Every team has someone whose absence pattern is suspicious. Every absence policy is designed around that person. This is why most absence policies are bad.
The person you are designing for is not the outlier. It is the parent whose child has chickenpox, the colleague recovering from surgery, the person managing a chronic condition well enough that you have never noticed. Build for them and the outlier is handled anyway, because outliers show up clearly against a well-measured baseline.
Measure frequency, not duration
The instinct is to count days lost. It is the wrong measure. Someone off for three weeks with a broken leg costs you three weeks of cover, once, predictably. Someone off for a single day, eleven times a year, at no notice, is operationally far more disruptive and shows up as a smaller number.
The standard tool here is the Bradford Factor: S² × D, where S is the number of separate absence spells and D is the total days. It deliberately weights fragmentation heavily.
One absence of ten days scores 10. Ten absences of one day score 1,000.
Use it as a trigger for a conversation, never as an automatic disciplinary threshold. The formula does not know that someone had four separate hospital appointments, and a policy that fires automatically on a number will eventually fire on someone with a disability. That is a discrimination claim, and it will not be a close one.
Have a return-to-work conversation every time
Every absence, however short. Five minutes, informal, not a disciplinary meeting.
Three things it achieves:
- It confirms the person is actually fit to be back. People return early out of guilt more often than they stay away out of laziness.
- It surfaces causes you can fix. Recurring back pain in a warehouse role is not a sickness problem, it is an ergonomics problem wearing a sickness costume.
- It is the single most effective deterrent to casual absence - not because it is threatening, but because it makes absence visible and personal rather than a text message into a void.
The consistency matters more than the content. If the conversation only happens for people you already suspect, it is a disciplinary process pretending to be a welfare one, and everybody can tell.
Make the reporting process trivially easy
A process that requires ringing a manager who may not answer, at 06:00, produces two bad outcomes: people who report late, and people who come in ill because reporting was harder than attending.
One channel, always available, that reaches whoever is covering. Record the time it was logged. That record protects the employee at least as often as it protects you.
Separate the three causes before you respond
Unplanned absence has three broadly different roots and they need different responses:
Genuine short-term illness. Nothing to manage. Cover it, welcome them back, move on. Treating this as a performance issue is how you build a team that comes to work with norovirus.
A pattern with a cause outside work. Caring responsibilities, a health condition, transport, a second job. The response is a conversation about adjustments - a different shift pattern, a later start, formalising the flexibility that is already happening informally. This is where most "problem" absence actually lives, and it is usually solvable.
A pattern with a cause inside work. Absence concentrated on particular shifts, particular tasks, or after particular colleagues are on the rota is not an absence problem. It is a signal, and it is pointing at something. Look before you discipline.
Only after all three are excluded is it a conduct issue - and by then you will have documentation showing you looked properly, which is exactly what you need if it does become formal.
Watch the seasonal shape
Absence has a shape. It rises in winter, spikes after long weekends, and clusters around school holidays. Knowing your own pattern lets you plan cover instead of scrambling for it.
If you have twelve months of data, look at absence by week of the year. Most operations find two or three predictable peaks they have been treating as bad luck every single year.
What good looks like
- Every absence logged the same way, through the same channel, with a timestamp.
- Every return followed by a short conversation.
- Frequency measured and reviewed monthly, with triggers that start conversations rather than processes.
- Adjustments offered before conduct is questioned.
- Cover planned around the seasonal shape rather than improvised.
None of that is soft. It is what a defensible absence process looks like - and it happens to also be the version that keeps people who would otherwise have left.