Going Cashless? SBP payment reports · six years of PDFs
The State Bank of Pakistan says 92% of retail payments are now digital. I rebuilt that figure from six years of SBP reports and tested it against the amount of cash in circulation.
Introduction
The State Bank of Pakistan reports that 92% of retail payments are now digital, a figure widely quoted as proof that Pakistan is going cashless.
Project
Rebuilt that figure from six years of SBP Payment Systems Reviews, published only as PDFs, split it into transaction count and rupee value, and tested it against the stock of cash in circulation.
Challenge
The 92% counts transactions processed by banks, not rupees, and treats ATM cash withdrawals as digital. By value, digital channels carry 40%, and real cash holdings haven't moved since 2019.
My role
Solo project: PDF extraction pipeline, tidy series, analysis, a market brief for a fintech product team, and hand-drawn SVG charts. Python, pdfplumber and pandas.
Not yet. Payments inside banks and payment apps are going digital fast, but Pakistanis hold as much cash as they did in 2019, after inflation.
The 92% figure is a share of the payments banks and payment providers process. Cash handed between people never reaches them, so it isn't in the figure at all. By value, digital channels carry 40% of retail payments, up from 17% in mid-2023. That rise in value is the real shift, and it comes from bank mobile apps and internet banking, not wallets.
What the 92% actually measures
SBP splits retail payments into 11 channels and counts everything except bank-branch and agent counter transactions as digital. Rebuilt from SBP's own tables, the figure comes out at exactly the published 92.3% for January–March 2026.
Two things are hidden in it. The figure counts transactions, not rupees. And SBP counts ATM transactions as digital, though 97.5% of them are cash withdrawals.
Share of formal retail payments made through digital channels
Show as a table
Count and value tell different stories
Wallet apps such as JazzCash and Easypaisa now handle most transactions, but they are small payments. Bank branches handle few transactions, but they move most of the money.
Retail payments by channel
Show as a table
Average payment size by channel
What the branch rupees are
Most of this is business and high-value payments. It is the largest pool of payments still done on paper, and wallet apps barely touch it.
Testing the headline against cash
If Pakistan were going cashless, the stock of cash would be shrinking. In rupees it has more than doubled since 2019: +142% to PKR 11.9 trillion. Consumer prices rose by the same 142%, so in real terms the cash stock hasn't moved.
Cash is slowly losing ground to bank deposits, which grew faster (+205%). People also keep withdrawing more cash: ATM cash withdrawals doubled, and SBP counts each of them as a digital transaction.
Currency in circulation
Cash as a share of broad money (M2)
ATM cash withdrawals
Show cash figures as a table
Both things are true. 92% of formal transactions are digital, and the cash stock is flat in real terms. Pakistan is digitising its formal payment system much faster than it is shrinking its cash economy.
Raast merchant payments: the fastest-growing piece
Raast is SBP's instant-payment system. Its person-to-merchant (P2M) payments, mostly QR codes at shops, went from 4.3 million to 55.9 million a quarter between Jul–Sep 2025 and Jan–Mar 2026, and to 86.0 million (provisional) in Apr–Jun 2026.
It is still small. P2M was 7.6% of Raast transactions and 1.4% of Raast value in January–March 2026, with an average payment of PKR 5,720 against PKR 28,420 for a transfer between people.
Raast payments by type
Show as a table
Merchants accepting QR
Merchant payments, Jan–Mar 2026
Acceptance is ahead of use. In January–March 2026, in-store wallet payments plus Raast P2M came to about 140 million against 2.5 million QR merchants. That is roughly 56 payments per merchant per quarter, under one a day. It's a rough indicator, because the two counts don't line up one to one.
The long view: banks' own channels since 2019
SBP's definition of bank channels hasn't changed since 2019, so this is the longest like-for-like view. Bank mobile apps went from 13 million to 753 million transactions a quarter and overtook branches in late 2022 and ATMs in late 2023. Branch counter transactions fell, but the value going through branches kept rising.
Bank and microfinance-bank channels
Show as a table
What is inside the wallet boom
Wallet payments doubled in a year. In-store payments grew fastest, from 21 million to 111 million a quarter, and online merchant payments grew 2.5 times. But the jump in July–September 2025 came mostly from "miscellaneous payments", which went from 108 million to 466 million in one quarter. SBP doesn't say what that category contains.
Branchless-banking wallet payments by purpose
Show as a table
What this means for a fintech product team
- Size the market by count and by value separately. A wallet that leads on count (60% of transactions) is competing for 5% of the value. Business-to-business payments, payroll and supplier payments compete for the other 95%.
- The biggest offline pool is cheques and branch cash. That is about PKR 94 trillion a quarter, or 56% of retail value. Business payment tools, bulk disbursement and digital cheque replacement go after it directly.
- Build around cash-in and cash-out. ATM withdrawals are still growing and agents handled 155 million transactions in January–March 2026. Products that make it easy to move between cash and digital will reach more people than ones that assume cash is going away.
- The merchant opportunity is getting QR used. Acceptance is spreading fast (3.8 million QR merchants) but each merchant takes few QR payments. Incentives, instant settlement and tools that give a shopkeeper a reason to prefer QR are where the value is.
- Treat SBP's quarterly numbers as provisional. Revisions can be large, and a big share of recent wallet growth sits in an unexplained category. Check the latest edition before building a forecast on any one quarter.
Data and method
The pipeline is written in Python. It downloads 27 SBP Payment Systems Reviews (18 quarterly, 9 annual, FY18–FY26), extracts every table with one generic PDF parser, and builds tidy time series. Cash, deposits, M2 and CPI come from SBP EasyData.
- One parser for every layout. SBP changed the table layout almost every year. The parser reads the shape every table shares (title, units, period header, label and numbers), so a new layout doesn't need new code.
- Latest revision, one report per quarter. Each quarter appears in up to five reports. The pipeline takes all of a quarter's channels from its latest report, so definitions never mix within a quarter.
- Built-in checks. Channels must add up to SBP's published total in every quarter, or the build stops. The chained CPI index matches SBP's published inflation to within 0.2 percentage points.
- No splicing across definition changes. The all-channel quarterly series starts in Jul–Sep 2023, the first quarter with wallets included and direct debit reported separately from branches. The bank-only series covers 2019 onward on one definition.
- Caveats. No dataset counts cash payments, so the cash tests use the cash stock, its share of broad money, and withdrawals. Raast overlaps the app and wallet channels. Card e-commerce changed its data source in FY23. The latest quarter is provisional.
- Calendar. SBP's fiscal year runs July to June, so FY26-Q3 is January–March 2026.