1 This is a claim for unpaid overtime pay. The claimant is a Bangladeshi national who was employed by the respondent company under a work permit as a food processing worker from around December 2023 to 8 December 2025. During his employment, he was deployed to work at a Bangladeshi restaurant operated by the respondent.

2 The respondent had not given the claimant any written record of the key employment terms in accordance with s 95A of the Employment Act 1968 (2020 Rev Ed) (the “EA”), but both parties agreed that the terms of employment were governed by the in-principle approval (the “IPA”) issued by the Ministry of Manpower. The IPA provided that the claimant was to work 44 hours per week for six days a week, at a basic monthly salary of $1,500 and a fixed monthly allowance of $500, giving a total monthly salary of $2,000. The overtime rate stipulated in the IPA is $11.80 per hour.

3 The claimant’s case was that throughout his employment, he had worked between 13 to 15 hours a day (excluding breaks), seven days a week. For the purposes of this claim, he limited his claim to the period from 1 April 2025 to 8 December 2025. He maintained that he had worked a total of 1,848.8 overtime hours during this period, which at the stipulated overtime rate would amount to $21,815.84. Given that the jurisdiction of this tribunal is capped at $20,000 for the present case, he claimed that sum. The respondent denied liability entirely. Its position was that the claimant had not worked any overtime at all.

4 I pause to note two features of the claimant’s case. Neither point was raised by the respondent, but I address them for completeness.

5 First, the overtime hours that the claimant purportedly worked far exceed the ceilings established under ss 38(5) and 38(8) of the EA. Section 38(5) provides that an employee must not work overtime for more than 72 hours a month, and s 38(8) provides that an employee must not work for more than 12 hours in any one day, save in certain specified circumstances which do not apply here. The claimant claimed to have worked more than 12 hours each day, and more than 200 overtime hours per month during the relevant claim period, for which he had not been paid at all.

6 Nonetheless, the General Division of the High Court in Hossain Rakib v Ideal Design & Build Pte Ltd [2023] 5 SLR 1529 held that s 38(5) does not prevent an employee from claiming overtime pay beyond 72 hours per month, as that provision (and Part 4 of the EA in which it resides) exists to protect employees and not to prejudice their rights to payment for work done, especially when required by the employer (at [41]). To hold otherwise would be to allow an employer to use its own statutory breach as a shield against paying wages it owes (at [61]). That reasoning applies as well to s 38(8): work performed in excess of 12 hours each day does not strip the employee of his entitlement to be paid for those hours.

7 Second, the claimant also claimed to have worked seven days a week, which means he had been working on all of his rest days. A claim for pay for work done on a rest day at the request of the employer would ordinarily fall under s 37(3) of the EA rather than s 38(4). The former would have been more favourable to the employee. It entitles her or him to a sum at the basic rate for two days’ work, plus one and a half times her or his hourly basic rate for each hour exceeding the normal daily hours.

8 However, the claimant did not frame his claim on that basis. His claim referral certificate, which is required under s 12(5) of the Employment Claims Act 2016 (2020 Rev Ed) (the “ECA”) to be issued in respect of every specified employment dispute for which the claim is lodged, had been issued only in respect of overtime payment under s 38(4). That meant I was not seised of jurisdiction to hear and determine a claim for pay for work done on a rest day under s 37(3). But there is nothing preventing an employee from claiming for extra work performed, even on a rest day, in excess of 44 hours per week under s 38(4). This was what the claimant had chosen to do, and for which a claim referral certificate had been issued. That choice means only that, should the claimant succeed, the amount he would be able to recover will be less than what he might otherwise have recovered had he proceeded under s 37(3) as well, subject of course to the jurisdictional cap of $20,000.

9 I turn now to the evidence.

10 The primary piece of evidence that the claimant submitted in support of his claim was an attendance table covering the entire period from 1 April to 8 December 2025, prepared with the assistance of a non-profit organisation after the end of his employment and the commencement of this dispute. The table stated that the claimant worked every calendar day during the entire period, generally from 5.00am to 9.00pm on Mondays to Saturdays, and from 7.00am to 9.00pm on Sundays. There were some deviations from this pattern—for example, on some weekdays he started work at 7.00am—but these were not significant.

11 Prior to the hearing, the claimant had not submitted any objective evidence to corroborate the attendance records reproduced in the table. His position, which he maintained consistently in his written witness statements throughout the pre-trial case management period, was that the respondent had maintained records of his attendance, and that these should be made available to him for the purposes of this dispute. The respondent, however, denied the making and keeping of any such records.

12 At the hearing, the claimant was therefore invited to explain the basis upon which the table had been prepared—in particular, whether it was based on memory or on contemporaneous records he had maintained.

