Prices on many online listings change several times a day without anyone deciding each one. The changes are produced by systems following instructions written in advance.

Rules come before models

Most repricing begins with explicit constraints: a floor below which the item may not sell, a ceiling, and a target position relative to competing listings.

Within those bounds the system adjusts continually, and much of what looks like sophisticated behaviour is simply a rule tracking another seller's published price.

The constraints matter more than the algorithm, because they prevent the sequence of small adjustments from arriving somewhere the seller never intended.

Demand signals drive the direction

Systems observe how many people viewed a listing, how many added it to a basket, and how many completed a purchase, then compare those rates against expectations.

Strong interest with weak conversion suggests the price is above what the audience will pay. Rapid conversion suggests room to move upward.

These readings require enough traffic to be meaningful, which is why automated repricing is concentrated on high-volume listings and rarely applied to slow sellers.

Inventory and time close the loop

For goods with a deadline, such as travel, events or perishable stock, the system also weighs remaining units against remaining time.

Selling ahead of the expected pace allows prices to rise, while falling behind forces them down as the deadline approaches and the alternative becomes zero.

This is why prices in those categories can move sharply in both directions rather than declining steadily toward the date.

Competing systems interact

When several sellers all reprice automatically against each other, the result is a feedback loop where each adjustment triggers responses from the others.

These loops can run prices down to their floors within hours, or produce oscillations that no individual seller chose or benefits from.

Floors, damping and update delays exist mainly to interrupt these interactions rather than to improve any single seller's pricing.

What the movement tells a shopper

Frequent small changes indicate an automated system tracking competitors, and the price is likely to return to a similar level after moving.

Large step changes usually indicate a human decision, a promotional period or a change in stock position, and are less likely to reverse quickly.

Neither pattern reflects anything about the individual viewing the page, which is a distinction worth keeping separate from personalised pricing.