Writing an Attendance Policy Somebody Will Follow

TL;DR

• The core decision: whether the policy is written for the rare disciplinary case or for the ordinary Tuesday when somebody can't get in.
• When doing nothing is right: when people know what to do when they can't come in, and managers handle it consistently without consulting anything.
• What has to be true: somebody can answer, from the document, who to tell and by when, in under a minute.
• How the options split: by how much discretion the manager keeps and how automatic the consequences are.
• Decision rule: answer the everyday questions in the first half. Everything formal belongs at the back, where it will be read once a year.
• Outcome to expect: a shorter document people actually use, and fewer situations where two managers do opposite things.

Written for the Case That Almost Never Happens

Somebody sits down to write an attendance policy. They look at a few examples, and the examples are remarkably consistent: definitions, then a points or occurrence scheme, then a ladder of stages, then the appeal process. Comprehensive, defensible, and built entirely around the situation where somebody's attendance has become a disciplinary matter.

That situation is rare. What happens constantly is that somebody wakes up and can't come in, and needs to know two things quickly: who to tell, and by when. Most policies answer those questions somewhere, usually in a short paragraph well into the document, surrounded by material about thresholds and stages that has nothing to do with the decision the person is making at six in the morning.

So the document doesn't get used. It sits on the intranet, and what actually governs behaviour is whatever each manager tells their team, which differs by manager and gets passed on informally. The organisation has a policy in the sense that a file exists, and no policy in the sense that anything consistent happens.

Recommended Tools

Best tools for Workforce Management

B BambooHR HR management system L Lattice Performance management D Deel Global payroll & EOR

View all recommended tools

The real problem isn't that attendance policies are badly written. It's that they're written for the wrong reader. The rare case needs a defensible process and that process needs to exist. The ordinary case needs a clear answer in the first thirty seconds, and putting the second thing first is most of what separates a policy people follow from one nobody opens.

When You Genuinely Do Not Need to Act Yet

Your current setup is genuinely fine. People know who to tell and when, managers handle similar situations similarly, and nobody has had to escalate anything. A small, stable team with a consistent manager often has this without any written document, and writing one changes nothing except the number of files.

Friction is starting to show. A new joiner has asked what to do when they're ill, or two managers have handled comparable situations differently, or somebody has queried whether they were treated fairly. Each points at the same gap. The cheapest check is to ask a recent hire what they'd do tomorrow morning if they couldn't come in, and see whether the answer is confident.

It has become a real cost. You're getting inconsistent outcomes across teams, or a manager has taken a step that's now being questioned, or absence is being handled entirely by improvisation. At this point the absence of a shared standard is producing unfairness, which is a cost even where nobody complains.

The edge case that forces it. Somebody's attendance is heading towards a formal process, or an absence may be connected to a health condition or a disability, or a manager has applied a threshold to somebody whose absence carries specific protections. Rules on sick leave, protected absence, evidence and dismissal differ sharply by jurisdiction and have been changing, and a policy applied blindly can create real exposure. Have the document reviewed locally before it's used, and take advice on the specific case.

Five Questions This Reader Asks at 11pm

What should the policy actually contain? Two things in sequence. First, the ordinary case: who to tell, by when, by what method, what happens next, and what to do if it goes on longer. Then the formal machinery for the rare case, clearly marked as such. Most policies have these in the opposite order, which is why the first half of the document is irrelevant to almost everyone reading it.

Do points systems work? As a way to make attention consistent, yes. As a way to produce automatic consequences, no, and this is where they cause harm. Counting occurrences treats a series of unconnected days as worse than one long absence, and applied without judgement a scheme will eventually penalise somebody whose absence shouldn't have counted that way. If you use one, use it to trigger a conversation.

What counts as excessive? Any number you write down is arbitrary, and stating one invites people to treat it as an allowance. The more useful framing is a point at which somebody has a conversation, chosen against your own data rather than borrowed, and explicitly not a threshold at which something happens to them.

Should the policy cover lateness? Yes, briefly, and separately from absence. They're different behaviours with different causes: lateness is frequently a transport or caring problem with a practical fix, while absence usually isn't. Folding them into one count obscures both, and it produces the odd result of a person being three minutes late counting the same as a missed day.

