§6.3The Lighthouses and the Pilot Centres

• DESIGN-TO-COST · TWO FIRST, THEN THREE TO FIVE, EACH WITH THE FULL LOOP

Rather than launching thin programmes across many centres, the plan starts with everything at two, side by side: Chinatown Complex and Geylang Serai Market and Food Centre, the two Lighthouses, opening together in Year 1. Two, and these two, for balance and for inclusion. One is deep in Chinese cooking and the other in Malay, so from its first day the plan is tested on more than one community's food, and halal stalls are in it from the start. Tekka Centre is the third pilot centre and opens in Year 2. Each remains subject to its own hawkers' vote (§6.2). If one declines, the plan seeks another centre with a similar cuisine strength, not simply another willing one.

Any of Singapore's 123 government hawker centres can host a pilot centre, whoever runs it: NEA's list of 17 August 2026 counts 27 that NEA manages (26 of them open), 80 that Town Councils manage and 16 run by socially-conscious enterprise operators that NEA appoints. All of them are eligible, and none is favoured for who runs it. Only privately owned coffeeshops and food courts are left out for now; one may join only if the board recommends it and the hawker-elected Hawker Council does not veto it (§3.8, §6.2). What decides the order is the heritage itself: the plan goes first where the heritage-worthy stalls are most likely to be found in numbers, because that is where a Fellowship, an apprentice and a documented dish are most likely to be needed, and most likely to be taken up.

Why the wider rule still works. NEA regulates the tenancies at all 123 centres, and it remains the tenancy landlord at every centre that it or a Town Council manages, so at those 107 centres rents, the rent bands, the ballot and the tenancies work the same way. Town Councils manage and maintain the HDB-owned centres, and the Minister said on 5 August 2026 that there are no plans to change that. There the Town Council looks after the common property and collects the conservancy charge, and the centre's hawkers' association manages the table-cleaning service on the stallholders' behalf (NEA tender notices, Dec 2022 and Aug 2024), so both become partners: the association re-tenders the cleaning contract, with the Town Council's agreement; the fee subsidy is the Government's (§2.2, §7.7). Whatever a pilot centre needs from whoever runs it is listed in §7.7 as an ask. Wherever the centre, the first step is the support of its own hawkers' association.

At each Lighthouse, every mechanism runs simultaneously and visibly: the rent-band and ballot pilot for vacant stalls, with the table-cleaning fee subsidy and the re-tendered cleaning contract; the first Buying Circles; the first Master Fellowships – eight at Chinatown Complex and two at Geylang Serai in Year 1 – each with two apprentices learning at the master's own stall; the first relief cooks, proving each recorded dish on the masters' days off; the Taste Council tasting blind at a partner kitchen; the Security Package renewing; the evaluator measuring.

How the pilot centres are chosen. By one ranking and four conditions. The ranking is the expected count of heritage-worthy stalls at each centre. In Year 0 that is a stand-in measure, which sizes the budget: the stalls the MICHELIN Guide, the Hawker Heritage Awards and the heritage records already name, with an allowance for the ones they miss (§7.1). Once the Regulars' List has run a year, it is the List's own count at each centre, which reaches stalls no guide has noticed (§3.6). Centres are taken in that order, subject to four conditions:

• A government centre, any of the 123; a privately owned one only by the exception above.
• Every community's cooking. Between them, the pilot centres must cover Chinese, Malay and Indian cuisine, with halal stalls among them; Geylang Serai is included from the start, as the centre where Malay cooking and the plan's halal arrangements are tested. Hawker culture belongs to every community that cooks it, and a pilot that tested one community's food would have tested too little.
• No major works in a centre's first year as a pilot centre, or a start timed after them.
• Consent: the support of the centre's own hawkers' association, then its hawkers' vote (§6.2), and the agreement of NEA or the Town Council (§7.7).

The three pilot centres, and two add-ons. Read in that order, Chinatown Complex comes first: it has the most heritage-worthy stalls of the five centres we counted, 27 on the planning count. Geylang Serai, with 8, is included from the start, for Malay cooking. Tekka, with 9, has fewer than Old Airport Road (16) or Maxwell (13), and I should say so plainly: the community condition, not the count, puts it third, because it is where Indian cooking is tested. Old Airport Road and Maxwell come next by count, which is why they are the add-ons. One test is still open: the works test is not yet met at Chinatown Complex (below). All five below are among the twelve centres whose heritage NEA and NHB have marked with heritage boards. Old Airport Road and Maxwell are examples of a fourth and a fifth centre, not choices, each subject to its own hawkers' vote; if the Regulars' List's count at the end of Year 1 puts another centre ahead of them, that centre moves ahead of them.

To be sure, the five are not the same size: Chinatown Complex has 226 cooked-food stalls and Geylang Serai 63. So the ledgers do not cost an average centre. Each is costed from its own shape – its heritage-worthy stalls for the Fellowships, apprentices and dishes, its stall count for the cover, the relief cooks and the fee subsidy (§7.1). Chinatown Complex costs the most. That is where the heritage is.

One centre, though, is one data point, and I have come to think it is too few. Not every hawker is going to be willing to try our schemes out, and a loop that closes at one centre may have closed because of that centre – its association, its mix of cuisines, its regulars – rather than because of the design. So the plan opens two Lighthouses at once and, in Year 2, widens to three centres, with Tekka beside them, and five at most. Each is chosen by the same tests, with its own hawkers' association behind it. Each must pass the same co-design veto, with its own hawkers willing (§6.2). Each gets its gifts first: the Security Package, the listening and the co-design run before it opens – through Year 0 at the Lighthouses, and in the six months before Tekka opens, in the second half of Year 1. And each runs the full loop (the whole chain at one centre, from master to paid-off successor), not a thin copy of it: Master Fellowships with apprentices at the masters' own stalls, a pool of relief cooks, the Security Package, the rent-band and ballot pilot, and the evaluator. I call all of them the pilot centres; the Lighthouses are simply the first two. What stays national is what should: the Taste Council, the repository and the core institution, which grow a little to serve more centres.

Granted, three centres cost more than one, and Part VII prints how much: the four-year programme comes to $19,683,808, and the add-ons would cost $5,586,777 for Old Airport Road and $4,412,940 for Maxwell (§7.1). Yet a pilot that cannot tell a good design from a lucky centre is not cheap at any price. Tekka runs the loop one year behind the Lighthouses – master → documentation → apprentice → graduate's own stall → a stall that lasts → succession signed – so the evaluator can watch the same joints at different centres and at different ages. If the loop closes at all three, scaling is replication with evidence. If it closes at one and not the others, Singapore learns which joint failed, and where – in public, because the evaluator publishes either way. For the visitor, each centre is something better than a policy: it is a place. You can eat the programme.