The 1.4-minute Shinkansen and what Japanese customers expect

Japanese customers don't measure your uptime against your competitors. They measure it against the train they took to work. The Tokaido Shinkansen's average delay per train in fiscal 2024 was 1.4 minutes, typhoons and earthquakes included. That one number explains a scene most foreign software companies in Japan eventually live through: the outage, the SLA credit paid exactly as promised, and a customer who is still unhappy, because what they wanted was the cause and the fix.

What does 1.4 minutes actually count?

Everything. JR Central's "average delay per train" is the mean delay across every train it ran, including delays caused by natural disasters. In fiscal 2024 that was 1.4 minutes, across 383 trains a day (extra services included) carrying 460,000 passengers a day. Since the line opened in 1964, no passenger on board has been killed or injured in a train accident.

The year-by-year series shows what the metric really measures. Fiscal 2016: 24 seconds. 2018: 0.9 minutes. 2019: 0.2. 2022: 1.1. 2023: 1.6. 2024: 1.4. The number jumps in years with typhoons, earthquakes and heavy rain. So it measures how fast the line comes back after nature stops it, as much as it measures timetable precision. And it still stays around a minute.

To get a feel for the scale: 383 trains times 1.4 minutes is about 540 minutes, so the whole line accumulates roughly nine hours of delay a day. Whether that sounds like a lot depends on where you sit. The fastest Tokyo to Shin-Osaka run is 2 hours 21 minutes, so per train it is about 1% of the journey.

Typhoons and earthquakes included, a Shinkansen train runs about a minute late on average

Tokaido Shinkansen, average delay per train in service (minutes, by fiscal year, including delays caused by natural disasters). JR Central annual reports and company site; only years confirmed in published material

• FY2016
  24 sec
• FY2018typhoon and quake year
  0.9 min
• FY2019
  0.2 min
• FY2022
  1.1 min
• FY2023
  1.6 min
• FY2024383 trains, 460,000 riders a day
  1.4 min

This is the yardstick your customers ride every day. They watch what you do when you're late more than the SLA number.

One comparison. Deutsche Bahn counts a stop as on time if the train is less than six minutes late. On that basis, 52.4% of its long-distance trains were on time in September 2026. The units differ, but the gap in what a daily passenger experiences comes through.

So Japanese trains never run late?

They do, constantly. Japan's transport ministry publishes a "delay visualization" for 45 Tokyo-area commuter lines. In fiscal 2018, the average line issued delay certificates, which are handed out for delays of five minutes or more, on 11.7 of every 20 weekdays. The Tokyo Metro Chiyoda Line did so on 19.2 days. Nearly every day. The largest cause of delays under ten minutes was simply people taking too long to get on and off: 48.3%.

So what is Japanese punctuality? My reading: it isn't the absence of delay. It's that every delay gets measured, announced and documented. Five minutes late and you can collect a certificate to show your boss. The platform board shows an estimated restart time. Failures still happen. Every one of them gets counted.

The rules for failure are written before the failure

Article 289 of JR Central's passenger rules says that if a limited express, which includes Shinkansen services, arrives at its destination two or more hours late, the full express surcharge is refunded. The clause makes no distinction by cause. Article 290-3 adds that the claim can be made "whether or not the cause is attributable to the company," and that no other claim can be made at all. A typhoon delay of two hours gets you the surcharge back. One hour fifty-nine does not. Nothing beyond that is on the table.

Set that beside the European Union's rail passenger rights and the difference in character is clear. In the EU you get 25% of the fare back for a delay of 60 minutes or more and 50% for 120 or more, but the operator owes nothing if the cause was an "extraordinary circumstance" such as severe weather or a natural disaster. The Japanese rule has no sliding scale and no cause test, and a hard ceiling instead. Which is better for passengers depends on the day. The Japanese line is simpler, and everyone can recite it. Passengers know the two-hour line. Station staff know it. Nobody negotiates at the counter on the day, because the line was drawn in advance.

The other direction gets the same treatment. In November 2017 the Tsukuba Express issued an apology because a train left Minami-Nagareyama station 20 seconds early. Nobody had complained. The following year JR West did the same for a train that left Notogawa 25 seconds early. Overseas this became a story about Japanese fussiness. I read it differently. Any departure from the promised timetable, in either direction, is something you publish and apologize for. In that system, leaving a deviation unexplained would be the strange act.

What "stable" means to a Japanese customer

Here is the point. When a Japanese enterprise customer judges how reliable your SaaS is, the baseline for "things that don't stop" was set on a station platform, years before your demo. Nobody does this consciously. But that is where the yardstick lives.

This is why 99.9% uptime, on its own, rarely wins a deal. What separates vendors is what happens after a stop: how fast you announce it and give a restart estimate, whether you explain the cause, and whether you put the fix in writing. Japanese customers ask for an incident report because they grew up in a system where every failure gets accounted for. They will take the refund if the rules say so. It won't settle the matter.

Concretely, the first email on the night of an outage should contain what the station board contains. What is happening, who is affected, the workaround for now, the estimated restart, and when the next update will come. You don't need the cause yet. Writing "we don't know yet" earns more trust than silence. A few days later comes the report: timeline, cause, temporary and permanent fixes, and what will prevent a repeat. The IT department of a large Japanese company has received dozens of documents in exactly this shape from its domestic vendors. If yours doesn't match the shape, they worry before they read the content.

The point where head office usually pushes back is disclosing the cause. Legal at home may argue that naming the cause admits liability. For a Japanese customer it reads the other way. A report with no cause in it means either you still don't know or you are hiding it. They hear the railway announce the reason for every delay, "vehicle inspection," "a person on the tracks," every day of their lives.

I wrote before that you should open the repair center before you scale. For SaaS, the repair center is an incident process and a person who can explain root causes in Japanese. It is also part of why newcomers struggle to move customers in the country with the most century-old companies. An incumbent has already shown, many times, what it does when things break.

This is an observation, and I haven't got numbers behind it. Smaller and younger Japanese customers are often more relaxed. But if you sell to large Japanese companies, I'd draft the first email you will send on the night of an outage before you draft the contract.

Back to 1.4 minutes. Behind that figure are a few thousand people who inspected track on a typhoon night and got the first train out the next morning. When a Japanese customer looks at a foreign product, I think the look is roughly that one. Who decided what on the night of the outage, and how did you explain it the next morning. Your SLA table doesn't have a column for that.

---