An attendance policy people follow answers four questions and stops: what time the day starts for this person, how much lateness counts as late, who to tell when they cannot come in and by when, and what happens after that.
Policies that run to nine pages usually fail on the first question, because they state one company start time and the company has four working patterns.
Question one: what time the day starts
The expected start has to resolve per person per day, not per company. Someone on a four-day week, someone covering an evening shift and someone in another time zone all need a different answer, and the policy should say where that answer comes from rather than printing a single figure.
We resolve it in a fixed order: the shift assigned to that person for that day, then the expected start recorded for their office, then the company default. An unset value keeps the chain moving rather than stopping it, so a shift row saved with a blank start time falls through to the office rather than resolving to midnight. The important part is that the order is written down, so a disagreement about whether someone was late has a resolution path instead of two opinions.
Working patterns carry the working week with them. The weekend is stored as weekday numbers per company, seeded by region, because a Gulf entity and a UK entity in the same group do not share a weekend and neither should inherit the other's.
Question two: how late is late
Pick a grace period in minutes, apply it from that person's scheduled start, and put the number in the policy. Ours is a company setting rather than a constant, and a clock-in counts as late only once it passes the scheduled start plus the grace.
There is a second threshold worth setting and almost nobody does: the hour after which an arrival stops being a late start and becomes an exception. We hold a late-arrival cutoff, defaulting to midday, past which an automatic check-in declines rather than quietly opening a shift. Someone arriving at four in the afternoon has a reason, and the reason belongs with a manager rather than in a timestamp.
Punches get corrected, and the policy should say how. A phone goes flat, a site fence fires late, a shift starts before anyone opens the app. Our correction path has the employee propose an earlier clock-in with a written reason and a manager approve or reject it. On approval the original punch is kept alongside the corrected one, so a corrected row can never be mistaken for an untouched one months later.
Question three: reporting an unplanned absence
Name the person, the channel and the deadline. A policy that says to inform your manager as soon as possible produces three different behaviours in three teams.
Acas guidance on creating absence policies covers exactly this ground: procedures for reporting unplanned absence, handling unauthorised absence, keeping in touch during time off, and supporting a return to work. Acas also says to share the policy with the whole organisation, make it easy to follow, and train people on it. The training line is the difference between a document and a habit.
On recording, Acas asks for three fields per sickness absence: the number of absences, the length of each one, and the reason. It also advises recording disability-related and pregnancy-related absence separately.
Question four: what happens next
This is where most attendance policies overreach. Acas states plainly that there are no legal rules about how or when absence should be reviewed, and that an employer should not use review points to automatically trigger disciplinary action for absence. A threshold produces a conversation. A person decides the rest.
Acas also warns against discrimination in the use of review points, giving the example of a disabled worker who needs time off for regular medical treatment and may therefore reach a review point sooner than a colleague.
So write the policy in two halves. The first half is mechanical and identical for everyone: the grace period, the reporting deadline, the records kept. The second half is about judgement, and it should say who exercises it and on what evidence.
The records underneath
Attendance is a record-keeping obligation in most places, with the detail set per jurisdiction. The US Department of Labor's recordkeeping fact sheet requires hours worked each day and total hours worked each workweek, lets employers choose their timekeeping method as long as it is complete and accurate, and sets retention at three years for payroll records and two years for the supporting documents such as time cards and work schedules. Check the equivalent duty for every country you employ in rather than applying one retention period everywhere.
One punch in our model carries the day, the clock-in and clock-out, how it was captured, and where it came from when a site fence was involved. That last part matters for a policy you intend to enforce, because a dispute about a recorded time is nearly always a dispute about where someone was.
What to do with the policy once it exists
Read it against three real weeks of your own data before you publish it. Count how many arrivals fall inside the grace period, how many fall past the late cutoff, and how many unplanned absences were reported through the channel the draft names.
A low third number tells you the draft names a channel nobody uses, whatever it got right about grace periods. Measuring absence: the rate, the trigger, the next step covers turning those records into a rate and a threshold, and Reconciling attendance before payroll closes covers the month-end job the same records feed.
Answered here.
- What should an attendance policy contain?
- Four answers, in plain language: the expected start and finish for each working pattern, the grace period before a late arrival is recorded as late, how and by when someone reports an unplanned absence and who they tell, and what happens after a recorded absence or a run of late arrivals. Acas guidance on absence policies also covers unauthorised absence, keeping in touch, review points and return-to-work support.
- How long should a lateness grace period be?
- There is no statutory figure, so it is a company decision that should be the same for everyone on the same working pattern. Set it once, write it in the policy, and apply it from the scheduled start of that person's shift rather than from a single company-wide clock.
- Can an employer deduct pay for lateness?
- That depends on the jurisdiction and on what the employment contract permits, so take advice for each country you employ in. As a matter of process, a deduction should be visible to a named person who accepts it before the pay run closes, with the specific dates attached, rather than applied automatically.
- What attendance records do employers have to keep in the US?
- The US Department of Labor's FLSA recordkeeping fact sheet requires employers to record hours worked each day and total hours worked each workweek. Payroll records must be kept for at least three years, and the records that support wage computations, such as time cards and work schedules, for two years.