The listing goes live with a price the seller did not intend: a misplaced decimal, a discount stacked twice, a campaign rule that fired against the wrong catalog. For a few minutes, the storefront is selling at the wrong number. Then someone notices, and the price returns.
Four stages, in order
- The rule fires. A repricing script or a manual edit publishes the wrong price. This is the only instant where the find is both real and unseen.
- The page is crawled. Deal feeds and price trackers read the listing on their own schedule, often several minutes later.
- The gap spreads. Once a feed publishes it, the window is measured in minutes, and the crowd arrives.
- The seller reverts. Either a system detects the anomaly or a support ticket arrives. Stock is often paused before the price changes.
What a node sees
A node reads the same listing on a tight schedule and reports the moment the price stops matching its own recent history. That is a much smaller claim than “a discount exists”: it is “this number changed, and here is what it was before”.
The original price field matters. Without it, a report cannot distinguish a genuine drop from a listing that was always cheap.
Why speed is the entire product
Two nodes reporting the same listing produce one find. The second report is a duplicate and pays nothing. That rule is not a penalty; it is the reason the network can pay for first-hand observation without paying twice for the same fact.
The value of a find is not the discount. It is the head start.
This is also why the reward follows the discount tier rather than the size of the discount itself. A deeper gap pays more, but the network is still buying one verified observation, not an order.
What it will always miss
Glitches that never reach a fetchable page, drops that require a login, and prices locked behind a cart flow are outside what a headless node can see. So are gaps that close inside the polling interval. Every monitoring system has a blind spot; the honest thing is to name it instead of implying total coverage.
The channel shows what was caught. The absence of a listing is not proof that nothing happened, only that nothing was observed in time.