Working Hour Cap

Software work has long carried the habit of piling hours in front of a deadline — staying until midnight for the two weeks before a release and coming in on the weekend. It looks like individual choice and it is usually the product of a schedule. Four months of work planned into two months gets made up out of people's time.

The problem was that those hours never stayed a one-time event. A release rescued by a crunch sets the pattern for the next plan, and somewhere in between people leave the team with burnout. Quality falls in work done while exhausted, and fixing it consumes the time the crunch was supposed to save.

Labor law already regulated working hours, but the exceptions were wide. The structure of 40 standard hours plus 12 overtime hours was the same, while hours worked on holidays were treated as counting separately from those 12 — an interpretation that allowed up to 68 hours in a week. A 2018 amendment to the Labor Standards Act closed it by counting holiday work inside overtime and fixing the ceiling at 52. The reasoning was that a ceiling which can be raised by agreement is not a ceiling.

The arithmetic is simple. Forty statutory hours a week plus up to 12 overtime hours agreed between employer and employees gives a ceiling of 52, and hours worked on a holiday count inside that 12. Five things matter in practice.

1. Scope. It applies to workplaces with five or more regular employees. A workplace of four is outside the ceiling, and falls under it in the month a fifth person is hired, so a growing company crosses into the rule without doing anything except hiring.
2. What counts as a week. Seven calendar days. Working 60 hours this week and 44 next week to average out is a violation unless one of the schemes below has been formally adopted — the week is the unit, not the month, which is why an intense stretch has to be arranged in advance rather than explained afterward.
3. Who is liable. Responsibility sits with the employer, not the person who worked the hours. The penalty runs to two years' imprisonment or a fine of up to roughly $15,000.
4. The flexibility mechanisms. A flexible working hours scheme averages across a unit period; a selective working hours scheme measures the total inside a settlement period, normally one month and up to three months for research and development work. The latter fits a schedule that concentrates before a release and thins afterward. For disasters and sudden emergencies, a separately authorized extension exists with approval from the labour ministry.
5. Records. The employer has to keep track of who worked how many hours. Without records there is no way for the company to show the ceiling was respected if a dispute arises, and the absence of a record tends to be read in the employee's favor.

One point trips companies up regularly. A contract that folds overtime pay into the annual salary in advance does not lift the hour ceiling. How compensation is paid and how many hours may be worked are separate questions, so money added up front is not grounds for going past 52.

The rule reaches planning in three places.

1. Building a schedule. Because each person has a weekly ceiling, a schedule becomes people multiplied by weeks. In the meeting that decides how much goes into a sprint, the cap acts as the real capacity limit and "make it up with overtime" drops out of the arithmetic. What's left is cutting scope, adding people, or moving the date.
2. Designing incident response. With an on-call rotation, the company has to decide how standby time and actual response time are counted. Many pair it with a rule that shortens the next day's hours for someone who worked two hours at 3am.
3. Setting up working arrangements. Letting people choose their own start and end times through flexible hours requires formally adopting the selective scheme above and fixing a settlement period. Told only to organize their own time, some people exceed the weekly ceiling and the company carries the liability.

The rule comes up most often in the meeting that sets a release date, because asking whether a schedule is achievable turns out to be the same question as asking whether it is lawful. A team that has internalized that tends to argue about scope earlier, when the argument is still cheap.

"This schedule doesn't fit inside 52 hours. Either scope comes out or the date moves."

Overtime isn't available as a variable in the calculation.

"Let's put next month inside a selective-scheme settlement period."

Measuring the total across a month rather than week by week.

"You handled the incident overnight, so tomorrow afternoon should come off."

Counting response time as working time and adjusting for it.

Pagedeck ran into this directly in the quarter it built enterprise features. The first plan had four engineers on it for six weeks; Theo's calculation showed that inside a 52-hour week the same work needed eight.

Grace chose to push two detailed items of permission management to the next quarter rather than extend the timeline further, which kept the release date the customer had been given. Expecting incidents to cluster in release week, she split the on-call rotation two ways and wrote down in advance the rule reducing the following day's hours by the time spent responding.

The first assumption is that consent makes longer hours acceptable. If the person wants to, why block it — the objection is natural. But this ceiling is not a condition the parties can waive. Hours beyond it are a violation even with written consent, and the employer is liable. The premise is that if consent could lift the cap, the side with less bargaining power would always consent and the cap would disappear. Hours someone stayed late voluntarily count the same way if the company knew and allowed it.

The second is thinking remote work falls outside it. Nobody is in the office, so the hours look uncounted. But the rule attaches to hours worked, not to where they were worked. Only the method of record-keeping changes, and remote records tend to be blurrier, which puts the employer at a disadvantage in any later dispute. Time spent replying to evening messages accumulates into the same problem, and a chat history is exactly the kind of record that gets read back later.