Managing overtime

Overtime goes wrong in two directions and most companies only defend against one. The first is uncontrolled cost: hours worked without anyone deciding they should be, discovered at the end of the month when they are already owed. The second, and the more damaging one, is unpaid work: people staying late routinely and either not claiming it or being quietly discouraged from doing so. That one carries legal exposure in most jurisdictions and it corrodes trust faster than almost anything else.

A process that only tightens approval solves the first and worsens the second, because the hours do not disappear, they just stop being recorded. The design goal is both at once: nothing should be worked without a decision, and nothing worked should go unrecorded. Those are compatible, but only if approval happens before the work rather than after it, which is the single most important thing in this guide.

Approval after the fact is not approval, it is a bill. The manager is choosing between paying for hours already worked and telling someone their evening did not count, and there is no good answer in that position. Asking first changes the question to whether the work is necessary, whether it can wait, and whether someone else has capacity, which is the decision that was supposed to be made all along.

This has to be lightweight or it will be routed around. A request naming the day, the expected hours and the reason is enough, and it should reach the person who owns the budget for that team. Two cases need a route of their own: overtime that could not be foreseen, such as an incident, which needs a fast retrospective path rather than a fiction; and overtime that is effectively permanent, which is not an overtime problem at all but a staffing one that the approval record will make visible.

The rate is set by a combination of local law, the contract and company policy, and the order matters: the law sets a floor that the contract cannot go below. Premiums for nights, weekends and public holidays are common, and in several countries mandatory at defined multiples. Some jurisdictions cap weekly hours outright, some require a rest period between shifts, and some exclude certain roles from overtime entitlement altogether. Assuming a single rule across several countries is how a multi country payroll ends up wrong in one of them.

Time off in lieu is a legitimate alternative where local law permits it, and it needs the same discipline as pay: recorded as a balance, with an expiry, and taken. Unbanked lieu time is an unrecorded liability that surfaces when someone leaves, which is the worst moment to find it. Whichever you use, decide it before the hours are worked. Choosing after the fact makes it look like the cheaper option was picked once the cost was known.

A request is a forecast. Someone who asks to work four extra hours may work four, or two, or six. Paying the requested figure rather than the recorded one is the most common overtime error, and it is systematic rather than random: over a year it is a meaningful sum, and it is indefensible in an audit because the record contradicts the payment.

The rule that resolves it is simple and worth stating explicitly in policy. Approval authorises up to a number of hours; payment follows the hours actually recorded, capped at what was approved. Anything worked beyond the approved amount is a conversation, not an automatic payment, and anything approved but not worked is not paid. That way the approval controls cost, the record controls accuracy, and neither is asked to do the other one’s job.

Individual overtime requests are operational; the pattern across them is management information, and it is usually ignored. Overtime concentrated in one team, or on the same two people every month, is not a cost problem to be squeezed. It is a staffing gap, a process bottleneck or a single point of failure, and cutting the overtime without fixing the cause simply moves the failure somewhere less visible.

Review it monthly at team level, and look for three things: who, how often, and why. Repeated overtime on one person is a burnout risk and a knowledge concentration risk at the same time. Overtime at the same point in every cycle is a process design problem, usually fixable. And overtime that appears in the weeks after someone leaves is the true cost of a vacancy, which is worth knowing when the replacement headcount is being debated.