Four decisions that put hidden costs on our own books

Repaying the recruitment fees our Senai workers had paid to get their jobs

In November 2020, after a Covid-19 outbreak in a workers' hostel near our Senai kitchen, I asked an independent social-compliance auditor to interview our migrant workers privately, in their own languages. We employed 560 of them among the kitchen's 820 people, recruited through licensed agencies in Malaysia and partner agents in their home countries. The interviews showed that 412 had paid fees to be hired, on average about RM5,000 and in some cases more than RM9,000, much of it borrowed at home at high interest. Nobody at Rumbia had charged them. We had simply never asked what our agency arrangements cost the people at the end of them.

The stakes were plain. Debt taken on in order to be hired is a recognised indicator of forced labour, United States customs authorities had blocked imports from several Malaysian manufacturers over such indicators, and a worker repaying a loan cannot easily refuse overtime or report a hazard.

I recommended repaying every affected worker in full and adopting the principle that the employer pays all recruitment costs. Our group CFO accepted the principle but wanted repayment limited to fees backed by receipts, which very few workers held. The managing director of our Malaysian subsidiary warned that agencies would drop us and that neighbouring employers would resent the precedent. We compromised on evidence, not on principle: the auditor verified each claim through interviews and agency records, and we capped any single repayment at RM9,000.

We repaid RM2.06 million by May 2021 and deferred a third production line at Senai by a year to fund it. Our Malaysian operating margin fell by 0.8 percentage points in 2021. Two agencies stopped working with us, and we replaced them with agencies willing to bill us for the whole cost, now between RM3,200 and RM3,900 per worker.

Annual attrition among migrant workers fell from 34 per cent in 2020 to 11 per cent in 2023, and first-pass yield on the lines rose from 93.1 to 96.4 per cent as trained people stayed. Recruitment cost now sits in our product costing.

I would do one thing differently. We covered our own employees first and reached the cleaning and security contractors' workers at Senai only a year later. They worked on the same floor, and I should have included them from the start.

Building a second kitchen on the Singapore side of the Causeway

When I built the Senai kitchen in 2016, I consolidated three older kitchens into one, two of them in Singapore. On cost it was right: Johor's labour and land cost a fraction of Singapore's, and unit costs fell 19 per cent. It also meant that every cream cake, custard tart and chilled sandwich sold in our 159 Singapore bakeries and cafés crossed the border overnight in our lorries.

On the eve of the Lunar New Year weekend in January 2023, seven of those lorries were held in congestion for between six and nine hours. Loggers on three of them showed chilled product above 8°C for long enough that our head of food safety ordered it destroyed. We discarded S$210,000 of stock, 61 outlets had no cream cakes on the busiest trading weekend of the quarter, and transit discards for the year reached S$640,000.

Two plans followed. Our Malaysian operations team and the group CFO proposed better-insulated trailers, earlier departures and use of both crossings, at about S$1.6 million. I proposed moving the riskiest 30 per cent of our volume, the chilled and cream lines, into a 3,800-square-metre kitchen at Senoko. It would cost S$19 million to build and S$3.1 million a year more to run, because Singapore wages and work-permit levies are higher. The CFO showed that the project would return about 6.5 per cent against our 9 per cent hurdle rate. He was right about the arithmetic. I argued that the hurdle priced a kitchen but not a festive weekend of empty counters, and the board approved the plan with a post-investment review due in 2027.

Senoko opened in January 2025 with 210 staff. Transit discards fell to S$90,000 that year, and no chilled line ran out during either festive season. Senai's utilisation dropped from 88 to 74 per cent. We brought it back to 83 per cent by making the frozen dough that our Thai franchisee had been buying from a local supplier.

What I would do differently starts with an admission: the concentration I was undoing was my own design. In 2015 I modelled the crossing as a fixed four-hour journey. I would now price border delay as a range with a long tail before deciding where anything is made.

