HR & attendance × Research note
Fix attendance before payroll: the cheapest payroll error is the one never paid
Payroll errors are common, expensive and mostly start as attendance errors. On a construction site, where overtime is paid at double rates, a missing clock-out or a mis-recorded shift turns straight into money.
Key takeaways
- An EY survey found one in five payrolls contains errors, each costing about $291, with an average of 15 corrections per payroll period.
- Time and attendance errors were the most common, occurring more than once per employee per year.
- Under construction labour laws such as India’s BOCW Act, building workers are paid twice the ordinary rate for work beyond the normal working day, so attendance errors become wage errors quickly.
- Three checks catch much of it before payroll: missing clock-outs, leave still pending approval and implausibly long shifts.
- The goal is review before release. Correcting a paid error costs far more than fixing an attendance record.
01What payroll errors cost
In 2022, EY surveyed 508 payroll professionals at companies with 250 to 10,000 employees. It found that about one in five payrolls contains errors, that each error costs an average of $291 to fix, and that the average organisation makes 15 corrections per payroll period[1].
The most common errors were in time and attendance and expenses, occurring on average more than once per employee per year and costing about $250,000 per 1,000 employees. A 1,000-employee organisation spent an estimated 29 workweeks a year fixing the most common payroll errors[1].
Most payroll errors start as attendance errors, and attendance is cheapest to fix before anyone is paid.
02Why construction attendance is error-prone
Site attendance has every ingredient for errors: large, changing crews; shifts that start before dawn and run late around concrete pours; biometric devices that fail; manual registers kept by supervisors under pressure; and workers who move between sites during a week.
Labour law raises the stakes. India’s Building and Other Construction Workers (BOCW) Act, 1996 is a typical example. Under Section 28, the government fixes by rule the hours of a normal working day and a weekly day of rest. Under Section 29, a building worker who works beyond the normal working day is entitled to wages at twice the ordinary rate[2]. A shift recorded two hours too long is therefore paid at double rate, and a missing clock-out forces someone to guess.
03Attendance and payroll together
These checks work on direct employees’ attendance. On most construction projects the larger exposure is labour supplied by contractors, where ghost workers, proxy attendance and padded overtime are the typical leakage. That is covered in the Construction Payroll note, which compares each wage line with its site, crew and history. The two controls work best together: clean attendance before payroll, then anomaly review of the wage lines before release.
04Case study: Construction HR
From my portfolio · project 14 of 15
Synthetic data- Problem
- Missing clock-outs, pending leave and very long shifts reach payroll unreviewed.
- Decision supported
- Which attendance records to fix before payroll is run.
- Data
- 2,000 attendance records for 200 employees.
- Method
- Three deterministic checks with separate worklists and source evidence.
- Result
- 67 missing clock-outs, 66 pending leave approvals and 67 shifts of 12+ hours.
- Limits
- A rule-based review on synthetic data; no break deduction or legal-limit judgement.
05Three checks before payroll
My Construction HR project runs three simple, explainable checks on 2,000 synthetic attendance records for 200 employees, each with its own worklist and source evidence[3]:
Figure
Attendance records needing review before payroll (2,000 records)
- Missing clock-out: a clock-in with no clock-out and no leave recorded. Without it, hours are a guess.
- Leave pending approval: leave recorded but not yet approved. An administrative flag, not misconduct.
- Very long shift: 12 or more hours between clock-in and clock-out, with no break deducted. It may be genuine overtime, a missed clock-out on the right day, or a recording error.
06How strong is the evidence?
Not every finding in this note rests on the same kind of evidence. This is how I would weigh each one before acting on it.
| Finding | Evidence | Strength | Main caveat |
|---|---|---|---|
| One in five payrolls has errors, $291 each | EY survey of 508 payroll staff | Indicative | One country’s firms of 250–10,000 employees |
| Overtime on building work is paid at double rate | BOCW Act 1996, s.29 | Strong | Normal hours set by state rules |
| Three checks catch attendance errors before payroll | Rules on my synthetic attendance data | Indicative | Agreement with labels is by construction |
07A pre-payroll routine
- Close attendance before payroll, with a cut-off and a named person to resolve every open record.
- Resolve missing clock-outs with the supervisor, not by default hours.
- Clear pending leave approvals before the payroll run, so leave is paid or deducted correctly.
- Confirm long shifts against the site diary or pour log before paying overtime.
- Record the normal working day and break policy per site, so overtime is calculated from rules, not judgement.
- Measure corrections after payroll each period; a falling count is the proof the routine works.
NotesSources
- EY (2022). EY survey: payroll errors average $291 each, impacting the economy. Business Wire.
- Building and Other Construction Workers (Regulation of Employment and Conditions of Service) Act, 1996, sections 28–29. Section 28 text (Indian Kanoon); Chief Labour Commissioner.
- Iwale, A. (2026). Construction HR. GitHub.
Figures are quoted from the sources above as published; where a source reports a range or a survey estimate, it is described that way. Results from my own projects say whether they use real public data or synthetic data.