Shifts & Overtime: Hours You Can Defend on the Payroll
Overtime is the line on the payroll that nobody can defend and everybody pays. Where the hours come from, why derived overtime beats claimed overtime, and the approval step that stops a shift pattern quietly becoming a permanent cost.
Ask a Kenyan employer what they spent on overtime last quarter and most can give you a figure. Ask which department it came from, whether it was approved before or after it was worked, and whether the underlying hours were recorded at the time or reconstructed at month end, and the answers get thinner with each question.
That is not carelessness. It is what happens when overtime is a claim rather than a derivation — when the number comes from a person telling you what they worked, weeks after they worked it, and nobody is in a position to say otherwise.
A shift is a definition, not a habit
Everything downstream depends on this. Before you can say somebody worked extra, you need a written statement of what normal was — and in most organizations that statement exists only as custom. "The warehouse starts at seven" is a habit. A shift is a record: a start, an end, a break, and how much lateness is tolerated before it counts.
| What a shift holds | What it decides |
|---|---|
| Start and end time | The expected hours for the day — the baseline everything measures against |
| Break minutes | Whether nine hours on site is nine hours worked or eight |
| Grace minutes | Whether arriving four minutes late is lateness or life |
| Whether it is active | That retiring a shift pattern does not rewrite the history of people who worked it |
Grace minutes deserve a moment. Without a defined tolerance, every attendance system becomes a source of daily friction over two- and three-minute arrivals — and supervisors respond by ignoring the data entirely, which costs you the whole record. A stated grace period is what makes lateness meaningful when it does appear.
Derived, not claimed
Here is the mechanism that changes the conversation. Instead of asking staff to submit overtime, compare what attendance actually recorded against the shift the employee was assigned, and create a pending overtime record for the excess.
-
Attendance is captured as it happens
Clock-in and clock-out against the day, with the verification your sites use. Nothing here is about overtime yet — it is just what happened.
-
The day is compared to the assigned shift
Worked hours against expected hours, allowing for the break. Where the excess clears a minimum threshold, an overtime record is raised for that excess.
-
It arrives as pending, not approved
This is the important part. Derivation produces a candidate, not an entitlement. HR or the manager approves, rejects or amends, with the decision, the decider and any comment recorded.
-
Approved hours reach payroll as an input
Overtime hours and unpaid leave days are resolved for the period and feed the payroll calculation, rather than being typed in by whoever is preparing the run.
Claimed overtime is a negotiation with somebody's memory. Derived overtime is a question about a specific Tuesday, asked while the answer still exists.
The threshold matters more than it looks. Without a minimum, every employee who stays eleven minutes late generates a record, the approval queue fills with noise, and managers approve in bulk without reading — which is functionally the same as having no control at all.
The approval is where the money is
Most organizations put their control effort into measuring hours accurately and then approve whatever the measurement produces. That is backwards. Measurement tells you what happened; approval is where somebody decides whether it should have.
Approving after the fact, in bulk
- The work is done and the moral case is already made
- Nobody can distinguish necessary from habitual
- A department discovers overtime is easier than hiring
- The cost appears at month end as one number
- Refusing a claim becomes a personal conflict
Approving per record, weekly
- Each entry is a specific day with a specific reason
- Patterns surface while they are still small
- The question "why is this recurring?" gets asked
- The cost is visible as it accumulates
- Rejection is about one Tuesday, not about a person
The pattern worth naming: overtime that recurs in the same team every week is not overtime. It is understaffing being financed at a premium, and it is almost always cheaper to fix with a hire or a schedule change. That comparison is impossible to make when the cost arrives as one monthly figure with no shape to it.
What the hours are worth is your policy, not a formula
This is where an honest article has to slow down. How overtime should be paid — the multiplier, the treatment of rest days and public holidays, which categories of employee it applies to — is a matter of Kenyan employment law and your own contracts of employment, and it varies by sector and agreement.
We state no rates or multipliers anywhere in this article deliberately. Confirm the treatment that applies to your employees with a qualified employment adviser, then configure the pay components to match. What the system contributes is the hours, approved, attributable to a named day — which is the input the policy needs and the part most employers cannot produce.
