Leave: three questions about the day somebody leaves
Nothing here changes how anybody books a day off, sees a balance, or gets an approval. All three questions are about one moment — somebody’s last day — and about one figure on one payslip. Nobody has left since the tracker began, which is why this is the only part of the leave system with no historical case to check an answer against.
How to answer
Questions 1 and 2 are for HR. Both already have our proposal filled in, with the arithmetic shown. The fastest useful answer is “yes, that’s what we do” — and if it is not, the single most useful thing you can send back is one real last-day date and the figure you would have given. That is exactly how the joiner rule was recovered, and it beats any description.
Question 3 is for Anwar, and it is a yes or no.
Nothing here needs anyone to read code, and nothing here is urgent. The Leave module is being built now and none of these three blocks it.
01What is already settled
Annual Leave stays granted whole on 1 January. Nothing changes during the year for the thirty-one people who are not leaving. Only on somebody’s last day does the platform work out what they had actually earned and settle the difference.
The joiner rule needs no confirmation. HR already prorates for people who join mid-year using a rule that has never been written down anywhere. Reverse-engineered from the three 2026 joiners, it is exact:
The rule HR already applies
Months earned = 12 − the month you joined. The joining month itself is free.
Entitlement = 14 × months ÷ 12, rounded to the nearest half day.
| Hired | Months | Entitlement | |
|---|---|---|---|
| Batoro Aji | 26 Mar 2026 | 9 | 10.5 |
| Carina | 20 May 2026 | 7 | 8.0 |
| Andre | 13 Jul 2026 | 5 | 6.0 |
Those three match what HR calculated by hand, so the rule is reproduced rather than invented. The three questions below are all about the other end of the year, where there is no such precedent.
02Question 1 · What does a leaver earn?
Does the leaving month count, or not?
for HRThe obvious mirror of the joiner rule is to count the month of the last day, including it. That is deliberately the opposite of the joiner rule, which excludes the joining month — because excluding the partial month at both ends would charge somebody twice for two half-months they did work.
Here is what each reading produces, on the standard 14-day entitlement:
| Last day | A. Count the leaving month | B. Exclude it, mirroring the joiner exactly |
|---|---|---|
| 31 January | 1 month → 1.0 day | 0 months → 0 days |
| 15 April | 4 → 4.5 | 3 → 3.5 |
| 30 June | 6 → 7.0 | 5 → 6.0 |
| 30 September | 9 → 10.5 | 8 → 9.5 |
| 31 December | 12 → 14.0 | 11 → 13.0 |
The last row is why we think it is A. Under B, somebody who worked the entire year from January to December earned 13 of their 14 days and owes the company a day back on the way out. That cannot be right.
- we propose
- A — the leaving month counts.
- we need
- Whether that is what you do today, or whether you have been doing something else. If neither column matches, send one last-day date and the figure you would give.
03Question 2 · Joining and leaving in one year
This one was not on anybody’s list. It came out of writing question 1 down and then checking it against the joiner rule.
The two rules overlap, and together they overpay
gapThe joiner rule and the leaver rule were each written for somebody who was here at one end of the year. Applied together to a person who was here for neither end, they produce a figure larger than what that person was ever granted.
Worked example. Somebody joins 20 May 2027 and leaves 30 September 2027.
| Granted on their first day, under the joiner rule | 12 − 5 = 7 months | 8.0 days |
| Earned at exit, under rule A read literally | month 9 | 10.5 days |
They would be owed 2.5 days they were never granted in the first place, encashed on their final payslip. Not intended by anybody — it is what happens when two rules that were each correct in isolation are put in the same year.
One formula covers every case:
The proposal
Months earned = (the month you left, or December if you are still here) − (the month you joined, or zero if you joined in an earlier year).
It reduces to both existing rules at the extremes, so nothing anybody currently calculates changes:
| Case | Months | Days |
|---|---|---|
| Andre — hired 13 Jul 2026, still here | 12 − 7 = 5 | 6.0 — matches HR’s figure |
| Long-serving person leaving 30 Sep | 9 − 0 = 9 | 10.5 |
| Joined 20 May, left 30 Sep, same year | 9 − 5 = 4 | 4.5 |
- we need
- Has this ever actually happened? If somebody has joined and left inside one year, what they were paid is worth more to us than any formula we can propose.
04Question 3 · When the deduction is bigger than the payslip
Is the remainder written off, or pursued?
for AnwarSomebody who took more leave than they earned has the difference deducted from their final payslip. The proposal on the table since July is that this is capped at the final net pay, with any remainder written off — pursuing a former employee for the balance costs more than it recovers.
It was proposed and not objected to, which is not the same as agreed. It is a rule that gives money away, so it should be a decision rather than something that happened by default.
The useful thing we found while checking it
On current figures, the cap almost never fires.
The deduction is priced at the gross rate of pay — 12 × monthly gross ÷ 260 — which is the same figure used for encashment, because it is the same thing in the opposite direction. Against a Singapore net pay after employee CPF, the worst realistic case is somebody taking their entire year’s leave in January and resigning on the 31st. Even then the deduction reaches roughly 76% of a month’s gross against a net of about 80%.
The cap only starts to bite above about 19 days of entitlement. Today exactly one person in the company clears that — Yashwin, on six days of service award — and only in the scenario where he takes all twenty days in January and resigns on the 31st.
So the question is not whether to write off a lot of money. It is:
- we need
- Do you agree the remainder is written off rather than pursued — accepting that on today’s numbers this writes off nothing, and that it exists for the case where the final payslip already carries another deduction (unpaid leave days, or a claim being recovered) alongside the leave settlement?
- worth knowing
- The 20% CPF figure and the 260-day divisor are among the things Finance is being asked to confirm separately. If either moves, the 19-day threshold moves with it. The rule does not change — only the estimate of how often it fires.
The reverse case is built either way. A leaver who took less than they earned is owed the difference, encashed on the final payslip. Building only the deduction would produce a rule that is exact when it costs the employee and silent when it costs the company, and that is not what is being proposed.
05If these go unanswered
Stated plainly, so that silence is a choice rather than an accident.
- Question 1 gets built as A — the leaving month counts. It is the only reading that does not tell somebody who worked all twelve months that they earned eleven.
- Question 2 gets built as the single formula, because the alternative pays people days they were never granted.
- Question 3 gets built as a cap, since it is already recorded as proposed and not objected to.
Everything else in the Leave module is unblocked and is being built now. These three change one figure on one payslip, on a day that has not yet happened to anybody — which is exactly why they are worth answering before it does.