Late minutes belong to a shift
Lateness is meaningless unless you say late for what. Measure it against one company start time and every person on a different shift is wrong by definition, all month, in a report someone will act on.
PeopleMuster counts late minutes against the assigned shift window.
The nine o'clock problem
Many attendance systems hold a single start time for the company. A punch after it is late, and the minutes between are the late minutes.
That works for exactly one arrangement: everybody starts together. The moment a second shift exists, the arithmetic silently breaks.
Someone on a midday start looks three hours late every day. Their on-time rate collapses. Nobody reads the report closely enough to notice why, and the number gets quoted in a review.
What an assigned shift changes
A shift is a named window with a start and an end. Morning might run ten to seven. A mid shift might run twelve to nine. Each person is assigned one, from a start date, with a record of who assigned it.
Lateness then has a denominator. Late means after the start of the window that person was assigned on that date.
The assignment history matters as much as the assignment. Someone moved from the morning shift to the mid shift in the third week should not have the first two weeks recalculated against the new window. The history is what stops that.
Where the number still misleads
A correct denominator does not make late minutes a good measure of anything else. It measures arrival against a window. Nothing more.
Two people with identical late minutes can be in completely different situations. One has a commute that fails twice a month. One leaves early too. The column does not know.
Aggregate late minutes are worse. A department total tells you almost nothing without the shift mix underneath it, and shift mix is exactly what a monthly export flattens.
The small design choices that follow
Once shifts drive lateness, a few other things have to follow or the number drifts back into nonsense.
A person with no assigned shift needs a defined answer, not a default. In PeopleMuster the self-serve check-in card refuses to guess: no shift, no self-serve punch, and the person is routed to a manual punch request that a human reviews.
Holidays need to be in the same calculation. A company holiday that attendance ignores produces a day of absences for a day nobody was meant to work.
- Lateness measured against the assigned window, not a global time.
- Assignment history kept, so past months are not recalculated.
- No shift means no automatic verdict, and a reviewable request instead.
- The holiday calendar respected by the same attendance logic.
Grace, and where to put it
Most teams want some tolerance, and the argument is usually about where it lives rather than whether it exists.
Put it in the shift and the record stays honest. A window that opens a little before the start time is a stated rule everyone can read.
Put it in the report instead and you get a second, informal rule that nobody wrote down. Five minutes is fine until a manager decides it is not, and now two people are being held to different standards.
Either way, write it down. An unwritten tolerance is the thing people discover by being told off for something that was fine last week.
What to check before reporting lateness to anyone
Three checks, and they take a few minutes.
First, everyone has a shift for the whole period. A gap in assignment produces days measured against nothing.
Second, the correction queue is empty. Hours still waiting on review are hours missing from whatever you are about to show someone.
Third, the holidays are on the calendar. A company holiday the system does not know about turns into a day of absences for a day nobody worked.
Reports built before those three are done get argued with, and the argument is usually right.
The shift nobody assigned
Every company has people the shift model does not fit neatly. Part-time staff, people on a rotating pattern, someone covering two roles across different hours.
The wrong answer is to give them a nominal shift so the report has something to measure against. That produces numbers which look valid and describe nobody.
The better answer is to leave them unassigned and accept the manual path. Their attendance becomes a reviewed record rather than an automatic verdict, which is a truer description of the arrangement anyway.
It costs a few minutes a month and it keeps everyone else's numbers meaningful. A denominator invented for convenience contaminates the aggregate it sits inside.
Questions people ask
What happens when someone has no shift assigned?
Self-serve check-in will not run without one. The card says so and routes the person to a manual punch request, which an administrator reviews. Nothing is guessed.
Does changing someone's shift rewrite their history?
No. Shift assignments carry a start date and the previous assignments stay in the assignment history, so earlier months keep the window they were actually measured against.
Is a low on-time rate a performance problem?
Not on its own. It measures arrival against a window and knows nothing about workload, commute or output. Treat it as a prompt for a conversation, never as the conclusion.