Scale
PKR 105B+
of credit enabled through partner banks, from a few hundred thousand rupees a month in 2021 to more than PKR 280 million a day in 2026.
How AdalFi turned banks’ own data into more than PKR 105 billion of credit for the Pakistanis the formal system could not assess, and what five years of performance shows.
Cumulative credit enabled, Jul 2021 to Jul 2026, across multiple partner institutions. Local-currency figures converted at PKR 280 per USD.
of credit enabled through partner banks, from a few hundred thousand rupees a month in 2021 to more than PKR 280 million a day in 2026.
people entered the formal credit system for the first time, in a country where only 12% of adults borrow formally.
90+ day default rate, on a book where 53% of borrowers had never borrowed before. Sector SME NPL runs 4–8%.
SME borrowers in Pakistan now runs on our rails. PKR 42B enabled, roughly 7 in 10 borrowing for the first time.
towns reached, with 69% of lending outside the four largest cities. Once the decision no longer needed a branch visit, distance stopped deciding.
of the book went to women, above the 9% national target.
Credit fails the two customers an economy needs most. Pakistan digitised its deposits, not its decision to lend.
Exclusion here is no longer about who holds an account; nearly everyone does. It is about who a bank can say yes to. Payments went digital, 88% of retail transactions and PKR 50 trillion a year on instant rails, while formal borrowing barely moved. The customer is thin-file or no-file: a real financial life, invisible to a system reading collateral, salary slips and a bureau record instead of behaviour.
But only a few million hold any credit relationship with their bank.
And ~75% could not be sure of raising emergency funds.
Against 31.6% across South Asia: a 5× gap.
Between them they account for most of the country’s economic activity, and almost none of its formal credit.
6.5% of firms hold a bank loan. Shops, workshops, small manufacturers and traders carry much of the country’s employment, and nearly all of them are undocumented: no audited accounts, often no property in the owner’s name, a business conducted in cash and relationships.
Nothing was built for them. Microfinance stops below them; commercial lending starts above them and asks for the collateral they do not have.
Two groups the system reached last, for different reasons. Pakistan ranks 148th of 148 on the gender gap, and 14% of women hold a full-service account against 56% of men. A woman is less likely to hold property in her own name, a documented salary, or a credit record, and where she earns, it is often at home and in cash.
Rural households face a different geometry: assessment required presence, presence required a branch, and branches follow density. Distance from a city became distance from capital.
The largest group and the least visible. People earning below the median, whose needs were small enough that assessing them was never worth the cost. Gig workers and freelancers, whose income is real but arrives irregularly and matches no salary slip.
And anyone new to credit, who cannot build a record without a loan and cannot get a loan without a record. Three in four adults say they could not readily raise emergency money. Most of them are working.
No collateral, and often no salary slip a bank will accept. The answer is no, not because they cannot repay, but because the branch cannot tell. What judgment there is leans on a loan officer’s read of a person: their name, their gender, where they live. So they turn to a relative, a committee, or a moneylender, at rates that turn a problem of timing into a debt.
A bank cannot profitably send a loan officer to assess a PKR 45,000 loan. Manual underwriting is too slow and too costly for small tickets, so mass-market consumer credit never pencils out. The result is a deposit base of ~100M accounts the bank simply cannot lend against, and millions of good borrowers it turns away sight unseen.
The vegetable vendor, the ice-cream shop, the workshop: each needs collateral-free working capital to grow. Instead it waits three to six weeks for a manual decision, and is usually turned away.
Microfinance is not the answer either: its limits sit far below what a growing business needs. A ceiling that low keeps a business alive without ever letting it expand: enough to replace stock, never enough to add a line, a machine, or a second location. Microfinance lets a business sustain itself. Formal credit is what lets it grow. Held at subsistence, the same firm has no buffer either, so one bad season, one illness or one late payment becomes existential.
SME underwriting is done by hand: a loan officer’s judgment, weeks per file. The bank spends real money processing applicants in order to decline most of them, and never reads the transaction data it already holds on the ones who would have qualified.
