For years, many lenders used a borrower’s text messages to judge whether they could repay a loan.
Bank alerts showed salary payments, debit alerts showed spending and repayment messages showed earlier loans, and software turned those messages into an estimate of income and habits.
That helped lenders in markets where credit bureau records are thin. As a stand-alone method, though, it has become much less dependable.
Customers and banks have moved to other channels, the messages are hard to read accurately, and phone platforms and regulators now limit what lending apps may collect.
This article explains what SMS scoring is, why it falls short and what lenders can use instead.
What SMS scoring is
A credit score is a number used to predict how likely someone is to pay back a loan on time, as the CFPB explains.
Alternative data means information from outside the traditional credit file, such as mobile money transactions, digital payments and business records, in the way the IFC describes it.
SMS scoring is one kind of alternative-data scoring. It estimates how likely a person is to repay from the financial text messages on their phone, such as bank alerts, mobile money confirmations and loan repayment notices.
With the customer’s permission, software finds the financial messages, sorts the transactions and pulls out figures such as how often money arrives, how regular salary or business deposits are, how much goes out and whether earlier loans were repaid.
A model turns those figures into a score, and the lender uses it to decide eligibility, set a limit or send the application to a person. Lendsqr’s documentation explains why lenders do this.
Banks and other institutions send SMS alerts for salary payments, transfers, airtime purchases, loan disbursements and repayments, so if a borrower has no bank statement, a lender can check the alerts to see whether they regularly receive money and pay bills.
Research shows that phone data carries real signal. A study in the World Bank Economic Review found that patterns in call records at a South American telecom predicted default and, for people with thin financial histories, beat models built on credit bureau data.
It used call and SMS metadata held by a telecom, so it shows that phone data has value and says nothing about reading customers’ messages.
Why SMS scoring falls short
1. Financial activity has moved beyond the SMS inbox
Banks send SMS alerts, and many customers now switch them off, because each message costs money. In Nigeria, banks including GTBank told customers that the SMS alert fee would rise from ₦4 to ₦6 per message from 1 May 2025, citing higher telecom rates.
Daily Trust reports that a growing number of customers are turning SMS alerts off. The pattern is similar elsewhere. In Kenya, Equity Bank began charging KES 2.26 per transaction alert in April 2023, and Business Daily reported that Co-operative Bank and Absa Kenya also charge for alerts.
In South Africa, Standard Bank says its My Updates notifications are free in the app or by email, that SMS notifications cost 45 cents each, and that it now delivers them to the app inbox for customers with a digital profile.
When customers drop SMS alerts, a lender reading their inbox sees only part of what happens in their accounts. Say a trader receives 600,000 a month through a business account whose bank sends app notifications only.
Her inbox shows a few messages from a second bank and nothing from the account that matters, so the lender underestimates her income. Several accounts, wallets and payment apps widen the gap, since each may use a different channel.
2. SMS data does not show real income
A transaction alert shows that money moved, and it does not show that the person earned it.
Customers receive transfers from relatives, move money between their own accounts, collect payments on behalf of a supplier or take short-term advances they must repay.
All of these raise the recorded inflows without raising the ability to repay a new loan.
Here is an example. A trader’s inbox shows 1,200,000 arriving in a month. Of that, 700,000 came from customers, 300,000 was moved over from her own savings account and 200,000 was a supplier’s payment that she passed on.
Her real income from sales is 700,000. A lender that counts the whole 1,200,000 as income may offer a loan she cannot afford. A text gives the amount and the date, and it cannot say what the money means.
3. SMS records are incomplete and hard to standardize
Each bank word alerts in its own way, so software has to recognize many formats and keep up when a bank changes a template, an abbreviation or a currency format. Lendsqr’s own tool has to turn raw SMS into transactional data, and a change in wording can break that step. Duplicates cause errors too.
A bank and a payment app may both send a message about the same transaction, and a system that counts both will overstate income.
Some alerts leave out the balance or the description, and some messages have been deleted, so lenders need to check what they extract against other records.
Read more: Does Android still allow lenders to scrap SMS from phones for scoring?
4. Access raises privacy and consent problems
An inbox holds much more than bank alerts. It also holds one-time passwords, private conversations and messages from employers, so broad access collects far more than a credit decision needs. Phone platforms and regulators have started to push back.
Google Play’s policy says an app must be the user’s default SMS, Phone or Assistant handler to request SMS permissions, and it bars apps from using other methods to get the same data, which closes this route for most lending apps.
