By mid 2025, Brazil had approximately 87 mobile devices per 100 inhabitants. Formal bank account ownership ranges from 55% to 65%, depending on the counting method. That difference is substantial, representing tens of millions with an active, data producing financial device but no bank account. For credit scoring, this is where the opportunity lies.
For the unbanked, mobile activity may offer more than the richer data trail held by most bureaus. Often, it is the only trail available. To assess whether a thin file borrower in Fortaleza or Belem can repay a R$1,500 personal loan, a lender may have no bureau record to consult. Yet that person often has two to 4 years of mobile account history, which can reveal a coherent financial behavior pattern when interpreted properly.
Why Mobile Data Fits Credit Signals Structurally
Mobile financial behavior matters as a credit signal because of its underlying economics, not because it is unusual or clever. A prepaid phone account involves managing a limited resource, such as airtime or data balance, against predictable demand and recurring expenses. The resulting patterns, including top up frequency, service lapses, and responses to a low balance, directly reflect how someone manages constrained resources during ordinary conditions and periods of stress.
This connection is supported by evidence rather than theory. Academic teams studying Sub Saharan Africa and South Asia have reported meaningful links between prepaid mobile management measures and microloan repayment results. Coefficients differ across markets, yet the direction is consistent enough for this to merit consideration as a signal category.
We avoid overstating the finding. The correlation is real and the causal mechanism plausible, but explanatory power is partial. Mobile behavior predicts a meaningful share, not all, of thin file borrowers' repayment variance. It is one layer of a multi signal architecture, not a standalone credit scoring system.
The Mobile First Borrower Profile
To treat mobile data as a credit source, first identify the mobile first borrower. In Brazil, this generally means someone age 18 to 35 in an urban or peri urban area, working in informal services or trade, using a smartphone with 2 to 5 years of mobile account history, and lacking formal credit history. The person is likely unbanked or underbanked, perhaps holding a basic neobank account but using few formal credit products.
This borrower's finances usually move through several channels: informal work paid in cash, Pix transfers through a basic bank account or digital wallet, prepaid mobile account activity, and utility payments. Although the digital footprint is thinner than a fully banked person's statement, it remains coherent over time. Reading that pattern can give a lender enough basis for a meaningful lending decision.
A mobile first approach also affects the lending experience. Lenders serving this group generally originate through a mobile app using a digitally native application flow. Permissions and consent for alternative data can fit naturally within that process. Someone who has already approved location sharing for an app and receives a clear explanation of mobile account access for credit scoring is unlikely to view that request as unusual friction within a smartphone native experience.
Mobile Activity Signal Categories
The mobile features we use span several groups, with differences in coverage, predictive strength, and the consent infrastructure each requires.
Carriers provide recharge and balance management data, including transaction times, amounts, and account balance history. In most mobile credit models, this is the central signal. It covers prepaid subscribers who have a data sharing agreement with the lender or scoring provider. Consent is comparatively straightforward: the borrower authorizes account record access, much as they would authorize bank statement access during a standard loan application.
Where available, mobile money and P2P transfer data reveals aspects of the borrower's financial network. Regular Pix receipts from a small group of senders, suggesting stable work or family support, differ from occasional transfers arriving from many senders. Even without formal payslips, transfer timing can provide an indication of income regularity.
If a lender operates a proprietary app and receives the required permissions, app behavior signals can include session frequency, usage by time of day, and financial activity inside the app. Because this information is more personal than carrier billing records, consent requires greater care. We use it as an opt in enrichment layer, not a standard input.
How Lenders Integrate Mobile Data
Lenders adding mobile data to underwriting generally choose among three approaches. One is a direct carrier API, supported by a formal data sharing agreement with one or more carriers. It offers the broadest coverage but demands investment in carrier relationships. Another is an aggregator with existing carrier partnerships that provides normalized mobile data through one API. The third is a scoring provider, like Supersymm, which manages data access and supplies a score instead of raw data, so the lender avoids that handling burden.
The main difference is a build or buy choice for data infrastructure. Lenders focused on originating credit and managing portfolios, rather than engineering data systems, will usually favor a scoring provider. At origination, a REST API call supplies the score and leading contributing signals, fitting into the existing loan origination system without requiring the lender to create or maintain mobile data infrastructure.
Supersymm built its API for this integration pattern. One call, using the borrower's identifier and permissioned data touchpoints, returns a score in under 800 milliseconds median latency, plus a confidence band and the leading signals with relative weights. This gives credit teams the decision format they need, rather than raw data requiring additional processing.
Limits of Mobile Data
Mobile first scoring does not replace the bureau system wholesale. When borrowers have bureau files, that information has validated predictive value based on decades of observed outcomes, with a depth mobile data cannot equal. For most lenders, the sound design is not to use mobile data instead of bureau data, but to use it where bureau information is absent or inadequate.
The population also has practical limits. Mobile scoring performs best when borrowers have at least 12 months of active mobile history. New SIM activations and frequent SIM switching leave too little signal depth. This excludes only a minority of the intended thin file lending population, but lender workflows still need to recognize the boundary.
Those helped most by mobile first scoring are established thin file borrowers, people who have used mobile services responsibly for years but whose creditworthiness has never been assessed because the relevant data was overlooked. This group is large, underserved, and ready for scoring. The data already exists. What has been absent is the infrastructure for interpreting it.