Working hours on a building site are decided at the tender

I think most of the debate in Japanese construction about the 2024 overtime limits has been held in the wrong department. Contractors have improved attendance systems, issued guidance on leaving site offices earlier and sent engineers on time-management courses. Each of those helps a little. None of them changes the fact that the hours an engineer will work are largely fixed when the company agrees a construction period with a client. My position, after six years of trying to bring Asakage's site engineers within the limits, is that working hours are a commercial decision first, that the head of people should sign the schedule assumptions in every large tender, and that a contractor's count of hours should include the partner-firm crews whose days its schedule also decides.

This is not an argument that HR measures are pointless, or that clients alone are to blame. Contractors offered short periods to take work for decades, and clients learnt to expect them. It is an argument about where the decision sits. If a company accepts a schedule that needs Saturday work and late evenings, it has decided its people's hours, whatever its attendance policy says.

Where the hours went when we counted them

When I returned to Asakage in 2012, site engineers recorded their own hours in books initialled by their site managers. The books showed an average of about 40 hours of overtime a month. That year we began comparing them with gate-entry records and log-ons in site offices. The real figure was closer to 80. Nobody had falsified anything in a way they would have recognised as dishonest. Engineers left out the hour after the last inspection, the evening spent on photographs for the client's monthly report and the Saturday spent checking drawings at home. Their managers had done the same when they were young.

Making hours visible changed the executive committee's discussion more than any argument I made. It also taught me that hours move. When we pressed site offices to cut evening work, drawings and inspection records began arriving at head office late at night. When we closed Saturdays, some engineers worked Sundays at home. When we built a document support centre in Nagoya to take paperwork off sites, I failed to count its hours for six months, and some of the overtime we had removed from sites had simply relocated there. A contractor that measures only the record it is trying to improve will usually see that record improve.

The lesson I draw is that compliance should be measured across every place hours can go: the site office, the document centre, home log-ons and the partner crews. The figure we now report to our audit and supervisory committee each quarter covers all four, and it is less flattering than the site-office figure alone.

A schedule page in every large tender

From April 2020 we required every bid above ¥1 billion to be priced on a schedule that closes the site eight days in every four weeks, with a page showing the engineers, the hours and the partner crews the schedule assumes. I sign that page, together with the head of the construction division. If the client asks for a shorter period, the page is revised and signed again, and the cost of additional staff or a second shift is priced rather than absorbed.

Our head of building construction argued against the page, and his objections were fair. Competitors would offer shorter periods, clients would find us rigid, and estimators would resent HR appearing in a process they had run for decades. He was partly right. Over two years we declined or lost nine tenders worth about ¥62 billion where the client would not accept the period. Accepted schedules lengthened by an average of 6%.

What we gained was a way to have the argument at the point where it could still be settled. A client who wants a warehouse handed over before its peak season can usually have it, if it pays for a second shift or a larger site team. A client who wants the date without the cost is asking our engineers and the partner crews to pay the difference in their evenings. The page makes that visible to the client, to our sales team and to our board. Most clients, once they saw it, accepted a longer period or paid for the staff. Several told us they had assumed the old schedules cost nothing.

The subcontractor's day belongs in the same count

The overtime limits apply to each employer's own workers. Asakage's site engineers are our employees; the formwork carpenters, rebar workers and scaffolders on our sites mostly are not. It would be possible to comply with the law on our own payroll while a partner firm's crew works the Saturday we no longer work. I do not think a general contractor can treat that as someone else's problem, because our schedule set their day as well as ours.

The difficulty is that many small partner firms still pay their craftsmen by the day. When we closed sites on Saturdays, those workers lost a day's pay each week, close to a tenth of their income in some months. The limits were meant to protect people like them, and our first version of compliance cut their pay. We raised the labour rates in our subcontracts by about 4% and asked partner firms to move craftsmen to monthly pay, but the increase reached them months after the Saturday closures, and two formwork firms lost young workers to competitors in the meantime. If I were starting again, I would settle the partners' pay before closing a single Saturday.

Some contractors argue that pricing longer schedules simply hands work to firms that will not. In the short term that is true, and our order intake in 2021 showed it. Over five years I think it is less true than it looks. Since April 2024 every contractor's engineers have been subject to the same limits, and the firms that kept bidding short schedules have had to find the hours somewhere. Our voluntary turnover among site engineers fell from 7.8% to 4.1% over the same period, and graduate applications for site roles have risen. Those are the people who will build the schedules of the next decade.

I would ask any contractor's board to read three documents. First, the schedule assumptions behind the five largest bids of the past year, and the names of whoever signed them. Second, the overtime figure for site engineers measured from gate and log-on records rather than self-report, including the document staff and home working. Third, the pay arrangements of the partner crews on its largest sites, as those crews actually receive them. If the three do not agree with the company's published statement on working hours, then the statement is not where the hours are being decided.