Lab 16 / the biggest unbuilt surface
Every other surface reads from the platform's own book. This one calls suppliers, and a supplier call costs time, costs a rate limit, and returns an answer that starts going stale the moment it arrives. So this screen has three jobs no other screen has: say what it is about to spend, say how old the answer is, and prove that two prices are for the same thing before it invites a comparison.
Underneath it are two doctrines. Items are never typed: anything the marketplace cannot sell comes from the item library, and a manual asset is created once and reused forever. And there are no suppliers on this screen. The agency is buying from Tripbooks. Which wholesaler is behind a rate is Tripbooks' business, it is supplier number one in the books, and no name, rate id or third party identifier appears anywhere a subscriber can read it.
That second rule turned out to improve the design rather than constrain it, and the section on rate options explains why.
Something failed to answer, and the strip says so: some rates could not be retrieved, this list may be incomplete, with a retry. The agent learns the one thing that affects their decision, which is that the cheapest rate on screen might not be the cheapest rate available. They do not learn who failed, because who failed is Tripbooks' problem to fix, not the agency's problem to interpret.
02 / the idea the whole screen is built on
The same Deluxe King at Park Hyatt Kyoto is available three ways. Most systems show that as three results and leave the agent to work out whether they are looking at one room or three. Here the property and the room are proven identical behind the scenes, so they collapse into one row, and what opens underneath is three rate options that differ only in cancellation and payment terms.
Removing the supplier names made this better, not worse. When the rows were labelled by who supplied them, the eye went to the brand and the decision looked like a choice of wholesaler. Labelled Cheapest, Recommended and Most flexible, the row says what the money actually buys, which is exactly the decision an agent is making.
03 / the honest comparison
On this search Tripbooks is $183.00 a night below retail on comparable terms, which is what happens the overwhelming majority of the time. The design question was never how to present that. It was what the component does on the rare search where retail wins.
It says so, at the same size, with the terms underneath. That case is uncommon enough to cost almost nothing and valuable enough to be the reason the other ninety-five per cent is believed. A comparison that only appears when it flatters us is not a comparison, and an agent will work that out the first time they check.
04 / the laws this screen adds
Not a name, not a rate id, not a third party property code, on the operator side or the customer side. The agency buys from Tripbooks; Tripbooks is supplier number one in the books, and everything behind it is plumbing that appears only in the ledger and the audit trail.
Fetched four minutes ago, tokens expire in eleven. Drawn with the same clock component as the fulfilment queue, so time behaves the same way everywhere.
When something fails to return, the list says it may be incomplete and offers a retry. It never says what failed. The screen would rather admit an incomplete answer than present a confident wrong one, and it can do that without narrating its own supply chain.
The property and the room are matched to a standard before two rates are ever shown as alternatives. Where that proof does not exist they stay on separate rows, because a false comparison is worse than a shorter list.
Cancellation and payment terms sit beside every figure in every view. A rate without its ladder is not a rate, it is a number.
If an agent overrides a marketplace rate, the item records what the marketplace quoted alongside what was entered. Undercutting a locked net routes to an approver, and the approvals worklist already knows how to show it.
Ruled, and one consequence to carry