting tends to be most impactful in products that are revolving or short-term in nature, and in portfolios serving mid-to-lower credit tiers. Lenders operating strictly in super-prime segments often see less incremental benefit because traditional models already perform well there.

That said, it raises an important question. During periods of economic stress, such as the 2008 financial crisis, when losses rose sharply even across prime portfolios, could real-time cash flow monitoring identify early signs of financial strain and better manage exposure?

One practical challenge is experimentation, particularly for non-bank lenders that lack large “on-us” datasets or a broad share of wallet across their customers. Unlike bureau data, cash flow data rarely comes with long historical archives. This makes traditional back testing difficult. Lenders can address this by partnering with fintechs that provide a cash score or by running pilots and starting with narrow use cases such as secondary approval paths.

Whether looking to implement alternative data or cash flow underwriting, a simple approach might look like this:

1. Define a clear business objective, such as expanding approvals or improving early risk detection
2. Launch a controlled pilot on a subset of applications or accounts
3. Measure outcomes against existing models
4. Scale gradually once performance and operational impacts are understood

Ultimately, success depends on pairing alternative data and cash flow insights with disciplined risk frameworks. Together, they strengthen underwriting decisions to expand access to credit while protecting portfolio performance.

Here at Payson, we’ve helped dozens of fintechs and banks implement alternative and cash flow data into their underwriting decisions. Whether you’re exploring these capabilities for the first time or looking to refine existing strategies, we partner closely with teams to translate data into practical, scalable outcomes. If you’re looking to build on your current underwriting strategies, reach out and we’d be happy to connect.

Payson is an Event Partner of Open Banking Expo Canada 2026, on March 5 at the MTCC in Toronto. Click here to find out more about the agenda, speakers and exhibitors – and register to get your ticket.