When scraping bots set your prices: the repricer feedback loop
Your price is an input to their bot, and theirs to yours
Price scraping is usually discussed as theft of information. The operational risk runs the other way: your published price is an instruction to every automated system watching it, including your own, through the competitor data it buys. A data provider that mixes your marketplace price, your store price, and a coupon site's invented price gives your repricer three versions of reality and no way to rank them.
Loops do not need malice. Two ordinary repricers with undercut rules and no floors will walk a shared product down step by step overnight. Sellers have woken to products sold at a loss by the thousand, each individual price move small, logged, and within the rules as written. The rules were the problem.
Bait prices and poisoned feeds
Deliberate poisoning exists because it works. A competitor or a rogue marketplace seller lists at an impossible price with no intention of fulfilling, your scraper records a market price that does not exist, and your system cuts to chase it. The tell is fulfilment: a price nobody will actually ship at is not a market price. Repricers that weight listings by seller standing, stock signals, and sales history are harder to bait than repricers that take the lowest number on the page at face value.
Set hard floors per product from your own cost stack, landed cost plus fees plus minimum margin, stored in the repricer and not derivable from competitor data. A floor computed from the market is not a floor. It is another way for the market to move your price.
Controls that keep the automation honest
Cap the size and frequency of automated moves, and alert on cumulative drift rather than single changes: a product down 2 percent is noise, the same product down 2 percent eleven times in a night is a loop. Exclude your own listings and your own marketplace storefronts from competitor sets, a configuration error that still turns up in audits. Keep a human approval band for high volume lines, where automation proposes and a person disposes each morning.
And log the input, not just the output. When a price goes wrong, the first question is what the system saw. Teams that can replay the exact competitor snapshot that caused a move fix loops in a day. Teams that only kept the new price argue about it for a month.
Related reading: How do bots exploit backorder and pre-order queues?, How do loyalty point bots drain rewards programs?.
Frequently asked questions
Is price scraping illegal?
Scraping publicly displayed prices sits in contested legal territory that varies by jurisdiction and by how access was gained. Retailers generally do not need a legal theory to act, though: the practical defences are technical and commercial, from bot management on the catalogue to repricer rules that refuse to chase unverifiable prices.
How do I tell a repricer loop from a genuine price war?
Check the timing and the steps. A loop moves in regular small increments at machine intervals, often overnight, and the competitor price moves in mirror image shortly after yours. A price war has human fingerprints: bigger steps, business hours, and prices that hold once they land.
Should repricers ever be fully automatic?
On low volume, low risk lines with hard cost based floors, automation is proportionate. On hero products that carry the month, the cost of one bad night outweighs a year of saved manual work. Match the autonomy to the blast radius, not to the vendor's default setting.
Can a competitor deliberately crash my prices this way?
They can try, by baiting the feeds your repricer reads. The defence is on your side of the wall: floors from your own costs, scepticism toward prices with no fulfilment behind them, and drift alerts. A repricer that cannot be dragged below your floor cannot be crashed, only annoyed.