Pricing delivery orders as a separate product

By 2022 delivery platforms carried 21 per cent of Rumbia Kopi's sales and 11 per cent of sales at the stalls we run ourselves in Laman Food Halls. The platforms charged commission of around 30 per cent of the menu price, more when we paid for promoted placement, and packaging cost us about S$0.38 an order. When my team and finance allocated those costs, together with the labour of packing orders during the breakfast peak, delivery contributed minus 1.8 per cent of its own sales. Many delivered orders of kaya toast and soft-boiled eggs lost us money, and too many arrived soggy or cold.

I proposed treating delivery as its own product. That meant cutting the delivery menu from 64 items to 31 that travel well; setting delivery prices on average 14 per cent above in-store prices, stated plainly on each listing; withdrawing our own stalls in 11 food halls where packing blocked the walk-in queue; and launching an order-ahead pickup service of our own. Our head of brands disagreed firmly. Platforms rank merchants partly on price and range, and she expected us to lose visibility to competitors who kept their prices level. Some food-hall tenants feared the withdrawal would reduce footfall, so we left their own listings untouched.

We launched in all 52 cafés in January 2023. Delivery sales fell from S$46 million in 2022 to S$37 million in 2023, and café revenue fell 3.4 per cent that year. One platform moved us down its rankings for about three months, which cost us more than I had modelled.

By 2024 delivery contributed 8.6 per cent of its own sales. Complaints about delivered items fell from 2.9 to 1.1 per thousand orders. Pickup orders reached 7 per cent of café transactions, with no commission attached. The executive committee now reviews delivery contribution by brand every month, beside in-store margin, and any proposal to join a platform promotion has to show its effect on that line.

What I would do differently: I would have tested the change in six cafés for a quarter before taking it to all 52. The ranking penalty would then have been a measured risk, and our head of brands would have had evidence on her side or mine rather than a forecast.

Pausing our Thai franchisee's expansion until its chillers could cope with April

Wiangkaew Foods has held the master franchise for Rumbia Bakehouse in Thailand since 2013. By early 2023 it ran 64 outlets in Bangkok, Chiang Mai and Pattaya and planned to reach 120 by 2027, largely through kiosks on petrol-station forecourts. We earned a 5 per cent royalty on its sales and a margin on the frozen dough and fillings we supplied. Thai expansion featured in the growth plan our group CEO had presented to investors.

Our audits in April 2023, during Thailand's hottest month, told a different story. At 17 outlets, mostly kiosks without air-conditioning, display chillers could not hold cream products below 5°C through the afternoon. Two outlets on low-lying Bangkok streets had flooded the previous September and still kept stock on the floor. Only 71 per cent of outlets passed our food-safety audit. Network sales averaged about 205,000 baht per outlet a month, but the forecourt kiosks averaged about 150,000 baht against roughly 260,000 baht they needed to break even.

I recommended halting new openings for twelve months, replacing chillers at the failing sites with the cost shared equally, closing sub-scale kiosks and cutting the royalty to 4 per cent for two years at refurbished outlets. Wiangkaew's owner argued that a pause would hand good locations to competitors. Our CEO was reluctant to withdraw a growth figure she had given investors, and said so at the board. She accepted the plan on condition that we wrote down the criteria for restarting.

We contributed 9 million baht towards chillers, 13 kiosks closed, and the network fell to 51 outlets. We gave up about S$1.1 million a year in royalty and supply margin during the reset.

By 2025, 96 per cent of outlets passed audit and average monthly sales per outlet had risen 22 per cent, to about 250,000 baht, largely because the weakest sites had gone. Expansion resumed in 2025 under the written criteria: air-conditioned sites only, a flood check for every location and a sales forecast reviewed by both companies. Six outlets have opened since.

What I would do differently: heat and flood criteria belonged in the franchise site manual from the first day in 2013. We wrote them only after a hot April showed us what was missing, and a franchisee's money paid for that lesson alongside ours.