13 The claimant explained that from March 2025, he had written down on a piece of paper his starting and ending work hours for each day until 8 December 2025, and that the attendance table was based on these personally maintained records. This piece of paper was not submitted in evidence prior to the hearing, nor could the claimant produce it at the hearing, as he explained that it remained in his room.

14 He was then asked why he had only begun maintaining this record from March 2025. He explained that it was around that time that the respondent switched from a punch card attendance system to an electronic facial recognition system. Under the punch card system, he had occasionally been able to obtain completed punch cards from his supervisor and photograph them as his own records. When the company transitioned to the facial recognition system from March 2025, he could no longer access those records, and so began keeping his own handwritten notes of his time in and time out each day. The claimant also submitted a photograph of himself standing in front of the facial recognition machine, and explained that this machine was maintained at the premises of a restaurant adjacent to the one he worked at, which was operated by a related company of the respondent. He said he was instructed to use that machine each working day.

15 Similarly, photographs of the punch cards had not been submitted in evidence prior to the hearing, which is understandable given that the punch cards would have related to work performed prior to the period of his claim. Nonetheless, as some of these photographs were stored on the claimant’s phone, I directed him to produce them at short notice during the hearing itself. He submitted photographs of punch cards bearing his name for the months of May, July and August 2024. The punch cards show that, for these three months, he worked hours consistent with those stated in his attendance table: on Mondays to Saturdays, he generally reported to work between 4.20am and 4.45am; on most Sundays, he reported shortly before 7.00am, although there were some instances where he reported as late as 8.36am; and he would generally leave work at around 9.00pm, with some days ending slightly before 9.00pm and many days extending closer to 10.00pm. He also submitted screenshots of the metadata of these photographs, showing that they were taken at around the same time as the punch cards themselves.

16 The respondent’s evidence in response to the claim was exclusively the testimony of its three witnesses: its office manager, who was also the representative at the hearing; its director; and its employee who works as a chef at the same restaurant where the claimant was deployed.

17 All three witnesses were unequivocal that the claimant never worked any overtime hours. When the respondent’s representative was invited to specify the working hours and working days of the claimant, she said that none of its employees has fixed working hours, and that start and end times are communicated verbally by the supervisor the day before. Similarly, rest days are not fixed but given whenever business is slow, and shifts are approximately eight hours long, inclusive of a one-hour break. She maintained that despite all this flexibility, the company was sure that the claimant did not work more than 44 hours a week — once he was done with his eight-hour shift, he would leave and the chef would take over his duties. She also explained that if any employee worked overtime hours, he would be paid in cash on the same day.

18 The respondent’s representative also maintained that the company did not make or keep any records of the claimant’s working hours. She denied knowledge of any punch card system and said she did not know where the claimant obtained the punch cards. She did not deny, however, that a facial recognition system was introduced by the respondent’s related company for the adjacent restaurant. But she claimed that this system was used only for employees of the related company working at the adjacent restaurant, and that the claimant had no access to it. When invited to explain why the claimant had a photograph of himself standing in front of the machine, she suggested that he must have entered the office without permission and taken the photograph to manufacture evidence for this claim.

19 The chef testified that he works approximately 11 hours a day—from 6.00am to 10.00am, and from 4.00pm to 11.00pm—for six days per week. He acknowledged that he worked overtime every working day but added that he would be paid in cash for two overtime hours daily. Although he regularly worked overtime, he maintained that the claimant never did. This was because, once he was done with cooking, he would take over matters of service from the claimant, and if any further cooking had to be done, the chef from the adjacent restaurant would assist.

20 The director’s evidence largely restated what the office manager and the chef had already said, save that he additionally impugned the claimant’s motives for bringing the claim, suggesting that it was brought purely in retaliation after the company cancelled the claimant’s work permit because he was a troublesome employee who had fought with customers on three occasions. He also aired grievances about the non-profit organisation assisting the claimant.

21 The task before me was to determine the extent of overtime hours the claimant worked, if any. The burden rested on the claimant to prove that it was more likely than not that he had worked overtime to the magnitude of 1,848.8 hours for the period from 1 April 2025 to 8 December 2025.

22 There was something striking about the claimant’s account—that he had worked 13 to 15 hours a day, well beyond the 44 hours per week that his IPA stipulated and without rest, for an extended duration, and seemingly without raising a formal complaint prior to this claim. The respondent made this point in its submissions, seemingly contending that the claimant’s account is inherently improbable, and further implied that the claim was commenced by the claimant as a form of retaliation against its decision to terminate his employment.

23 I confess that I had initially found the respondent’s improbability argument appealing. That instinct arose, perhaps, from an intuitive assumption that an employee would not ordinarily tolerate the conditions alleged by the claimant for an extended period without protest—and, yet, on the claimant’s own account, that was precisely what he had done, raising the matter only after his employment was terminated.

