61% of Japanese Companies Still Run Legacy Systems. 12% Wrote It Into a Plan

In 2018 the Ministry of Economy, Trade and Industry published a report warning that if Japanese companies did not modernise their systems, the country would lose up to ¥12 trillion a year from 2025. It called this the 2025 Digital Cliff. The phrase went everywhere. Every vendor in Japan put it in a slide deck.

2025 has been and gone. In May 2025 METI published the report of the Legacy Systems Modernization Committee, run with the Digital Agency and IPA, which sat from July 2024 to March 2025 to find out where things actually stood. It came with a survey of around 4,000 user and vendor companies, fielded between 17 December 2024 and 14 February 2025, with 799 responses.

Two numbers from it have stayed with me.

61% of user companies still hold legacy systems. Among large companies it is 74%.

12% of user companies mention introducing or replacing a large system anywhere in their medium term management plan.

Nobody fell off the cliff

Start with the framing, because the framing was wrong and the wrongness matters.

A cliff is an event. You reach a date, something breaks, and afterwards the world is different. That is a useful story for getting a budget approved and a terrible description of what old software does to a company.

Old systems do not fail on a date. They get more expensive to change, in small increments, forever. A request that would have taken two weeks takes six because three people have to be consulted about a table nobody documented. A feature ships without the part that touched the mainframe. A product decision quietly becomes whatever the system already supports.

None of that shows up as an outage. It shows up as a company that is slower than its competitors and cannot say why.

So 2025 arrived and nothing visibly happened, which is the worst possible outcome for the argument. Seven years of warning produced a deadline that passed without incident, and now the people who spent those years asking for modernisation budget have to explain why the thing they predicted did not occur. It did occur. It just does not look like anything.

The 12% is the real finding

The 61% figure gets quoted. I think 12% is the one that explains it.

A medium term management plan in a Japanese company is a public document. It is what the company tells its board, its bank, its shareholders and its staff that it intends to do over the next three to five years. It is where the commitments live.

If replacing a core system is not in that document, it is not a commitment. It is a line item in an IT budget that any competent finance director can defer for one more year, and deferring it costs nothing this year, and it costs nothing next year, and the person who eventually pays for twenty years of deferral will not be the person who deferred it.

The committee report is blunt about the mechanism. Information system departments are consumed by keeping the current systems running. They have limited contact with business departments. Management treats systems as a cost line, which puts the IT department socially below the departments it serves, so system problems never get promoted into management problems. Nothing in that loop produces a plan.

88% of Japanese companies have not written down what they intend to do about the software their business runs on. That is not a technology gap. That is a governance one.

What separates the companies where it works

The useful half of the survey is the cross tabulation, because it says what the companies making progress have in common. The report tested these with a chi squared test and reports only pairs with a significant association, which is more rigour than most industry surveys bother with.

Where the information system department and management actually share information about the systems, 71% of companies have documented and visualised their system specifications. Where they do not, 66% have not. Same split on building in house: 87% of the companies with that information flow do some development themselves.

Where a company has appointed a CxO, a named executive owning the function, 75% have visualised their specifications. Where there is no CxO, 52% have not. On progress, 41% of companies with a CxO say modernisation is going smoothly. Among those without one, 71% say it has stalled.

The report also finds that companies not modernising are the ones that outsource every phase of operation, maintenance and development to a vendor. Companies that have documented their own systems tend to bring work back in house, and companies that have brought work in house tend to modernise.

You can read that chain in either direction and it still holds. Knowing what you have and being able to change it are the same capability described from two ends.

The part I recognise from the outside

I do product design, so I meet this from a specific angle. I get called in to improve an interface and I find out in week two that the interface is downstream of a system nobody present can describe.

The symptom is always the same. Somebody can tell me what the screen does. Nobody can tell me why it does that, or what happens if it stops. The answer, when it comes, comes from a vendor, in a meeting, three weeks later, and it is a number rather than an explanation.

That is what the 61% looks like at the surface. A form with nine fields where four would do, because those nine map to nine columns, and the columns cannot move, and so the form cannot move, and so the user does nine fields worth of work forever.

Nobody chose that. It is a decision that got made once, by somebody who has left, and then stopped being a decision and became a fact of the building.

The thing I want people to notice is that the ¥12 trillion is not paid in a crisis. It is paid in those four extra fields, several million times a day, across a whole economy, by people who have no idea they are paying it.

Where I would start

Not with a replacement project. The survey is clear that large modernisation efforts run for years, and that a plan revised mid flight runs for years more.

I would start with the two things the data actually links to progress, both of which are cheap.

Write down what you have. Not an architecture diagram for a consultancy, an honest inventory of which systems exist, what each one is for, who understands it and what would happen if it stopped. Most companies discover during this exercise that at least one answer to "who understands it" is a person who retires next year.

Then give it to somebody with authority. The CxO correlation in this survey is not really about the title. It is about whether a system problem has a route into a room where budget gets decided. Without that route, the information system department can be completely right about the risk for a decade and never be heard, which is roughly what the last decade looks like.

METI says it will build maturity indicators and self diagnosis tools so companies can assess their own IT assets. That is sensible and I hope it works. But an indicator only helps a company that has already decided to look. The 12% number says most of them have not decided anything yet, and that is the part no tool fixes.