The deeper constraint is arithmetic. No institution can manually assess millions of account holders. A bank with several million business and salaried depositors would need an army of analysts to form a view on each one, so it forms a view on almost none of them and waits for the few who apply. The only way out is to stop assessing on application and start scoring the entire base continuously, in the background, without anyone asking.
The IFC puts the credit gap for formal micro, small and medium enterprises across emerging markets at roughly US$5.7 trillion: tens of millions of firms that cannot obtain the financing they need. Small business lending is the largest identified unmet credit market in the world, and the constraint in most of it is the one described above: not capital, but the cost of deciding who to give it to.
The one channel built to close this gap is going backwards. In the year to December 2025 Pakistan’s microfinance sector lost roughly a million borrowers and its sector capital adequacy ratio turned negative. Over the same period the bank channel added roughly 1.8 million, from 4.6 million to 6.4 million.
Read together, those two movements describe a handover nobody planned. The institutions built to serve the excluded are shrinking, and the institutions with the balance sheet to scale it were never able to underwrite it. Inclusion is not stalling for want of demand. It is stalling in the gap between the two.
Getting a person through that first opening, from an account to a first loan, is exactly where the market fails, and exactly where our work begins.
Beneath it sits self-omission: the people who wanted credit but never applied, certain the answer would be no. They appear in no rejection statistic, and no decline rate captures them. Pre-qualified credit is the one form that finds them, because the offer arrives before the question has to be asked.
The people at the top of that funnel are not sitting still. They already borrow, from committees, shopkeepers, suppliers and moneylenders, at prices that punish the very cash-flow gap they are trying to bridge.
the effective cost of informal credit to a farming household, from commission agents charging a mark-up close to 50% to professional moneylenders documented near 80%.
is how quickly most small traders are back in debt after it has been cleared for them, in studies of vendor borrowing across South and South East Asia.
effective annualised cost on some app-based nano-lending in this market, on tenors of only a few weeks.
The pattern holds across emerging markets, and the harm is not only the price. Informal credit leaves no record, so repaying it faithfully for a decade earns a borrower nothing at all. And where digital lending has arrived without underwriting behind it, it has reproduced the same trap faster: opaque terms, punitive rollover, and collection practices that work on shame.
The evidence was always there. Nobody could afford to read it.
Every problem in the last section has the same root: the cost and the bias of a human decision. So we rebuilt the decision.
Every bank already holds years of a customer’s deposits, transfers and cash flow. AdalFi reads that signal, billions of transactions, and turns it into a real-time credit decision, so a bank can lend to a customer it already has but never recognised as bankable.
No branch visit, no collateral, no loan officer’s read of a person, no weeks of paperwork.
For the borrower it is a yes that finally reflects how they actually manage money; for the lender it is a way to say that yes profitably, at scale, across the whole shelf, from a first salary advance to an SME line.
The partner bank’s own account and transaction history: the customer’s real financial behaviour, already on file.
Our models turn that behaviour into an instant, calibrated credit assessment, and improve with every repayment they observe.
The bank lends, on its own balance sheet, in minutes, at scale, to towns a branch network never reached.
Tens of thousands of repayments flow back into the models every month. Each one sharpens the next decision: a flywheel that gets more accurate, and harder to replicate, the more we lend.
Which is why the economics invert: every lender that joins makes the next decision cheaper and more accurate, so the borrowers who were least worth assessing under the old cost structure become the easiest to serve under this one.
The cost of a decision fell to almost nothing, and the incentives inverted with it.
There is a consequence worth naming plainly. Under the old cost structure the borrowers least worth assessing were the ones with the smallest, least documented, least conventional needs, which is to say the people for whom credit would have mattered most. Exclusion was not a judgment anyone made about them. It fell out of the price of looking.
And the field teams did not disappear. The work moved: time that went into collecting documents, chasing signatures and ferrying files went into serving customers instead.
A first request, on a thin file.
The barrier was never the first “no.” It was that the first “no” was final.
