Japan's Code Payments Grew 270 Times in Seven Years. Electronic Money Went Flat
On 31 March 2026 the Ministry of Economy, Trade and Industry published Japan's cashless payment ratio for 2025. The headline was 58.0%. The year before, the same ministry published 42.8%.
A fifteen point move in a national payments statistic in one year would be extraordinary. It did not happen. Most of that gap is a change to the denominator, and the ministry says so in the release. I want to walk through the honest version, because it is more interesting than the headline, and because underneath it there is a result about interfaces that I did not expect to find.
Two numbers, two denominators
The cashless ratio is card and code payments divided by consumer spending. The argument has always been about the second half.
Until this year the denominator was private final consumption expenditure. That figure includes something called imputed rent for owner occupied housing, which is the rent a homeowner would notionally pay themselves if they were a tenant. Nobody pays it. It cannot be paid by card, because it is not paid at all. It is about ¥57 trillion, roughly 17% of the total. Leaving it in the denominator put a hard ceiling on the ratio no matter how much of the real economy went cashless.
So in December 2025 a METI study group replaced it. The new domestic indicator divides by household final consumption expenditure minus imputed rent. The old formula survives as the international comparison indicator, because everyone else still computes it that way.
Here are both, from the METI release.
| Year | Domestic indicator | International indicator |
|---|---|---|
| 2010 | 16.6% | 13.2% |
| 2020 | 29.8% | 29.7% |
| 2024 | 52.8% | 42.8% |
| 2025 | 58.0% | 46.3% |
The comparable year on year move is 42.8% to 46.3%. Three and a half points, which is a good year and an ordinary one. The government's old target of roughly 40% by 2025 was met early. The new targets are 65% on the domestic indicator by 2030 and 80% eventually.
None of this is hidden. It is in the release, with the formulas. But the number that travels is 58%, and if you are comparing it to something you read last year you are comparing two different things.
What actually moved
The composition is where it gets useful. Of the ¥162.7 trillion paid without cash in 2025, credit cards were 82.7% (¥134.6 trillion), code payments 10.2% (¥16.6 trillion), electronic money 3.7% (¥6.0 trillion) and debit cards 3.4% (¥5.5 trillion).
Now compare 2018 to 2025 on both value and transaction count. The Shinkin Central Bank Research Institute did this arithmetic from the METI data and I am borrowing it.
| Method | 2018 value | 2025 value | 2018 transactions | 2025 transactions | 2018 average | 2025 average |
|---|---|---|---|---|---|---|
| Credit card | ¥66.7tn | ¥134.6tn | 10.1bn | 22.8bn | ¥6,629 | ¥5,895 |
| Debit card | ¥1.3tn | ¥5.5tn | 0.24bn | 1.37bn | ¥5,578 | ¥4,047 |
| Electronic money | ¥5.5tn | ¥6.0tn | 5.9bn | 5.8bn | ¥936 | ¥1,035 |
| Code payment | ¥0.2tn | ¥16.6tn | 0.05bn | 13.5bn | ¥3,017 | ¥1,229 |
Read the transaction column twice. Code payments went from 50 million transactions a year to 13.5 billion. That is a factor of 270 in seven years.
Electronic money went from 5.9 billion transactions to 5.8 billion. It did not grow. It shrank slightly while the market around it roughly doubled.
The worse interface won
This is the part I keep turning over.
Electronic money in Japan means Suica, PASMO, iD, Edy, WAON, nanaco. At the moment of payment it is close to perfect. The card is already in your wallet or your phone case. You do not unlock anything. You do not open anything. You do not wait for a network. You hold it near the reader for about half a second and a chime tells you it worked. I have watched people pay for a coffee without breaking a sentence.
Code payment is worse at every one of those steps. Unlock the phone. Find the app. Wait for the code to render, which needs a connection. Either present it to a scanner or point your camera at a printed sheet and type the amount in yourself, then show the merchant your own screen so they can verify a number you entered. On a bad connection it fails in front of a queue.
By any reasonable read of the payer's experience, the tap is better. The tap lost by a factor of more than two on transaction volume.
It lost because the payer's interface was never the binding constraint. The merchant's was.
Accepting electronic money means a certified reader. That is hardware, an installation, a contract, a settlement account and a per transaction fee negotiated by someone who has done this before. For a ramen shop with four seats it is a capital decision and a paperwork afternoon.
Accepting code payment, for most of the growth period, meant printing a piece of paper and putting it in a stand on the counter. The scanner was the customer's own phone, which they had already bought and were already carrying. The onboarding cost on the merchant side went to approximately zero, and merchant coverage is the thing that decides whether a payment method is usable at all.
So the method with the worse moment of payment got to be present in far more moments.
The same effect, in the credit card row
There is a smaller version of this in the credit column, and it points the same way.
Credit card value doubled, but the average transaction fell from ¥6,629 to ¥5,895. Cards did not get used for bigger things. They got used for smaller things, which is what happens when contactless removes the signature and the PIN pad from a ¥600 purchase. Take out three seconds and one moment of social friction and a payment method walks into a category it was previously too heavy for.
Nothing about the card changed. The step count at the counter changed.
What I take from this
I spend most of my working life on the thing that appears on the screen, so I have a bias to declare. The lesson here cuts against my own interests, which is usually a sign it is worth keeping.
When a market has two sides, the side with the higher cost of adoption decides the outcome. Japanese electronic money spent fifteen years polishing the half second the consumer experiences, and lost to a printed QR code that solved a purchasing decision for a shop owner. If you are building anything that needs a counterparty to say yes, count the steps on their side before you count the steps on yours.
That holds until the experience gets bad enough to break the task, and only until then. Code payment is annoying. It is not broken. Above that line, the better experience is a preference, and preferences lose to availability.
The last one is about the metric rather than the market. The cashless number went up five points and got reported as fifteen because a committee fixed a genuine flaw in a formula. The fix was correct. The comparison people will draw from it is not. When you change how you count something, the number stops being evidence about the thing and becomes evidence about the change, and it takes about a year for anyone to notice.
Japan is often described as slow to go cashless. On the international indicator it went from 13.2% to 46.3% in fifteen years and hit its government target ahead of schedule. That is not slow. What is interesting is not whether it moved. It is which of the two interfaces moved it, and that the answer is the one I would not have picked.