THREE WAYS TO READ ABSENCE DATA BEYOND THE ABSENCE RATE

An absence rate treats a single five day illness and five scattered days off as the same event. Three readings pull them apart, using leave records you already keep.

Emmanuel Duncan · 7 September 2026 · 6 min read

Absence usually reaches the board as one percentage. Days lost over days available, rolled up across the whole organisation, compared against last year. The figure is easy to produce because payroll and leave records already hold everything it needs.

It is also almost impossible to act on. A rate of four percent could describe an organisation where a handful of people are on long medical leave and everyone else is at their desk. It could equally describe one where two hundred people each take an unplanned day off every month. Those are different organisations with different problems, and the same number covers both.

What follows are three readings that pull them apart. Each uses the leave and attendance data you already keep. None of them requires a new system.

Reading One: Frequency Against Duration

The first split is the most useful, and it takes one afternoon.

For each person, count two things over a twelve month window. How many separate absence episodes they had, and how many days those episodes cost in total. An episode is a continuous run of absence, so a person off for five consecutive days has one episode of five days, while a person off for five scattered single days has five episodes of five days.

Those two people have identical absence rates. They are not presenting the same issue to the organisation. The five day block is usually a genuine illness, it is often notified in advance, and the team can plan around it. The five scattered days arrive without warning, each one breaks a shift or a schedule, and the disruption cost is far higher than the day count suggests.

The Bradford Factor formalises this. It multiplies the number of episodes by itself and then by the total days, so frequent short absence scores heavily and a single long absence barely registers. I use it as a sorting device rather than a judgement. It ranks the list so you know who to talk to first. It does not tell you what you will find when you talk to them, and an organisation that turns a Bradford score into an automatic disciplinary trigger will punish people with chronic conditions and people with caring responsibilities before it reaches anyone who is simply taking days.

The practical output is a scatter of your staff, episodes on one axis and days on the other. Most people sit near the origin. The interesting cases are the two tails, and they need entirely different responses.

Reading Two: Absence by Day of Week and by Month

Plot unplanned absence by day of the week. Then plot it by month.

A flat line across Monday to Friday tells you absence is being driven by health and by life. A line that spikes on Monday and Friday tells you something else is happening, and it is usually one of three things: a scheduling problem where weekends are being extended informally, a leave policy that makes formal short leave hard enough to apply for that people call in sick instead, or a supervision gap on the days when senior staff are least present.

The monthly view does similar work. In several Ghanaian organisations I have looked at, unplanned absence climbs in the weeks around school reopening and around major holidays. That points at the leave calendar rather than at individual discipline. The working calendar and the lives of the people doing the work have drifted apart, and the fix sits in leave policy.

Both plots take a pivot table. Both of them tend to change the room's assumption about what absence is.

Reading Three: Concentration

Work out what share of your total lost days comes from what share of your staff.

The usual finding is heavy concentration. A small group accounts for most of the days, and a large group has near perfect attendance. When you know the shape of that distribution, you know which lever matters. If eighty percent of lost days come from fifteen people, an organisation wide attendance campaign is the wrong intervention, and fifteen individual conversations are the right one. If the days are spread evenly across hundreds of people, individual conversations will not touch the total and the cause is structural.

Run this by unit as well as organisation wide. Concentration inside a single department, particularly a department doing physically demanding work or running unpopular shifts, is a job design finding rather than an attendance finding.

What to Do About the Reasons Field

Most leave systems record a reason. Most reason fields are close to useless, because the categories were designed for payroll rather than for diagnosis, and because the person entering the record is usually a clerk working from a phone call.

Before building anything on that field, pull a hundred recent records and count how many carry a category that would let you act. If the answer is low, you have two options. You can leave the field alone and rely on the three readings above, all of which work without it. Or you can narrow the categories to a short list that a supervisor can pick from honestly, and accept that the data only becomes useful from the date of the change forward.

I would not spend six months cleaning historical reason codes. The return is poor and the three readings do not need them.

Handling This Fairly

Absence data touches health, and health information carries obligations that a headcount figure does not.

Three rules I hold to. Aggregate figures can circulate; individual records stay with the people who need them to manage the person, which normally means the line manager and HR and nobody else. Long term medical absence comes out of the headline attendance figure and gets managed as its own case, because leaving it in distorts every unit rate it touches. And no analysis of absence gets published at a level small enough to identify one person, which in practice means a floor of around eight staff before a unit rate is shown at all.

There is also a reporting question worth settling early. Attendance figures used in performance conversations change behaviour, and not always in the direction you want. People come to work ill, which raises the infection cost and lowers the output of the day. If you are going to report attendance by team, say plainly what it will and will not be used for.

Where to Start

Take twelve months of leave and attendance records into a spreadsheet. You need four columns: person, start date, end date, and unit.

From those four, build the episode count and the day count per person, then the day of week and month plots, then the concentration curve. All three are pivot table work. None of them needs a licence.

Take the result to one conversation with the operations lead before it goes near the board. The first pass will surface data problems, and it is better to find those in a working session than in a slide.

The number you have been reporting will still be there afterwards. It will simply have three readings sitting behind it that tell you which of the four or five different problems it has been describing all along.

Emmanuel Duncan is the founder of Aletheia Lunara Consult, a professional education and consulting practice working with organisations across Ghana and beyond on people analytics, workforce research and organisational capability.