Salary regularised, balances steadier, a small loan repaid. Months pass. No re-apply needed.
The model re-scores continuously. A decline is a snapshot of a moment, not a verdict on a person.
Credit extended instantly. The door doesn’t stay shut; it reopens on its own.
A small business no longer chases a lender to be assessed on accounts it does not keep. The lender reads the transaction history it already holds, sees which businesses qualify, and makes the offer. Weeks of waiting to be evaluated becomes an offer that is already there.
Nothing in the decision requires a branch, so how far a household sits from a city stops being a credit variable. And no name, gender or address is an input, so the discretion a woman’s application used to meet is simply not in the process.
A gig worker, a freelancer or a daily-wage earner has no salary slip, but the money still moves and the pattern is there to read. And someone with no credit history stops being invisible, because the account itself is the history.
Not what we lent: what would not have happened otherwise.
When a decision no longer needs collateral or a branch visit, the people the old system could never see are the first to be reached. Tens of thousands crossed into formal credit for the very first time: a first loan, a first record, the first rung of a financial life. And a first loan is rarely the last: once a borrower is visible, one approval becomes a relationship that compounds for years, and a deposit base the bank could never lend against becomes a book it can build on.
people entered the formal credit system for the first time, in a country where only 12% of adults borrow formally. Each one is now visible to a bank for the rest of their credit life.
The median borrower earns about Rs 50,000 per month, below the national average household income.
Borrower monthly income against the national average household income. Demographic breakdowns are computed across the consumer and SME products for which we hold borrower-level records; the PKR 105B book covers 28 products.
What it is: collateral-free business finance, revolving working capital lines and term loans, for the “hidden middle” too big for microfinance and too small or too new for traditional commercial lending.
This is hard because SME underwriting is broken for both sides. Assessing a business loan by hand takes a bank four to six weeks and a loan officer’s judgment, and most applicants are declined at the end of it, costly for the lender, slow and humiliating for the business. Transactional scoring inverts the exercise: the bank reads its own transaction data, sees which businesses already qualify, and offers them credit in minutes. Faster processing and a lower cost of origination are what expand the underserved segment: when a decision costs minutes instead of weeks, tickets this small become economic to write.
Two numbers give the scale, and the first is a national figure rather than one of ours. The state-backed collateral-free SME scheme, the country’s flagship effort at exactly this problem, reached roughly 8,000 businesses over four years across eight banks. On AdalFi, roughly 6,400 businesses have borrowed, about 1 in 43 SME borrowers in the country’s banking system.
Note what sits in the middle. A large share of small business owners finance their businesses on personal credit, because that is the only product they can qualify for. They should be receiving a business line sized to their turnover; they are offered a personal loan sized to a salary they do not have. The ladder is not only about how much a borrower can reach; it is about whether the product they reach is the right one.
The largest single line within SME lending is revolving credit: about PKR 23.5B. A business draws on it when cash is tight and repays as revenue arrives. The informal version is everywhere: stock taken from a distributor at a premium, or a rotating committee among neighbours. Both cost more than they appear to while building no record at all. A formal line is cheaper, and every repayment becomes the file that earns the next, larger one.
Median loan size by product, converted at PKR 280. Not to linear scale.
The gap it fills: women are routinely asked for collateral they don’t own and credibility the system won’t extend. Data-driven scoring judges cash-flow behaviour, not gender.
AdalFi reached ~15,600 women borrowers, a 12% share of the book, against a national policy target of 9% for women borrowers at banks. That is not a quota we set out to hit. It falls out of a model that decides on cash-flow behaviour and never takes a name, a gender or an address as an input.
on personal loans, women receive larger loans relative to income than men in every income band:
evidence the model extends real trust, not a token limit, once a woman clears it.
Axis runs 0–25%. Dashed line marks the 9% national policy target for women borrowers at banks. Overall share across the book: 12%. The ceiling is product design and the device gap, not demand.
What it is: seasonal, revolving agricultural credit that follows the crop cycle, so farmers buy seed and fertiliser at the right time instead of borrowing from a middleman against the harvest at punitive rates.