24 It is here that the words of Lord Bingham, written in an extra-judicial essay, give one considerable pause. He observed (see Sir Thomas Bingham, “The Judge as Juror: the Judicial Determination of Factual Issues” (1985) 38(1) Current Legal Problems 1 at p 14):

25 The point is that how people respond to any given set of facts will vary across persons of diverse social and cultural backgrounds and individual experiences. The adjudicator may simply not be competent to judge how likely it would be for a person whose background and experiences are far removed from her or his own to behave in a particular way.

26 Lord Justice Popplewell made a related observation in his extrajudicial speech (see “Judging Truth from Memory: The Science”, speech at the Commercial Bar Association (16 November 2023)). Drawing inferences about inherent probabilities, he said, is a “subjective value judgment” that may involve “an element of overreach in our asserting what would be expected in given circumstances”, since “the individual experience of any single judge… is rarely full enough to be well informed about the typical behaviour of the individual in the particular factual context” (at [13], [16]). He also made the observation that will ring true to most: that “we have all had experiences in life and litigation in which the apparently improbable turns out to be what happened” (at [17]).

27 These words of caution counsel against too readily accepting the argument that it was inherently improbable for one to have behaved as the claimant did. Indeed, on further reflection, I was not persuaded that there is anything inherently improbable about a worker accepting such conditions for as long as the claimant did and raising such matters only after her or his employment ended. Employees tolerate poor working conditions for all manner of reasons—fear of conflict, hope that things will improve, reluctance to risk a steady income, difficulty with finding alternative employment, or simple resignation that complaining will achieve nothing. None of these are uncommon responses for any employee in Singapore.

28 Some of these reasons operate with greater force for foreign workers, who inhabit a different labour reality from most employees in Singapore. A foreign worker’s right to remain in Singapore is tied to her or his employment. Raising employment issues during the period of employment risks repatriation. If she or he came to Singapore carrying recruitment debts, repatriation could mean financial catastrophe. It may also be difficult to seek alternative employment, since foreign workers generally require employer consent to transfer. Hence, a foreign worker with unpaid overtime or lack of rest days may, quite reasonably and rationally, consider that it is better to keep a bird in the hand—in the form of one’s monthly salary—than two birds in the bush—by pressing for overtime pay or rest day pay and risking repatriation.

29 I make clear, however, that I made no positive findings that any of these reasons in fact applied to the claimant. No evidence was led before me on this. The point I make is narrower: that it was not inherently improbable that the claimant would have raised these matters only after his employment had been terminated. By then, the risks that may have previously counselled silent tolerance would cease to exist. In that limited sense, the respondent may well be right that the claim was only commenced in response to the termination. But I did not think that the timing of the present claim, without more, necessarily impugned the factual foundation upon which it rested.

30 I therefore rejected the inherent improbability argument. The better approach, rather, was to determine the matter on the inferences drawn from the documentary record and witness evidence before me.

31 The only documentary evidence relating to the hours worked by the claimant was the attendance table he submitted. This provided a useful starting point. The table was said to have been prepared based on handwritten notes the claimant had been maintaining contemporaneously since March 2025. That evidence was not produced before this tribunal. Even if it had been, I would have had concerns about self-produced timesheets standing alone. An opportunistic employee could easily fabricate such records to support a claim against a former employer.

32 That said, I considered the attendance table produced by the claimant to be credible, for two reasons.

33 First, the claimant was able to give a detailed and internally coherent account of how the table had been prepared, and why he began to maintain the handwritten notes from a specific point in time. This was linked to his explanation of how the attendance system worked and evolved, which was corroborated by the photographs of the facial recognition system and the punch cards.

34 There was only one discrepancy. In an earlier witness statement, the claimant had described the electronic attendance system as a thumbprint system. At the hearing, he testified that it was a facial recognition system. When the discrepancy was put to him, he expressed surprise and maintained that it was the latter. His explanation was that his witness statement was likely recorded wrongly, given that he required assistance to record his evidence in English, a language he neither spoke nor read. I found this to be a minor discrepancy and accepted that it was more likely than not the result of miscommunication in the course of interpretation.

35 Second, although the punch cards had not initially been submitted in evidence—apparently because the claimant did not appreciate their relevance—I found them to be credible and authentic contemporaneous records that corroborate the attendance table. To be sure, these punch cards predated the claim period, which ran from 1 April 2025. Nonetheless, the claimant’s case had always been that his working hours were generally consistent throughout his employment, and the punch cards revealed working hours consistent with those recorded in the attendance table. They therefore bore corroborative value.

36 In my view, the claimant’s evidence established a prima facie case, and one which rose to the level where the evidential burden shifted to the respondent to show that the claimant did not work overtime to the magnitude of 1,848.8 hours for the relevant period.

