A launch sells out or it leaks, and that is decided before the doors open

Two towers launch in the same quarter. Comparable product, comparable location, comparable price per square foot, comparable spend. One clears its release schedule in six weeks. The other takes seventy per cent of its volume in the first fortnight, stalls for four months holding the units nobody wanted, and finishes with an incentive programme that gives back the margin the launch was supposed to earn.

The market explains that difference as demand. Usually it was control.

The scarce resource is not demand. It is the order of release.

Off-plan is one of the few sales environments where the seller controls supply precisely. A release schedule is a pricing instrument. Holding the best-facing units back until the third tranche, pacing release against absorption, keeping a reserve for the buyers who arrive late with the most money — these are decisions about scarcity, and over a launch they are worth more than any single campaign.

They also survive contact with launch day for about an hour.

The release exists in two places. It exists as a plan, in a document senior people agreed. And it exists as a state, in whatever file the sales floor is actually working from. Within an hour of the doors opening, those two things have diverged. Someone released a unit early for a buyer who was standing in front of them. Someone placed a hold and did not record it. Two agencies were told the same unit was available. The plan is still true in the deck and no longer true in the building.

Ask a developer's sales operation what is available right now and you rarely get one number everyone agrees with. That is not a technology failure. It is that availability is being maintained by people, and people maintain it at the speed of conversation while a launch moves at the speed of a room.

Allocation is a governance problem wearing a sales costume

Every agency wants inventory held. The request is reasonable — they cannot sell what they cannot promise. The difficulty is that a hold is a loan of scarcity, and loans need terms.

Without an expiry the system enforces, holds accumulate. Two weeks into a launch a developer can be fully allocated and barely sold: every good unit spoken for, almost none of it converted, and no way to distinguish an agency working a genuine buyer from an agency parking stock so that a competitor cannot have it.

Without a rule, allocation follows the volume of the request. The largest agency asks most persistently, receives the best tranche, and converts it at whatever rate it converts. The agency turning one hold in three into a reservation is treated identically to the agency turning one in twelve, because nobody is measuring the difference. Nobody is measuring it because the hold record is a message thread.

Allocation discipline is not complicated to describe. How many units, in which tiers, to whom, for how long, and what happens automatically when the period ends. What makes it hard is that each of those is also a relationship decision, and relationships are precisely what a governed rule exists to protect from short-term pressure. A rule the system applies is a rule the head of sales does not have to defend personally, twenty times a day, to people they will need again next quarter.

Registration is the record that decides who trusts you

This is where launches actually leak.

An off-plan buyer speaks to three agencies. That is normal behaviour, not bad faith. One registered the client on Tuesday and did the work. One met them at the sales gallery on Saturday and closed. Both will claim the introduction, and both will believe themselves.

If registration is not timestamped, scoped and expiring, the dispute is settled by whoever escalates hardest. The immediate cost is a commission argument. The lasting cost is that your strongest agencies learn what the real rule is, and it is not a rule. They rarely stop working with you after a dispute. They do something more damaging: they keep working with you, and quietly route their best buyers to the developer whose registration record they trust.

A registration record that genuinely resolves disputes needs a few unglamorous properties. It identifies a person by something durable rather than by a name, because there are a great many buyers with the same name and one of them is always the one in dispute. It carries a timestamp nobody can edit afterwards. It has a scope, because registering a buyer against a project is a different claim from registering them against a unit type. It expires, because a registration that never lapses is a permanent claim on a buyer the agency stopped calling months ago. And it has a stated precedence rule for overlaps, agreed before the launch rather than negotiated during it.

There is one further piece, and it is the one developers avoid. The direct channel has to register under the same rules. If the developer's own team can claim a buyer an agency registered a fortnight earlier, the network learns that the house wins ties, and it prices that into how much of its pipeline it gives you.

What leaking looks like, in order

It rarely arrives as a single failure. It arrives as a sequence, and the sequence is recognisable.

• Availability drifts, so the floor stops trusting the inventory position and starts confirming by phone.
• Holds outlive their purpose, so the project is allocated but not sold.
• Registration disputes begin, and each is resolved by judgement rather than by record.
• The strongest agencies deprioritise the project without announcing it.
• What remains is the hardest stock in the building, and the release levers that would have helped were spent in week one.
• Incentives start. The launch pays for its own operational failure, in margin.

By the time anyone calls it a demand problem, six operational decisions have already produced it.

The launch does not end at the reservation

A reservation is not a sale, and a sale is not the end of the operation. Reservations that do not convert to a signed agreement inside a defined window should return to inventory without anyone having to remember. Payment plans create obligations running for years, and a missed instalment is a relationship event before it is a finance event.

Then there is handover, which most developers treat as an exit. The buyer becomes a unit number in a management system at precisely the moment they are most likely to buy again or introduce someone who will. The cheapest demand available for the next launch is the owner list from the last one, and it is routinely discarded because nobody owns the record once the keys change hands.

None of this is exotic. It is inventory, allocation, registration and obligation held as governed state rather than as institutional memory. A developer selling its own stock through a broker network is not a brokerage with different inventory; it is a different operating model, and it needs the operating layer arranged differently — which is what the property developers page describes.