How much discretion should managers have? More than a strict scheme allows and less than none. The workable split is that the process is fixed and the judgement is local: when a conversation happens is the same everywhere, what comes out of it depends on the situation. That gives you consistency where people can see it and flexibility where it matters.

Three Honest Categories the Approaches Split Into

Nothing written, handled by managers. No document, each manager applies their own judgement. It's right for genuinely small teams with a stable manager who knows everybody, and it has the advantage that judgement gets applied: a long illness and a pattern get handled differently without anybody codifying the distinction. It fails on consistency and on defensibility. Two people in similar situations get different outcomes depending on their manager, nobody can point to a standard, and when a case does become formal there's no record of what was said or agreed. It also puts every new manager in the position of inventing a policy privately.

A short statement of expectations with manager discretion. A page covering the ordinary case, with what happens afterwards left to judgement. It's right for most organisations under a certain size, because it answers the questions people actually have while avoiding machinery you don't need. It's also the version most likely to be read. It fails at the point where something has to escalate, since there's no defined route, and it fails where managers vary a lot, because the discretion produces exactly the inconsistency the document was meant to reduce.

A structured scheme with defined thresholds and stages. Points or occurrences accumulate, and defined steps follow. It's right at scale, in unionised settings where a negotiated standard exists, and wherever inconsistency has already caused a problem. It gives managers cover for conversations they'd otherwise avoid and makes outcomes predictable. It fails when the steps run automatically. A scheme applied without judgement will eventually advance somebody whose absence relates to a health condition or carries specific protection, and the fact that it was automatic is not a defence.

Five Diagnostic Questions You Can Self-Assess Against

Ask a recent joiner what they'd do tomorrow if they couldn't come in. If the answer is hesitant, or includes checking with a colleague, your policy isn't doing the one job that matters most often. This takes two minutes and is the most direct test available.

Time how long it takes to find the reporting instruction in your own document. Open it and look. If it's past the first screen, or buried under definitions, then the document is organised for the reader who almost never exists rather than the one who reads it at six in the morning.

Ask two managers the same scenario. Give them an identical situation and compare the answers. Wide variation tells you the real policy is whatever each manager believes, and that no document currently in existence has changed that.

Check whether anybody has ever been through the formal stages. If nobody has, and the scheme has existed for years, it's decorative machinery occupying most of your document. If several people have, look at whether they had anything in common, because a scheme that catches the same category of person repeatedly is telling you something.

Look at whether the threshold has become a target. Where a stated number of absences is permitted before something happens, some people will treat it as an entitlement, and usage will cluster just below it. That clustering is visible in your own data and it's the clearest evidence that a published threshold changes behaviour in ways nobody intended.

Five Attendance Policy Structures, Reviewed

No written policy at all

Nothing documented, handled by managers as it arises. It earns its place in small organisations with stable management, where everybody knows the expectation and the manager knows the people. Judgement is applied case by case, which is genuinely better than any rule for the hardest situations, and nothing has to be maintained.

Where it falls short is the moment anything is questioned. There's no standard to point at, no record of what was said, and no way to show that two people were treated comparably. It also leaves every manager to construct their own approach privately, which means the organisation has as many attendance policies as it has managers and no way to see the variation.

It's defensible as a deliberate choice while you're small. It stops being defensible the moment somebody is treated in a way they want to challenge.

The cheap halfway step is worth knowing, because most organisations in this position think their only options are nothing or a full policy. Write down the reporting instructions alone, on half a page, and leave everything else to judgement. That fixes the question people actually have, costs an hour, and commits you to no machinery. Almost all the practical benefit of having a policy comes from that half page rather than from anything that follows it.

A short statement of expectations

A page: who to tell, by when, how, what happens next, and a note that persistent problems will be discussed. It earns its place by being read, which no other structure on this list reliably achieves. It answers the ordinary case directly and leaves the rare one to judgement, which is the correct proportion for most organisations.

It falls short at escalation. When something does need to become formal, there's no defined path, so it gets improvised under pressure, which is where errors happen. It also relies on managers exercising judgement consistently, and where they don't, a short policy provides nothing to correct them with.

It's the right starting point for most organisations, provided somebody knows what the escalation route is even if it isn't printed in the same document.

One addition makes it considerably more durable without lengthening it. Say who to ask when the document doesn't cover something, by role rather than by name. Short policies generate edge cases by design, and a named route for the uncovered situation is what stops each manager inventing an answer privately, which is the failure mode that quietly rebuilds the inconsistency you were trying to avoid.

