42.68 Million People Visited Japan. When the Interface Failed, 68% Reached for Their Own Phone
42,683,600 people visited Japan in 2025. That is a record, up 15.8% on 2024, which was itself a record at 36,870,148. Australia passed one million visitors for the first time, becoming the seventh market to do so, and twenty markets set an all time annual high. The numbers are JNTO's, published on 21 January 2026.
They spent ¥9.45 trillion, up 16.4%, an average of ¥229,000 each.
The Japan Tourism Agency also runs a survey of departing visitors asking what went wrong on the trip. The latest results, published on 28 April 2026, cover 4,110 travellers surveyed at Narita, Haneda, Kansai, Fukuoka and New Chitose between 18 November 2025 and 13 January 2026.
One line in it is the reason I am writing this. When visitors could not communicate with staff, 68% dealt with it by using an ICT tool. When multilingual signage was missing or unclear, 57% did the same.
They took out their own phone.
What people said went wrong
| Problem | Share |
|---|---|
| Too few rubbish bins | 17.2% |
| Communicating with staff | 15.4% |
| Crowding at destinations | 12.9% |
| Using transport | 11.3% |
| Too little or unclear multilingual signage | 10.9% |
| Using a credit or debit card | 10.7% |
| Immigration procedures | 10.3% |
| Nothing went wrong | 43.7% |
Rubbish bins top the list for the second year running, which is a real and slightly comic feature of visiting Japan and which I am going to leave alone because it is not a design problem I can help with.
The rest of the list is one problem wearing five hats.
Communication trouble happened in restaurants (38%) and at railway and bus stations (27%). Signage trouble happened inside railway stations in cities (30%) and in restaurants in rural areas (23%). Transport trouble was mostly non shinkansen railways in cities (73%) and buses in the regions (42%), and what specifically went wrong was searching for the route to a destination (45%) and working out where to get on and off (38%). Card trouble was restaurants again (32%). Immigration trouble was the wait, at 58%.
Restaurants, stations, buses. The two places a traveller has to transact with a stranger under time pressure.
The number that went the wrong way
43.7% said nothing went wrong. The year before, that figure was 51.1%.
So satisfaction on this measure fell by 7.4 points in the year that arrivals rose 15.8%. That is not a mystery and it is not a scandal. It is capacity. The same signage, the same staff and the same platform width, meeting six million more people.
It does tell you where the country is on the curve, though. Japan is past the point where more visitors is straightforwardly good news, and the constraint is not hotel rooms. It is the small interactions that scale badly.
The front end Japan does not own
Here is the part I find genuinely striking as a designer.
Two thirds of the people who hit a language wall in Japan resolved it with a device Japan did not build, running software Japan does not control, over a connection sold by a company that may or may not be Japanese.
Google Translate and Google Maps are the actual interface layer of a ¥9.45 trillion export industry.
Every yen of that ¥9.45 trillion passes through a moment where somebody works out which platform, which bus, which dish, which price. In more than half of the moments where that fails, the recovery is the visitor's phone.
I want to be fair about this. It works. It works well enough that 43.7% of people had no trouble at all, and a camera pointed at a menu is a genuinely good solution to a menu in a language you cannot read. Nobody should feel bad that travellers use their phones.
But if you are designing anything in Japan that a foreign visitor will touch, this changes what you are designing. You are not designing the primary interface. You are designing the input to somebody else's.
What that implies, concretely
Once you accept that the phone is doing the work, a set of ordinary design decisions become obvious, and most of them are cheap.
Signs need to be photographable. A translation app reads a sign through a camera, at an angle, in bad light, held by someone walking. Low contrast type, text over a photograph, decorative fonts, vertical text mixed with horizontal, and text baked into an image at low resolution all degrade badly. Flat high contrast type on a plain background survives. This is not an accessibility nicety, it is now a functional requirement, and it happens to be the same requirement as accessibility.
Names have to survive translation. If a station exit is called 東口 and the app returns "East Exit" but the sign in the building says "Exit 3", the traveller is holding two correct answers that do not match. The single highest value thing a station can do is make sure the machine readable name and the printed name are the same string. That costs nothing and it is wrong almost everywhere.
Restaurants came top for both communication and card trouble, so the menu is the piece of paper in Japan where a small change does the most work. A printed menu with clear type, a price next to every item and a card symbol on the door removes most of a conversation that neither party wants to have.
Route search is the failure that actually strands people, at 45%. Transport operators publish timetable data at wildly different quality. If your bus route is not in the data, the visitor's phone does not know your bus exists, and they will take a taxi and remember your town as difficult.
And the immigration wait, at 58% of the immigration complaints, is a queue design problem in a building the state controls entirely. It is the first interface in the country and it is the one Japan has the most direct authority over.
Why I think this matters beyond tourism
There is a version of this article that ends by saying Japan should build a better translation app, and I do not believe that. It should not. That fight is over and it was lost to a company with a fifteen year head start on the data.
The opportunity is the other way round. The phone is a fixed, universal, free piece of infrastructure that 42.68 million people are already carrying and already using. Designing for it is mostly about not breaking it.
Most of what is on the list above costs a sign, a data file or a menu reprint. None of it requires an app, a budget cycle or a committee. The reason it has not happened is that nobody has framed the visitor's phone as part of the system, so nobody is responsible for whether it works.
42.68 million people, ¥9.45 trillion, and the front end is Google. That is worth someone in Japan writing down as a design brief rather than a complaint.