37 It would not have been difficult for an honest and responsible employer to discharge that burden. Section 95(1) of the EA requires an employer to make and keep employee records containing prescribed particulars for every employee and former employee, including the hours worked each day and the duration of any breaks. An employer which had made and kept such records could easily have produced them and identified any discrepancies between its own records and the employee’s evidence.

38 The respondent’s case was that the company had not made or kept records of employees’ hours worked. As to the claimed ignorance and disavowal of any punch card system, I found this unconvincing in the face of the photographs of the punch cards submitted by the claimant.

39 As to the facial recognition system, I rejected the respondent’s representative’s explanation that it had been used only for employees of the related company at the adjacent restaurant, and not the respondent’s employees. The evidence before me, including the respondent’s own witnesses, established that the two companies operated, for all practical purposes, as a single integrated entity. The representative herself worked for both companies. Employees from the related company’s restaurant were said to have also cooked food for the respondent’s restaurant—this may raise questions about compliance with the relevant work permit conditions, though I made no finding on that. When the representative was asked whether it was accurate to say that the two restaurants shared all their resources quite extensively, she agreed that this was so. In my view, the insistence that the facial recognition system was conveniently ring-fenced for the related company’s employees, and that the claimant had no business using it, was not convincing.

40 I therefore found that the respondent had in fact maintained attendance records. It simply refused to disclose or make accessible such records by maintaining the position that no such records ever existed. This amounted to a breach of s 95(2) of the EA, which requires that such records be readily accessible to the employee or former employee to whom they relate. And this breach engaged s 21(2) of the ECA, which provides that the tribunal may draw any inference it thinks fit from a party’s failure to comply with specified statutory obligations including those under s 95 of the EA, and that such inferences include one that any evidence not produced on account of that failure would, if produced, have been unfavourable to that party. In my judgment, this was an appropriate case to draw such an inference.

41 That inference was further reinforced by the weakness of the respondent’s own evidence. The respondent’s position was not that the claimant had worked overtime but considerably less than 1,848.8 hours, or that he had worked overtime to some extent and was already paid for it. It was that he had not worked any overtime at all.

42 Even setting aside the attendance records, when a former employee brings a claim of this magnitude—every day worked, no rest days, overtime hours numbering in the thousands across months—she or he takes on a correspondingly greater risk of being contradicted. The broader the claim, the easier it becomes for an employer to find at least one day, or one record that does not fit. In the age of digital communication, such evidence would not be difficult to come by. A message, a roster, a record of any kind showing that the claimant was not at work on a particular day or worked fewer hours than she or he claimed.

43 But the respondent produced not a single well-evidenced counter-example. All communication regarding working hours, it says, was conducted verbally and face to face. This was hard to accept given the representative’s own evidence that staff did not have fixed working hours or rest days and were informed of their hours only the day before. If that were true, it was difficult to see how there would be no trace of any digital communication over the nearly two years of his employment to support these assertions.

44 There was a further reason to disbelieve the respondent’s account. The restaurant had only three employees: a cleaner, the claimant who was primarily responsible for service, and the chef. The restaurant opened at around 6.00am to 7.00am and closed at around 11.00pm. I had considerable difficulty understanding how any meaningful shift system could have been implemented between the claimant and the chef, with no other workers available to relieve the claimant after his purported eight-hour shift ended.

45 The internal incoherence of the respondent’s account was put in sharper relief by the representative’s own concession that the claimant regularly started work as early as 5.00am to 6.00am and worked shifts of approximately eight hours, which would have put the end of his shift at around 2.00pm to 3.00pm at the latest. Yet the chef’s own evidence was that his morning shift ends at 10.00am, and he did not return until 4.00pm. On the respondent’s own account, that left a gap during which neither the claimant nor the chef would be at work. It was more likely than not that the employees had to work overtime given the operating hours of the restaurant, and the chef’s own evidence—that he worked an average of 11 hours per day—further supported the claimant’s case.

46 In summary, the respondent did not seriously challenge the claimant’s attendance table in any specific respect. It did not point to a single entry or produce any objective evidence to show that on any particular day, the claimant had not been at work or worked fewer hours than claimed. In the absence of any such challenge, and given the claimant’s evidence as a whole, I was satisfied that the claimant had worked the hours stated in the attendance table from 1 April to 8 December 2025. Any concerns regarding the precision or accuracy of the hours stated were, in my view, immaterial, and should in any case be resolved in his favour given the adverse inference drawn from the respondent’s refusal to disclose the attendance records in its possession.

47 For the reasons given, I allowed the claim in full and awarded the claimant the sum of $20,000. In addition, I ordered the respondent to pay the claimant $400 in costs and $60 in disbursements.