A points or occurrence system with thresholds

Absences accumulate points or occurrences, with defined levels triggering defined responses. It earns its place on consistency and on manager confidence. Everybody is measured the same way, and a manager who'd otherwise avoid an awkward conversation has a neutral reason to have one that isn't a personal judgement.

It falls short in the arithmetic and in automatic application. Counting occurrences rather than days means four single days score worse than one long absence, which may be right for coverage and is a strange thing to say about a person. Worse, a scheme that advances people automatically will eventually process somebody whose absence relates to a health condition or carries specific protection, and requirements around that differ sharply by jurisdiction. Have the design reviewed locally before use, and build in a stop.

The safe version treats a threshold as the trigger for a conversation, with what follows depending entirely on what the conversation finds.

There's a design detail that changes how these schemes behave over time, which is whether points expire. A scheme where occurrences roll off after a period treats attendance as a current state, which is usually what you want. One where they accumulate indefinitely means somebody's record gets steadily worse regardless of recent behaviour, and the people it catches are those with long service rather than those with a current problem. If you build one, decide the rolling period deliberately rather than inheriting it.

Trigger-based review without automatic consequences

A threshold prompts a review meeting, and nothing else is predetermined. It earns its place by keeping the benefit of a scheme, which is consistent attention, while removing the part that causes harm, which is consequence without judgement. It's also easier to defend, since each outcome rests on a reasoned decision rather than on a count.

It falls short when the review itself becomes a formality. If the meeting always reaches the same conclusion, you've rebuilt an automatic scheme with extra steps, and people will recognise that quickly. It also asks more of managers, who now have to reach a judgement rather than apply a rule, and some will find that harder than the process it replaced.

This is the structure that suits most organisations that have outgrown a short statement, and the thing to protect is the genuineness of the review.

A simple check tells you whether it's still genuine. Look at the last several reviews and see whether any of them ended with no further action. If none did, the meeting is a stage rather than a review, whatever the document calls it, and both managers and staff will have worked that out well before anybody in HR does.

A full disciplinary ladder with fixed stages

Defined stages with defined outcomes, up to and including dismissal. It earns its place where it's genuinely required: unionised environments with a negotiated procedure, regulated settings, or organisations where a documented, uniform process is a condition of operating. In those contexts, predictability is the point.

It falls short as the everyday framework for everybody else, because it makes the rare case the organising principle of the whole document. It's also slow, it's adversarial by design, and it tends to be written in language that's precise for a hearing and useless for somebody trying to find out who to call. And it interacts poorly with protected absence unless somebody has deliberately built in the places where the ladder must stop.

Keep it, where it's needed. Put it at the back, and write the front of the document for the person who just wants to know what to do this morning. Where a negotiated procedure exists, remember that it sets the floor for the formal part and says nothing about the rest of the document. Nothing in an agreed disciplinary route prevents you from putting a clear page of reporting instructions in front of it, and doing so is rarely contentious, because it helps the people the agreement exists to protect.

The Decision Table

The second row is where most policies should start and almost never do. A single page that answers who to tell and by when, placed first, does more for actual behaviour than any amount of machinery behind it.

What the Policy Has to Answer on an Ordinary Tuesday

The test of a policy is whether somebody can get what they need from it quickly, while feeling unwell, before work. Most fail that test, not because the information is missing but because it's placed behind material aimed at a different reader.

The second row is the one that produces the most avoidable friction. As soon as reasonably practicable means nothing to somebody deciding whether to ring at five in the morning or wait until seven, and the result is that every team develops its own unwritten convention. A stated time removes the whole problem.

The last row is worth answering honestly rather than reassuringly. People will assume absence is recorded and counted, because it usually is, and an evasive answer confirms the suspicion while adding distrust. Saying plainly what's recorded, what prompts a conversation, and what doesn't happen automatically is better received than a soothing sentence that nobody believes. It also removes a common source of unnecessary attendance. People who are unsure what counts against them tend to come in when they should not, which spreads illness in public-facing and care settings and produces a worse outcome than the absence would have.

Where Discretion Belongs and Where It Does Not

Every attendance policy is a settlement between consistency and judgement, and most get the split in the wrong place.

