KT Sparks

Construction consortium · Europe · building a metro line, 1,200+ people badging through site gates

Catching Buddy Punching on a Metro Construction Project

When someone clocks in for a missing colleague, the company pays for an hour of work that never happened. Across 61,000 time records a year, no one can see that by eye. Our UiPath robot picked out clock-ins too perfectly matched to be genuine and tested each against 1.4 million site gate events. Roughly 14% of reviewed staff showed malicious clocking behaviour.

Catching Buddy Punching on a Metro Construction Project
Industry
Construction & Infrastructure
Function
Legal & Compliance

Results

14%
of staff reviewed flagged for clocking fraud
An audit finding the delivery team reported, across the 154 employees reviewed.
61,000
time records analysed, day by day
61,138 records covering 154 employees over 13 months, extracted from the timekeeping report.
1.4M
gate events checked against clock-ins
Around 1.39 million events from access control, across 98 site locations.

01

The challenge

The client is a construction consortium delivering a metro line through the heart of a major European city. That means dozens of active sites, underground stations and shafts, and over 1,200 people passing through the gates in a year. Workers used a timekeeping system to clock in and out. Anyone entering a site also went through an access-control gate. Nobody had ever compared the two.

This opened the door to two forms of abuse. One is buddy punching, where an employee clocks in or out for a colleague who is absent. The other is employees recording more hours than they actually worked. The company ordered an internal audit to measure how widespread both were.

What was at stake

  • Wages for work never done. Each false stamp is paid as a real hour, with employer contributions on top.
  • Too big to see. Nobody can compare 1.4 million gate events and 61,000 daily time records a year by hand.
  • Safety and security on site. On a live underground construction site, someone logged as present who is absent, or the other way round, is a serious problem.
  • Bad habits spread. When abuse is never caught, it grows, because colleagues notice nobody challenges it.
  • Harder to defend costs. Where labour cost is passed through and disputed, padded hours weaken every number the company needs to stand behind.

02

What we did

We built a UiPath robot to act as the analytics engine for the internal audit. It converted the timekeeping reports the company already produced into data, searched for clock-ins that were statistically too well synchronised to be real, and then checked each suspect against access-control records.

Five steps to audit evidence

  1. Turning the report into data. The robot read the PDF timekeeping report and used pattern matching to parse each employee block and each day. The result was a table of 61,138 daily records covering 154 employees over 13 months, holding role, site, totals, and as many as three entries and exits per day.
  2. Finding synchronised clock-ins. Day by day, the robot compared every employee's entry and exit times with those of all other employees. It flagged pairs whose entry and exit times were 0, 1, 2, 3 or 4 minutes apart, the tell-tale pattern of a single person using two badges.
  3. Counting weekend work. As a second indicator, it counted each employee's Saturday and Sunday stamps.
  4. Testing against the gates. For every suspect, the recorded shifts were compared, one month at a time, with about 1.4 million access-control events from 98 site locations, to confirm whether that person really went through the gate.
  5. Giving auditors the evidence. Results went into workbooks split by check and by time interval, with the matching gate events next to each case, so the audit team could go through all of them.

Cutting the noise

On a busy site, entries in the same minute happen all the time: over 2,000 entry-only matches at zero minutes apart. Once both the entry and the exit had to match, the list dropped below 30 pairs per interval. The gate cross-check then placed each suspect's own badge events next to the clock record. What the auditors reviewed was evidence, not a statistic.

Scoped to the investigation

This was quick, single-purpose analytics, carried out at the direction of the audit on data the company already had. It was never meant to be a permanent production robot.

Stack

LayerTools
Automation and analyticsVB.NET, UiPath Studio, Excel and data-table activities
Document parsingRegular expressions, text extraction from PDF
SourcesAccess-control export (Excel, monthly), timekeeping report (PDF)
OutputWorkbooks of audit evidence for each check and interval

03

The outcome

For the first time, the audit team had something it could never have built manually: each suspect clock-in placed next to the gate record that either backs it up or contradicts it.

BeforeAfter
Timekeeping compared with site accessNot doneReconciled across 13 months
How much was checkedImpossible manually61,138 time records plus ~1.4M gate events
Buddy punchingHidden in individual recordsCaught at intervals of 0-4 minutes
EvidenceNo systematic checkingOne workbook per check, with matching gate events
  • Roughly 14% of reviewed employees displayed malicious clocking behaviour.
  • Several hundred hours paid for but never worked came to light over the review period. With construction labour costing about €24 an hour on average, we estimate that at around €7,000-€22,000.
  • 13 months of timekeeping reconciled against 1.4 million gate events, which no one could do manually.
  • Case-by-case evidence to support the internal audit investigation.

What teams tend to overlook

A timekeeping system confirms that a badge was used. It does not confirm who used it. Fraud that depends on a colleague's help looks entirely normal in each individual record. It becomes visible only when records are compared against each other and against a second, independent system. The second pitfall is the law on employee monitoring. Checking staff is regulated, GDPR among other rules, so the analysis must use records the company already holds lawfully, under an audit mandate, and people must review the findings before any action is taken.