A leave policy that survives growth
Policies written for thirty people fail at three hundred in predictable ways. Most of them are about edges nobody wrote down.
A small company runs leave on judgement. Someone asks, a manager says yes, and the balance is roughly remembered. That works until two people ask for the same week, or until someone leaves with a balance that has to be paid out, and suddenly the policy has to be exact rather than reasonable.
The first thing to make explicit is accrual. Does leave accrue monthly or land as an annual grant? Does it start on the join date or on the first of the year? A policy that does not answer this produces two different balances for two people who joined a week apart.
The second is tenure. Almost every organisation eventually gives longer-serving staff more days, and almost none of them writes down what happens in the year the threshold is crossed. Decide whether the extra days apply from the anniversary or from the start of that year, and write it somewhere the system can read.
The third is the calendar. A leave request that spans a public holiday should not consume a day of balance, which means the holiday calendar has to exist as data rather than as something the HR team knows. Weekends differ by country, too, and that difference is a frequent source of quiet miscalculation.
Carry-over is where most policies are vague on purpose, because being precise means telling people they will lose days. Being vague does not avoid that conversation; it just moves it to December and adds an argument about what was promised.