Consistency belongs in the process. When a conversation happens, who has it, what gets recorded, and what route exists if something escalates should be identical across the organisation. That's the part people compare, and variation there is experienced as unfairness even when each individual decision was sensible. It's also the part that's cheap to standardise, because it doesn't require anybody to reach a judgement.

Judgement belongs in the outcome. What follows a conversation depends on things no policy can anticipate: whether there's an underlying health matter, whether the pattern has an explanation, whether something about the job is contributing. A scheme that fixes the outcome removes the ability to respond to any of that, and it removes it precisely in the cases where responding matters most.

The failure mode in both directions is worth naming. Too much discretion produces inconsistency that staff can see and nobody can defend: the same situation handled two ways in two teams, with the difference visible to everybody. Too little produces a process that runs over the top of the facts, and the people it damages are usually the ones with a genuine reason.

There's one place where discretion has to be bounded rather than free, and that's anywhere absence may be connected to a health condition, a disability, or another protected reason. Those situations carry specific requirements that differ considerably by jurisdiction, and a manager exercising ordinary judgement may not recognise that the ordinary rules have stopped applying. The policy should say plainly what makes a case one to stop and escalate, and managers should know they won't be criticised for stopping. That last part matters more than the instruction itself. A manager who suspects that pausing will be read as failing to follow the process will carry on, which is precisely the situation the stop was designed to prevent, so it needs saying explicitly and it needs to be true the first time somebody does it.

What to Put in Writing

The policy is the artefact, and a few things around it determine whether it works.

The third row is the one that protects managers and the one most often missing. Somebody applying a policy in good faith needs to know which situations take them outside it, and needs to be confident that stopping is the right call rather than a failure to follow process.

Questions to Ask Before You Commit

On the ordinary case. Can somebody find the reporting instruction in thirty seconds? A bad answer is that it's in the document.

On consistency. Would two managers handle the same scenario the same way? A bad answer is that they should.

On triggers. Does a threshold prompt a conversation or a consequence? A bad answer is that it depends on the manager.

On the stop. What makes a case one to pause and escalate, and do managers know? A bad answer is that HR would spot it.

On the number. Is the threshold published, and has usage clustered below it? A bad answer is that nobody would do that.

On review. Who has checked this against local requirements, and when? A bad answer is that it was adapted from a template.

What Getting This Wrong Costs

The first cost is that the policy simply doesn't operate. A document written for the disciplinary case and organised accordingly won't be read by the person deciding whether to ring in, so what governs behaviour is whatever each manager has told their team. The organisation believes it has a consistent standard and has instead got a collection of local conventions, and nobody discovers the difference until two cases are compared.

The second cost is inconsistency that becomes visible at the worst moment. Where managers improvise, similar situations get different outcomes, and people compare notes. That's experienced as unfairness regardless of whether each individual decision was reasonable, and it's particularly damaging because it suggests that what happens to you depends on who you report to rather than on what you did.

The third cost is the one with real exposure attached. A scheme that advances automatically will eventually be applied to somebody whose absence relates to a health condition, a disability or another protected reason, and in many jurisdictions applying it blindly in that situation creates a problem the organisation then has to answer for. The automatic nature of it, which was the selling point, is what makes it indefensible, because nobody exercised the judgement that would have stopped it.

So before you write anything, work out which problem you actually have. An information problem means people don't know what to do when they can't come in, and the fix is a page with the reporting instructions at the top. A consistency problem means managers handle things differently, and the fix is a standard process with local judgement on outcomes. A defensibility problem means you need a documented route because something is heading towards formal, and the fix is a proper procedure reviewed locally. Most organisations have the first and write for the third.

When You Are Ready to Go Further

None of this needs a system. It needs the reporting instructions written plainly and placed first, a stated point at which somebody has a conversation, and clarity about what makes a case one to stop and escalate.

The step beyond the document is the thing that determines whether any of it works, which is whether managers can actually have the conversation. A policy that triggers a discussion is only as good as the discussion, and most attendance problems are resolved or not resolved in that exchange rather than by anything written down.

HROpsLab publishes independent comparison work across HR tooling, applicant tracking and payroll. We sell nothing, we take no vendor money, and we publish no paid placements. If the next step is looking at what your current systems record and prompt, our comparison work is one place to start.

Frequently Asked Questions

Most people open an attendance policy once, early in the morning, wanting one answer. Write that part first.