A marketplace account rarely dies from a dramatic event. It dies from a decimal. Order defect rate drifts from 0.7 to 1.1 percent across a bad month of returns and late replies, and a store that felt healthy discovers its selling privileges were resting on three thresholds it never wrote down anywhere.
Amazon states the thresholds plainly in Seller Central's account health documentation. Sellers go under anyway, because plainly stated and actually monitored are different things. The metrics are knowable. The discipline is the product.
Order defect rate, negative feedback plus A to Z claims plus chargebacks as a share of orders, must stay under 1 percent, measured over a rolling 60 day window. Late shipment rate must stay under 4 percent for seller fulfilled orders. Pre fulfillment cancellation rate must stay under 2.5 percent. Each is a hard line with restriction of selling privileges on the other side, per Amazon's own program policy language.
The windows are the trap. A defect from a bad week rides in your 60 day ODR long after you have fixed the cause, which is why a metric can worsen while operations improve, and why watching the trend weekly beats checking after a warning email. The warning email is not the early signal. It is the late one. Related reading: Serial Number Tracking for Counterfeit Control: Capturing the Unit, Not the Style.
Layered over the individual metrics sits the Account Health Rating, a 0 to 1000 score that starts new accounts at 200 and treats 200 and above as healthy, 100 to 199 as at risk, and below 100 as critical. Violations deduct points in chunks; order volume earns them back slowly. A critical violation can drop the rating to zero outright, which is why policy compliance outranks shipping speed in any sane triage.
Read together, the system is legible: keep every metric under its line with margin, because margin is what absorbs the freak week, the carrier outage, the one furious customer with three claims. A store running at 0.9 percent ODR is not healthy; it is one bad weekend from a case.
Fix causes in the order Amazon weighs them: defects first, because ODR is the heaviest. Work the actual sources, listing accuracy where buyers say not as described, packaging where damage claims cluster, response time where unanswered messages ripen into claims. Then document everything as if a plan of action will be demanded, because at some point in a long selling life, it will be.
Do not chase the metric with theater. Asking buyers to revise honest feedback violates policy and converts a performance problem into a compliance violation, trading points in the recoverable column for points in the dangerous one.
Under 1 percent, over a rolling 60 day window. Order defect rate combines negative feedback, A to Z Guarantee claims, and chargebacks. Above 1 percent, Amazon's policy says selling privileges may be lost or restricted, and experienced sellers target under 0.5 percent to keep a safety margin.
The rating runs from 0 to 1000, new accounts start at 200, and 200 or above is the healthy zone. Between 100 and 199 is at risk, and below 100 is critical, where deactivation is on the table. Violations deduct points; sustained order volume rebuilds them.
Late shipment rate must stay under 4 percent for seller fulfilled orders, and pre fulfillment cancellation rate under 2.5 percent. Both are Amazon program policy thresholds, and breaching either can restrict selling privileges even when your defect rate is clean.
Weekly at minimum, in Seller Central under Performance and Account Health, and daily during peak season or after any operational change. The rolling windows mean problems compound quietly; a weekly check catches a drifting metric while it is still a process fix rather than an appeal.