How to Decide What to Build When You Cannot Hire

If you work in technology in Japan you have seen the number many times. By 2030 the country will be short roughly 450,000 IT workers, and 790,000 in the pessimistic case. It is in every recruitment pitch and every DX consulting deck here.

The number is real. It comes from a METI projection published in April 2019, produced jointly with the Ministry of Health, Labour and Welfare and the Ministry of Education, based on a survey of 3,000 IT companies and 3,000 user companies with 2,173 responses.

What does not get quoted is the sentence just below it.

METI writes that if labour productivity in the sector rises 3.54% a year, supply and demand for IT workers are expected to balance by 2030. Not narrow. The shortfall disappears.

I have heard the 450,000 figure in meetings here more times than I can count. I have never once heard that sentence.

What the model actually says

The arithmetic is simpler than the way it gets discussed.

On the supply side, METI projected IT workers in Japan rising from 1.03 million in 2018 to 1.13 million in 2030. That is growth, in a country with a shrinking working population. The pipeline is not collapsing.

On the demand side there are three scenarios for market growth. At 4.4% a year, the high case taken from what companies themselves reported, 2030 demand reaches 1.92 million, and against 1.13 million of supply that is the 790,000 figure. At 2.7%, the middle case, the gap is 450,000. At 1%, demand barely moves.

All of it assumes labour productivity rising 0.7% a year, which is what Japan's information and communications sector actually managed through the 2010s. The same document notes the sector averaged 2.4% a year since 1995.

So 3.54% is demanding. Roughly five times the recent trend, about half again the long run average. I am not going to pretend that is easy.

But it is a different problem from the one most companies are trying to solve. Nearly everyone treats this as a hiring race. Nationally the hiring race cannot be won, because 1.13 million is 1.13 million however you distribute it. The productivity path is the other road in the same document.

What productivity means in software

This is where the conversation goes wrong, and it is a definitional problem.

When a company hears productivity it reaches for tooling. Better CI, more automation, an AI coding assistant, a low code platform for simple internal apps. All of that is real and worth doing.

Every one of them makes building faster. None of them affects whether the thing should be built.

In my experience the largest single source of wasted engineering capacity is not slow engineers. It is engineers building, correctly and on schedule, something that turned out not to be needed.

A feature requested by one loud stakeholder. A settings screen for a setting nobody changes. A fourth report duplicating three that exist, because the person who asked did not know about them. A migration built to preserve a workflow that only exists because of a system limitation removed years ago.

None of that registers as a productivity problem. Everyone involved was productive, the quality was fine, the date was met. It simply should not have been started.

Tooling does not fix that. It gets fixed upstream, by putting the decision before the sprint rather than after it, and that is design work regardless of what the person doing it is called.

The role Japan is actually missing

This is not my interpretation. It is what companies say when asked which people they lack.

IPA's DX Trends 2025 asks which of the five roles in the national digital skill standard companies are most short of. The United States and Germany both answer data scientist. Japan answers business architect, and the margin is large.

The standard defines a business architect as the person who carries an initiative from objective setting through implementation to verifying the result, coordinating everyone involved. In other words, the person who decides what is worth building and stays with it long enough to find out whether it worked.

Japan is not primarily short of people who can build. It is short of people who decide what gets built. A shortage of deciders produces exactly the pattern the rest of that report describes: effort flowing to efficiency rather than revenue, partial optimisation rather than whole systems, weak links between management, IT and business units.

Two more figures from the same report make it hard to call this only a labour market issue.

Asked whether the quality of their DX people is adequate, 3.8% of Japanese companies said yes. The US said 52.9%.

And among the more than 80% of Japanese companies reporting a shortage of DX people, 19.4% are carrying out no talent acquisition activity at all. One in five companies with the problem is doing nothing about the problem. That is not a supply constraint. It is a company that has not decided what it is looking for, which is the same failure as not deciding what to build, one level up.

What to do this year

Stop using the shortage as a reason to wait. If the plan is to execute the roadmap once you have hired, and the national projection says those people do not exist, then the plan is to not execute the roadmap.

Cut the roadmap instead, deliberately and in public. Rank by what changes for a customer, not by what was promised to whom. The bottom half of most roadmaps I see would not be missed, and every item on it is consuming capacity from the top half right now.

Put someone in the deciding seat even if you cannot hire one. It does not need to be a new person. It needs to be a named individual with both the authority to say no and the obligation to check afterwards whether it worked. Most companies already have this person and have never given them the second half.

Use outside people for deciding, not only for building. IPA found American and German companies contract specialist firms and individuals far more readily than Japanese ones, which lean on internal training and moving people between departments. Internal training is good and slow. 2030 is not far away.

Measure the thing you want to improve. Not lines of code, not story points. How long it takes from someone identifying a need to something being in front of a user, and what proportion of last year's shipped work is still in use. The second number is uncomfortable everywhere. It is also the productivity number METI's 3.54% is really asking about.

The point

The 450,000 figure is used, almost always, to explain why something cannot be done. It has become the most respectable excuse in Japanese enterprise technology.

The document it comes from does not support that use. It offers two roads, and the country has spent seven years discussing only one of them, which happens to be the road where the constraint is external and nobody has to change how decisions get made.

The other road is deciding better. Harder to buy, easier to start.

References

• METI, IT人材需給に関する調査（概要）, April 2019. Supply of 1.03 million in 2018 rising to 1.13 million in 2030 (table 2); gap of 450,000 in 2030 under the middle scenario (table 1); high scenario demand of 1.92 million giving 790,000; balance at 3.54% annual productivity growth; sector productivity of 0.7% in the 2010s and 2.4% since 1995
• IPA, DX動向2025 and the full report PDF, figures 3-2, 3-3, 3-4 and 3-5