Leave management policy guide.

A leave policy is a small set of decisions with large consequences: how much leave people get, when the leave year runs, what happens to unused days, who approves requests, and how public holidays interact with the counting. Most disputes trace back to one of these being undecided or ambiguous rather than wrong. The goal of the policy document is not elegance, it is that two reasonable people reading it reach the same number for the same situation, every time.

Write the policy as rules a system could execute, because eventually one should. "Managers may approve reasonable requests" is a sentiment. "Requests route to the line manager, who sees team coverage before deciding, and requests longer than two weeks also route to HR" is a rule. Every ambiguity you leave in the text becomes a precedent someone sets by accident on a busy afternoon, and precedents are much harder to unwind than sentences.

Start from the statutory floor in each country where people work, then decide what the company adds on top. Entitlement is also shaped by working pattern: part-time staff accrue pro rata, and joiners and leavers accrue for the fraction of the year they are actually employed. These calculations are simple individually and error-prone at scale, which is why they should be computed from policy, tenure and working pattern rather than typed into a tracker by hand.

Decide early whether entitlement grows with service, whether different groups get different amounts, and how you will handle the awkward cases: mid-year changes of hours, unpaid leave, and moves between countries. None of these are rare. A policy that only covers the full-time, full-year, single-country case covers perhaps half of a real workforce, and it is the other half that generates all of the disputes.

Unused leave at year end has three possible fates: it lapses, it carries over, or it is paid out. Unlimited carry over sounds generous and quietly builds a liability that grows every year, while strict lapsing pushes everyone into a December scramble that empties the office at once. The common middle ground is a small capped carry over with a use-by date, for example a few days that must be taken early in the new year. The exact numbers matter less than their being explicit and applied evenly.

Whatever rule you choose, apply it automatically and record every movement. Carry over applied by hand is applied inconsistently, and the inconsistency is discovered in the least convenient forum: a resignation negotiation, where untaken leave becomes money. A ledger of accruals, taken days and adjustments turns year end from an argument into an arithmetic check, because every balance can show its working.

The default flow is simple: the request goes to the line manager, the decision lands back on the record, and the balance moves. Complexity should be added only where risk justifies it: longer absences might need a second approver, and some teams may need minimum staffing rules. What matters more than the shape of the flow is its consistency. When approvals happen in chat for some people and formally for others, the informal channel always wins, and the tracker stops matching reality within a month.

Give approvers a deadline and a fallback. A request that sits unanswered for two weeks is its own kind of unfairness, because the person cannot book anything. A simple service standard, answered within a few working days with escalation to the next manager if not, keeps the system humane. And every decision, including declines, should carry a recorded reason, because the pattern of declines is worth seeing too.

Most bad approvals are not bad judgement, they are blind judgement: the manager who says yes without knowing two others on the team are already off that week. The single most valuable feature of any leave process is showing the approver who else is away for the requested dates at the moment of decision. It converts approval from a favour into an operational decision, and it protects the team from the resentment that builds when the same people always end up covering.

Coverage visibility also changes employee behaviour for the better. When people can see the team calendar before requesting, they self-organise around obvious clashes, and the requests that arrive are mostly approvable. That is a far better experience than requesting blind and being declined for a reason that was invisible when the request was made, and it spreads the popular weeks more evenly without anyone having to enforce it.

A leave request that spans a public holiday should not consume a day of entitlement for it, which means the day count depends on the calendar of the country where the person works. In a single-country company this is a detail; in a distributed one it is a monthly source of error, because each country keeps its own holidays and they move year to year. The rule to write down is simple: day counts exclude the public holidays of the employee’s own country, automatically, from whatever calendar applies that year.

Half-day conventions, regional holidays inside one country and religious holidays whose dates are confirmed late all complicate the picture further. The practical answer is the same in every case: the calendar is data, kept per country and per year, and the counting is computed from it. The moment holiday arithmetic is done in someone’s head, the same request costs different people different amounts, and someone eventually notices.

The Bradford factor is a score that weights frequent short absences more heavily than rare long ones, on the reasoning that many separate absences are more disruptive than one extended one. It multiplies the square of the number of absence spells by the total days absent, so ten single days score far higher than one ten-day absence. Its value is as an early signal: a rising score surfaces a pattern that is easy to miss when each individual absence looks unremarkable on its own.

Its danger is being used as a verdict. A high score says nothing about why the absences happened: it flags the parent of a chronically ill child and the disengaged employee identically, and using thresholds to trigger automatic sanctions punishes people for their circumstances. Treat the score as a prompt for a supportive conversation, never as a disciplinary trigger on its own, and be transparent with staff about how it is calculated and what it is used for.