In a developing economy, physical distance is itself a credit barrier. Assessment required presence, presence required a branch, and branches follow density and deposits, so distance from a city became distance from capital. Removing the branch visit from the decision removes distance from the equation. What it displaced was the advance taken from a buyer against the coming harvest: credit priced without competition and repaid in crop.
What it is: a short advance against wages already earned, released before payday.
The alternative was rarely a bank. It was a relative, a colleague, a shopkeeper, or a moneylender priced by the month. Each carries a cost that is not financial: you have to ask. Asking a family member for money in the middle of the month is a small humiliation that people arrange their lives to avoid, and it is one reason a medical bill becomes a crisis rather than an inconvenience.
An advance against earned wages removes the asking. The money is already the worker’s; only the timing changes. There is no conversation, no explanation, and no obligation to anybody in the room.
To roughly 56,600 workers: a median of seven advances each.
Some opportunities have a window. A provincial programme put electric two-wheelers within reach of tens of thousands of households, and the constraint was never demand or capital. It was whether a lender could stand up a bespoke programme, assess a population it had never scored, and disburse at volume before the window closed. Under manual origination that is not possible, so opportunities of this kind are usually watched rather than taken.
Roughly 20,000 electric two- and three-wheelers were financed on AdalFi. What made that possible was not a lending product. It was four AdalFi properties.
A scheme with its own eligibility rules, subsidy mechanics and partner set, configured rather than rebuilt.
Applicants included self-employed parents with no salary slip and no credit record, assessed instead on obligations they were already meeting.
Funds to the end customer without a queue forming behind manual processing.
Dealers, verification agencies and government systems connected into a single flow.
Environmental figures are modelled estimates on a stated assumption set, not measurements. Most of this volume arrived inside a provincial scheme that has since largely concluded, so the total is a stock rather than a run rate.
More than PKR 105 billion, written through the crunch. From a few hundred thousand rupees a month in 2021 to processing more than PKR 280 million a day.
Growth is the least interesting thing in this report, but it has to be established first, because everything that follows depends on this being a real book rather than a pilot. Three questions: how fast it grew, when it was written, and what it was lent against.
Agriculture, SME finance and revolving business lines, counted as separate segments.
Personal loans and top-ups, salary advances, vehicle financing and cards.
A deliberate weighting toward the segment with the widest credit gap. Two things make this growth unusual. First, its timing: it happened as inflation peaked near 38% and the policy rate sat at a record 22%, precisely when most lenders retreated into government securities. Second, its breadth: what began with one product at one bank now spans 28 products across 11 institutions, from Rs 16K salary advances to Rs 3M business lines.
A 0.2% default rate, on a book where most borrowers had never borrowed before. Reach and scale count for nothing if the book goes bad.
The people the system called “risky” were simply unseen.
The bottom line for partners: inclusion and asset quality are not a trade-off. Reaching new-to-credit, lower-income borrowers and keeping losses near zero is a design outcome of data-driven underwriting.
businesses reached in four years, across eight banks, by the flagship state-backed collateral-free SME scheme
SME borrowers in Pakistan now borrows on rails we run. Four years of subsidy moved a number that digital underwriting moves in months. The constraint was never appetite for lending; it was the cost of deciding.
the State Bank’s national target for women as a share of borrowers
of our borrowers are women. On personal loans, women receive larger loans relative to income than men in every income band.
hold the overwhelming share of Pakistan’s formal credit and its branch network
reached, with 69% of lending outside the four big metros. When underwriting stops depending on a branch visit, the branch map stops deciding who can borrow. Some of the districts we reach are among the country’s poorest.
net-new borrowers added by Pakistan’s entire bank channel in the year to December 2025
of net-new bank borrowers, and of net-new SME borrowers, nationally. Our direct footprint in a country of 245 million is small, and we would rather say so, but a measurable share of every new borrower the banking system gained last year came through this model.
