How to analyse employee attendance indicators
06.09.2026
Attendance analysis is the stage that comes after the report. The report shows what happened; the analysis explains why. Most companies stop at the first stage: the numbers exist but no decision follows. This guide shows which indicators are worth looking at, how to read them, and what decisions the analysis should lead to.
Which indicators are meaningful
Dozens of indicators can be calculated, but six are enough in practice:
- Attendance rate - the share of scheduled days actually worked.
- Average lateness - the average number of minutes when lateness occurs.
- Lateness frequency - how many times a month.
- Actual against planned hours - was less or more worked?
- Overtime share - what percentage of total hours.
- Number of corrections - how many records were changed by hand.
The last indicator says something about the system rather than the employee: if the correction count is high, the problem lies in the recording process.
How to read a number
A single figure is misleading. The basic rule of analysis is comparison, in three directions:
- Over time - this month against last month.
- Across groups - between departments, branches and shifts.
- Against plan - relative to an internal target.
An attendance rate of 92% is by itself neither good nor bad. If it was 97% last month, that is a signal. If other departments run at 91%, that is the norm.
Typical patterns and what they mean
| Observation | Likely cause | What to check |
|---|---|---|
| Lateness rises early in the week | Schedule or commute | Shift start time |
| Constant overtime in one department | Understaffing or planning | Workload distribution |
| Check-out records are missing | The point is badly placed | Position of the recording point |
| High number of corrections | The rule is unclear | Exception list |
| One branch is unusually high | Weak recording discipline | Accuracy of the records |
The last row deserves attention: an unusually good figure should also be investigated.
A practical example
At a logistics company the warehouse team's attendance rate was higher than every other department - close to 100%. Management read this as a positive.
The analysis showed something else: in the same team the overtime share was very low and check-out records almost always landed exactly on the scheduled time, even though there were days when deliveries arrived late.
The investigation revealed that records were entered in bulk by one person at the start of the shift. The number looked excellent because it was not real. After the recording rule was changed the rate fell slightly, but the indicators reflected reality for the first time - and the case for additional warehouse headcount was built on those numbers.
How often to analyse
Daily is too much, yearly too little. A practical rhythm looks like this:
- Weekly - exceptions only (long gaps in records, many corrections).
- Monthly - the six core indicators, by department.
- Quarterly - trends and schedule decisions.
From analysis to decision
Analysis should not end up in an archive. Every monthly review should produce at least one concrete decision: changing a schedule, moving a recording point, clarifying a rule, or raising a headcount question. An analysis that produces no decision leads to the same numbers repeating next month.
The human factor
These indicators measure people's behaviour, which calls for care. Two rules help: never discuss a number without its context, and let employees see their own data. With transparency the indicator becomes a subject of discussion; with secrecy it becomes a source of resentment.
How indicators complete each other
The most common mistake in analysis is reading indicators in isolation. In practice each number carries meaning only alongside its pair.
The attendance rate should be read together with the overtime share: a high rate with high overtime suggests the team is understaffed, while a high rate with low overtime suggests a stable workload.
Lateness should be read together with early leaving: if both are high, the problem is in the schedule; if only lateness is high, the cause relates to the morning.
The correction count is read together with the number of incomplete records - if both are high, the position of the recording point or the design of the process needs reviewing.
These pairings are the core tool of analysis and require no complex calculation.
Seasonal and cyclical effects
Many indicators depend on the season, and comparisons made without accounting for that produce wrong conclusions.
In retail and logistics the end of the year, in education the start of a term, and in manufacturing the order cycle all shift the figures systematically. Comparing January with December in such cases is meaningless.
Two practical approaches solve this: compare the same period against the same period of the previous year, or use a three-month rolling average. Both smooth out the seasonal effect.
Special situations should also be noted: public holidays, stocktaking periods, training weeks. Without those notes it becomes impossible, a year later, to explain why a number jumped.
How to present the results
The value of an analysis also depends on how it is delivered. Sending a table of numbers usually achieves nothing - nobody opens it.
A format that works is short: three key observations, the figure behind each one, and one proposal for each. An analysis longer than a page does not get read.
The second rule is to give context. "Lateness rose by 12%" creates alarm on its own; "lateness rose by 12%, 80% of the increase comes from one team, and that team's route has changed" leads to a decision.
The third rule is consistency: the same indicators in the same format every month. When the format changes, the ability to compare disappears.
Building a comparison base
Analysis needs a comparison base, and it should not be chosen at random.
Three types are used in practice. A historical base uses the unit's own figures from a previous period - the simplest and most common option. An internal base compares departments and branches; units with similar working conditions should be compared. A target base uses goals the company sets for itself.
The third is the most useful and the least used, for a simple reason: setting a target requires a decision. Once a concrete goal exists - "attendance must not fall below 95%" - the way the report is read changes as well, because the figure is measured against the target.
Targets must be realistic: derived from current performance and raised in stages. An unreachable target removes the indicator from the set of management tools.
In summary
The purpose of attendance analysis is not to evaluate employees but to find where the schedule, the process or the planning does not fit. Reading indicators in pairs against a comparison base makes that possible.
The practical minimum is: six indicators, a monthly cycle, comparison by department, and at least one concrete decision at the end of every review. An analysis that leads to no decision repeats with the same numbers next month.
The next step
Starting an analysis requires no complex tooling. Comparing three months of data by department is enough - the differences raise the questions by themselves.
QRGate provides the indicators in ready-made views and supports export. See what QRGate can do or learn more.
