How India pays, one scan at a time.

Each dot is one payment system. The further right, the larger an average payment. The scale jumps by ten at each gridline, because the systems differ by that much: an RTGS transfer is thousands of times a UPI payment.

An average is pulled up by a few very large payments, so it can sit well above what an ordinary payment looks like. The RBI does not publish the typical (median) payment, so we do not know it.

A falling average does not mean people spend less. It means UPI took over smaller and smaller payments, the ones once made in coins and small notes, so each added payment pulls the average down. A rising number of payments with a falling average is what a habit spreading looks like.

Two lines: UPI in blue, and every other retail system added together in grey. The grey line has grown too, but UPI's has grown so much faster that the rest looks flat beside it. A crore is ten million, so 100 crore payments a month is a billion.

This is a count of payments, not of people. One shopkeeper receiving a hundred payments a day adds a hundred; we do not know from this table how many Indians use UPI.

The left panel counts payments, the right panel counts rupees. The same systems sit in the same order in both, so read across: a long bar on the left and a short one on the right is a system of many small payments; the reverse is a system of few large ones.

Each line is the rupees a system moved in a month. RTGS sits far above the rest because it carries the settlements of companies, banks and governments; a single transfer can be crores. March towers each year: it is when the financial year closes, taxes fall due and accounts are squared.

These sums dwarf India's GDP because the same rupee is counted every time it changes hands, and much of it is money moving between financial institutions. They measure how busy the plumbing is, not the size of the economy.

Most Indians with a bank account have a debit card; far fewer have a credit card. Yet credit cards now carry more payments. The debit card's everyday work, paying a shop from your own account, went to UPI, which does the same thing from a phone. Credit cards kept what UPI did not offer at first: credit, reward points and online shopping.

The left panel counts trips to the ATM; the right panel is all the notes and coins held outside banks. If digital payments had replaced cash, both would fall. Only the first has. People draw cash less often but in larger amounts, and the stock of cash has kept growing with the economy. The deep dip in the right panel is November 2016, when old ₹500 and ₹1,000 notes were withdrawn.

Cash held is not cash spent: much of it is savings kept at home, or money held for weddings, farming and travel. We do not know how many payments are still made in cash.

A QR code costs a shopkeeper nothing but a printed sticker; a card machine has to be bought or rented. That difference is much of why UPI spread to vegetable carts and tea stalls that never took cards. White-label ATMs are run by companies that are not banks. Micro ATMs are the handheld devices of banking correspondents, the agents who bring banking to villages.

These are counts of things in place, not of things in use. A QR code printed once and never scanned is still counted.

IMPS was the instant transfer before UPI, and much of its work has moved to UPI. NACH debits are payments you agreed to once and that then repeat: loan instalments, insurance premiums, SIPs, utility bills. NACH credits and APBS go the other way, from employers and governments to people. AePS cash withdrawals are rural India taking benefit money out at a banking correspondent. UPI over USSD works on a basic phone with no internet, and is used very little.

The channel is the screen a payment starts on, not the system that carries it: a UPI payment from an app is a mobile payment, and so is a NEFT transfer made in a bank's app. Businesses still move large sums through net banking on a computer, which is why internet banking carries few payments but a great deal of money.

Each bar is one fiscal year, April to March. The blue part is UPI, which began in August 2016; the grey part is everything else. The year in progress is left out until its March is in.

The dotted line in November 2019 is where the RBI changed its table, so the lines stop and start again rather than run through. NEFT reads much the same on both sides. Card payments do not: the old table's "usage at PoS" is close to the new table's card payments of every kind, which suggests it also counted online card payments that the new table puts under "others". The RBI does not say, so read the two card stretches as different measures. The jump in card payments late in 2016 is demonetisation, when cash ran short for weeks and shops that had a card machine suddenly used it.

For decades the cheque was how salaries, rents and business bills were paid. The count rose with the number of bank accounts until NEFT, cards and then UPI took its work. Cheques have not disappeared: businesses, landlords and governments still write them.

A clearing house was once a room where banks of a town exchanged the cheques drawn on each other. The Cheque Truncation System replaced the paper with images, so cheques no longer travel. Centres were folded into three grids, at Mumbai, New Delhi and Chennai, and then into one national grid. That is why the city lines stop: the cheques did not vanish, they moved to the grid.

Each bar is the e-rupee outstanding on one 31 March. The RBI's table counts e-rupee held, not how often it is spent, so we do not know how many people use it or for what. Retail e-rupee is issued in tokens of fixed denominations, like notes.