One partner runs the comparison for us: the same products, the same market, the same brand, with a manual desk deciding one way and AdalFi deciding the other. The only variable is the underwriting.
personal loans a month originated by the manual desk running alongside AdalFi
the manual desk’s monthly volume: between 1,300 and 1,600 loans a month, decided on AdalFi.
of every card the institution has ever issued came through its manual channel, accumulated over its entire history
of that lifetime book was added on AdalFi in three years: 6,900 new cardholders.
of cardholders through the manual channel are women
of AdalFi-issued cards at the same institution go to women, 60% higher for the same product in the same market. The model reads cash flow; it never reads gender.
And the credit reaches down, not up. Most borrowers earn below the national average, the people for whom a first loan changes the most.
The finding in this report is not really about AdalFi, and it is not only about Pakistan. It is that the people a credit system turns away, or never asks, are very often creditworthy already.
Tens of thousands of the borrowers here held no credit record at all on the day they borrowed, and repaid at a rate the sector does not achieve on its most documented customers.
That distinction matters, because a cost is something anyone can remove, and the conditions that created it are not particular to one market. Across much of South Asia, South East Asia and Africa the same three facts hold at once: almost everyone now has an account, almost no one has the documents a lender was built to read, and a credit decision costs about the same to make whether the loan is fifty dollars or fifty thousand.
Nothing described in these pages required new capital, a new branch network, or a change in regulation. It required reading evidence a lender already holds about customers it already has. Wherever that is true, so is the rest of it.
There is a version of each of them in every market with this shape. None became a different person when the answer changed. They were simply, finally, legible.
What stands between hundreds of millions of people and a first loan is not their risk. It is that nobody has looked.
Credit reaching households below the median income, absorbing shocks that would otherwise force an asset sale.
Seasonal finance timed to a crop cycle rather than to a buyer’s advance against the harvest.
An assessment that takes no name and no gender as an input, in the country ranked last of 148 on the gender gap.
Working capital for the firms that carry most of the country’s employment.
Financing that put roughly twenty thousand electric vehicles on the road.
And first formal credit for tens of thousands of people who had none, which is the one this report is really about.
Roughly 129,000 people sit behind the figures in this report: one country, one platform, five years.
The number of people worldwide who still cannot be assessed runs into the hundreds of millions, and a great many of them hold no account at all, which means the method described here cannot yet reach them. A disproportionate number are women. What became of the people it did reach is the subject of the next report, and it will be a harder one to write.
The gap was never the people.
Credit enabled is compiled from AdalFi’s monthly ledger, July 2021 to July 2026, across all partner institutions. Revolving facilities are recorded on a limit basis. The 2026 figure covers January to July and is partial. Borrower characteristics are computed from the products for which borrower-level records are held, deduplicated by national identifier, which is a subset of the full book. Local-currency amounts converted at PKR 280 per USD.
State Bank of Pakistan, Financial Stability Review 2025, SME Finance Review and Payment Systems data. Karandaaz Financial Inclusion Survey 2024. World Bank Enterprise Surveys and Global Findex. IFC MSME Finance Gap, updated estimate published 2025. Pakistan Microfinance Network. Small and Medium Enterprises Development Authority. Pakistan Bureau of Statistics, Household Integrated Economic Survey 2024/25. World Economic Forum Global Gender Gap Report 2024.
Costs of informal agricultural credit are drawn from State Bank of Pakistan published research on grower mark-ups and from studies of the commission agent system by the Pakistan Institute of Development Economics and the International Growth Centre, alongside historical survey work on professional moneylender rates in Sindh. The speed at which small traders return to debt is from vendor borrowing studies by Innovations for Poverty Action and the Kellogg School of Management. Pricing on app-based nano-lending is from market reporting.
AdalFi provides the assessment layer; partner banks originate and hold every loan. Government-subsidised programmes are executed by partner banks on AdalFi, so outcomes there are co-produced and are described throughout as credit “enabled” rather than as lending by AdalFi. Environmental figures are modelled from a stated assumption set and reported as central estimates, not measurements.
Smarter credit decisions. Lending beyond barriers.
Book a demoGet the report↓