Buy construction capacity before you buy land

For most of my career the scarce resource in Japanese housing was land. Developers competed for sites, carried land banks measured in years of launches and judged their own strength by the size of those banks. Construction was something you bought when you were ready. If one general contractor's price was too high, another would bid. I learned the business on that assumption, and in 2023 it stopped being true for us.

When a tender comes back empty

In the year to March 2024, nine Mikurane condominium projects in the Tokyo area received no acceptable bid. For some, no contractor offered at all; for others, the lowest price exceeded our budget by more than 15 per cent. Our tender prices for mid-rise condominiums were 31 per cent higher than three years earlier. The overtime limits that began to apply to construction work in April 2024 were expected to lengthen programmes further, and contractors were choosing among projects rather than competing for them.

The idle sites cost us ¥3.4 billion a year in holding and finance costs. Worse, they distorted every other decision. Our land team was still buying at the pace the launch plan required, because the plan assumed construction would follow. Our sales team was promising launches that our construction department could not schedule. And our board was being shown a land bank as a measure of future revenue when part of it could not be built at prices buyers would pay.

The industry's usual response was to wait for costs to settle, raise prices where the market allowed and tender again. I did not think waiting would work. Demand for construction labour was rising from several directions at once, from public works to factories and data centres, while the construction workforce was shrinking and ageing. Nothing in the evidence pointed to a return to the earlier balance.

One design, negotiated once

In early 2024 I proposed that we reverse the sequence. We would secure construction capacity first, through four-year framework agreements with three mid-sized general contractors for about fifteen buildings a year in total, and buy only the land that capacity could build. To make the agreements worth a contractor's commitment, we offered what developers rarely give: volume they could plan crews and suppliers around, and designs they had helped to shape.

Our design team and the contractors agreed two structural grids for mid-rise buildings and a catalogue of prefabricated bathroom units, kitchens and balcony elements. Prices are set against an open-book index for materials and labour, and savings and overruns against it are shared. Each contractor knows a year ahead roughly which buildings it will build for us and can plan its subcontractors accordingly. We cut land purchases from about ¥120 billion a year to ¥75 billion to match.

The head of our condominium business opposed the plan, and his objections were sound. Launches would fall from about 4,300 homes a year to about 3,400. We would lose central Tokyo sites to competitors who were still buying hard, and we lost four in 2024. Standard designs would not suit every site or every buyer. An outside director added a different warning: depending on three contractors concentrates risk. We limited any one contractor to 40 per cent of our volume and kept the right to tender outside the framework.

What a land bank no longer tells a board

The results after two years are mixed but clear. In the year to March 2026 one tender went unawarded, against nine two years earlier. Standard buildings are completed in about 22 months against 26 for bespoke ones, and holding costs on idle sites have fallen to about ¥0.7 billion a year. Our gross margin on condominiums fell from 23.6 to 22.4 per cent, partly because we now share cost movements with contractors instead of squeezing them at tender. I regard that as the price of reliable capacity.

We also made mistakes. In the first year we applied the standard designs to every site, including a sloping one in Bunkyo where they did not fit, and lost a year redesigning. We now reserve about a fifth of our volume for bespoke designs. We also brought the contractors' site managers into the design catalogue later than we should have; they found problems our designers had missed.

The larger lesson concerns what a board is shown. A developer's land bank used to be a reasonable proxy for future revenue. It is now a list of options that are valuable only if someone can build on them at a workable price. When I present our pipeline to the board today, each project shows whether it holds a contractor slot, and land without one is reported separately. Directors who see only the size of a land bank are being shown the old scarcity.

For directors, the useful questions are simple to ask. How many projects in the pipeline have a contractor committed at an agreed price or index? How long have sites without one been waiting, and what does each month of waiting cost? What share of volume depends on a single contractor, and what happens if that contractor's order book fills with more profitable work elsewhere? And when management reports a fall in launches, is it a failure of sales or a deliberate match between land and the capacity to build on it? A board that hears clear answers will understand its pipeline far better than one shown a land bank measured in years. None of these questions needs engineering expertise. They need only that construction be treated as a strategic constraint rather than a procurement detail settled after the land committee has met.

I do not argue that land no longer matters. A poor site cannot be rescued by a good contractor. But the order of decisions has changed, and developers that keep buying land first will find themselves holding more of it than they can build, at prices that assumed construction would always be available. In an ageing country with a shrinking workforce, the people who build are the scarcer resource, and strategy should start with them.