In Nigeria, the FCCPC’s guidelines for digital lenders require applicants to confirm that their apps will not have access to customers’ call logs, contacts or photos and gallery.
The World Bank’s page on data privacy in digital credit lists the risks, including data collection that goes far beyond the purpose a customer agreed to. It also notes that, in a survey of 11 markets, only one regulator required lenders to tell customers which data they used for credit scoring.
A lender should be able to explain why it needs each data source and how long it keeps the data.
5. SMS scoring can disadvantage customers with little digital activity
A model that depends on messages assumes every applicant generates plenty of them. A market trader who is paid in cash and through informal arrangements may run a steady business and leave almost nothing in her inbox.
A salaried worker whose bank texts every credit may score higher simply because their activity is easier to see. The model rewards visibility and can overlook people who are able to repay.
The IFC’s May 2026 report, Cracking the Credit Code, says the benefits of alternative-data scoring are uneven and that women, informal workers and microentrepreneurs are still underrepresented in credit portfolios.
What lenders can use instead
Bank statements give the most direct view of deposits, withdrawals, balances and transfers. With the customer’s permission, a lender can use them to tell regular income from one-off inflows.
A salaried applicant’s statements show salary deposits, loan deductions and regular bills, and a trader’s statements show customer payments, supplier transfers and swings in balance.
A high turnover does not prove strong repayment capacity, so the lender still has to read the figures with care. Lendsqr’s announcement of its partnership with Mono says that Mono covers more than 30 financial institutions and collects statements with the customer’s explicit consent.
Other sources fill different gaps. Mobile money and wallet records show how often a customer is paid and how steady their balance is, which helps with people who work outside banks.
Credit bureau reports show recorded loans and repayment history across lenders, and they catch debts that never appear in the applicant’s own statements, though they say little about people who have never borrowed formally.
For small businesses, point-of-sale reports, invoices and supplier statements show sales and costs. APIs let authorized systems share this data directly, which reduces the need for screenshots and inbox analysis.
How Lendsqr supports lenders moving beyond SMS
Lendsqr’s decision model page lists the data a lender can build into its approval rules: credit bureau queries, ecosystem data on how a borrower has behaved with other lenders on the platform, scoring and velocity rules, and bank statement data that customers share by consent through MyBankStatement, Mono and Fenestra. Lendsqr also offers an SMS tool for Android apps.
A lender that uses it should treat the result as one input among several, check its permissions against Google’s policy and local rules, and compare it with other records before relying on it. To try the platform, sign up for a Lendsqr account.
Seven steps for moving beyond SMS-only scoring
Step 1: Audit your current process
Check which messages you collect, which figures you extract and how much they count in approvals. Compare the model’s predictions with real repayment.
Step 2: Decide what underwriting needs to know
List the questions you must answer, such as whether income is regular, what debts the borrower already has, how much the business sells and how they repaid before.
Step 3: Add verified data sources
Start with sources tied directly to repayment capacity, such as bank statements, bureau reports, mobile money statements or verified business receipts. Make sure you have a legal basis and the customer’s permission for each one.
Step 4: Test against past loans
Compare the SMS-based model with a version that adds other data. Look at defaults, missed repayments, approval rates and results for applicants with thin files.
Step 5: Keep affordability checks in underwriting
A score estimates the chance of default, and it does not tell you whether the installments fit the borrower’s budget. Review income, essential costs, existing debts and when cash arrives before you approve.
Step 6: Monitor performance and data quality
Track repayment results and investigate cases where the model’s prediction and the borrower’s behavior differ. Review each data source when a bank changes its alerts.
Step 7: Explain data use to customers
Tell applicants what you collect, why you need it and how it affects the decision. Limit access to authorized staff and keep the data only as long as you need it.
Read more: Why gig workers are denied loans, and how better credit scoring helps
Where SMS data still fits in lending decisions
SMS scoring helped digital lenders reach borrowers that bureau data could not, and it still has a place when the messages are available, relevant and collected with consent.
On its own, though, it depends on partial records, unclear income signals, tighter limits from phone platforms and regulators, and uneven coverage.
Lenders get a clearer view when they use SMS-derived data alongside bank statements, bureau reports, mobile money records and business records, keep affordability checks in underwriting and test every source against real repayment.
The goal is an assessment that matches how people earn, receive, spend and repay money, built on data the borrower has agreed to share.