Public holidays and the working week
Both are configurable per organization: public holidays, because holidays that only exist in someone's head produce leave and attendance calculations that quietly disagree with the calendar, and the working week itself, so a six-day operation counts leave days on its own calendar rather than an assumed Monday-to-Friday one. Set the working week before the first leave request rather than after a disputed balance.
What we do and do not do
What AWRA OpsHub does today
- Shift definitions with start, end, break minutes, grace minutes and an active flag.
- Shift schedules and assignments, so an employee's expected hours on a given day are a record rather than an assumption.
- Overtime derived from attendance against the assigned shift, raised as pending for the excess above a minimum threshold, and skipping days that already have a record.
- An approval decision on every overtime record — approver, timestamp and comment retained.
- Approved overtime hours and unpaid leave days resolved into the payroll run as inputs rather than typed figures.
- Configurable public holidays, and attendance regularization for the days the capture genuinely got it wrong.
What it does not do
- No overtime rate logic in the derivation. It produces hours. What those hours are worth — multipliers, rest-day and holiday treatment — is your policy expressed in pay components, not a formula we impose.
- No automatic rostering or shift optimisation. Schedules are built by you; nothing suggests who should work when, or balances a rota.
- No cap or block on overtime. The system will not stop an employee working, or a manager approving, beyond a limit you have set.
- A Saturday–Sunday weekend assumption in leave calculations, which is wrong for organizations whose week runs differently.
The first line is deliberate rather than a gap. An overtime multiplier hard-coded by a vendor is a legal opinion issued by software, and employment law is not ours to state on your behalf.
Getting it running without a fight
- Write the shifts down as they actually are, including the break, before changing anything. Half of what a shift system reveals is that the written pattern and the real one already differ.
- Set a grace period you can defend, and say it out loud. Silent tolerance is worse than none, because it is applied inconsistently.
- Set the overtime threshold high enough that the approval queue is readable. A queue nobody reads is a control nobody has.
- Approve weekly, not monthly. The reason for a Tuesday is retrievable on Friday and lost by the 30th.
- Read the recurring patterns quarterly and ask the staffing question, because that is where the actual saving is.
Our take
Define the shifts properly, let overtime be derived from attendance rather than claimed from memory, and approve weekly at a threshold that keeps the queue readable. Then treat recurring overtime in a team as a staffing signal rather than a payroll line — that is where the money actually is, and it only becomes visible once the hours have a shape.
See attendance and shift management
Shift definitions with breaks and grace, assignments, attendance capture with regularization, derived overtime with approvals, and hours that reach payroll as inputs.
Explore attendanceFrequently asked questions
What rate should overtime be paid at in Kenya?
We deliberately state no rates or multipliers. Overtime treatment — the multiplier, how rest days and public holidays are handled, and which employee categories it applies to — is governed by Kenyan employment law and your own contracts, and it varies by sector and by agreement. Confirm what applies to your employees with a qualified employment adviser, then express it in your pay components. What the system supplies is approved hours attributable to specific days, which is the input that policy needs.
Where do the overtime hours come from?
From attendance compared against the shift the employee was assigned for that day. Where the worked hours exceed expected hours by more than a minimum threshold, a pending overtime record is created for the excess — days that already carry a record are skipped, so re-running does not duplicate. Nothing is payable until somebody approves it, and the approver, the time and any comment are retained on the record.
Can we stop overtime exceeding a monthly limit?
Not by a system rule — there is no cap that blocks further hours or refuses an approval beyond a threshold you set. The control is the approval step itself, which is why approving weekly rather than monthly matters: a limit enforced by a manager reading five records on Friday is real, while a limit checked at month end is a report about something that already happened.
What if attendance captured the day wrongly?
That is what regularization is for — a recorded correction to an attendance record, rather than an edit that leaves no trace. It matters because the alternative is staff losing trust in a system that occasionally gets their day wrong with no route to fix it, and once that trust goes they revert to keeping their own book, which is where you started.
Does the system build the roster for us?
No. Shifts, schedules and assignments are yours to define and allocate; nothing suggests who should work when or balances a rota across a team. If automated rostering is a hard requirement — as it can be for hospitals, hotels and security firms — treat it as a distinct capability to scope rather than assuming it comes with attendance.
