A competitor price is evidence, not an order is the whole article in one line, and the reason it matters is that the naive version is not just risky in isolation - it is reflexive. If the other side runs the same rule, two repricers chase each other to the floor overnight and neither system logs anything that looks like an error. Rate limiting the direction of travel matters as much as the absolute guardrail: a floor stops the catastrophe, but capping how far a price can move per hour is what gives a human time to notice. The confidence model is the part I would push hardest on, because most bad repricing decisions trace back to input quality rather than logic - a stale cache, a marketplace offer from a seller who cannot actually fulfil, a coupon price scraped as the list price. Carrying an age and a source on every observed price, and refusing to act on anything below a confidence threshold, removes more incidents than tightening the formula ever does.