Order Defect Rate and Account Health on Amazon: The Thresholds You Cannot Cross

There are two kinds of metrics in a seller’s life. The ones that tell you how much you are earning, and the ones that tell you whether you will still have a business next week. Net margin, ACoS and inventory turnover belong to the first group. The Order Defect Rate belongs to the second, which is why it gets treated differently: you do not optimize it, you watch it.

The trouble is that almost nobody watches it until the alarm is already ringing. Most sellers open Seller Central to look at sales, not at the account health tab. And when they do open it, it is usually because an email arrived saying the account is at risk — which is exactly the worst moment to find out, because ODR is measured over a rolling sixty-day window, and that means the damage already inside does not get erased: it leaves the window slowly, on its own schedule, not yours.

The other half of the problem is scale. A seller doing 150 orders in two months and one doing 3,000 face the same percentage threshold, but not the same real tolerance. For the first, a single claim can eat most of the allowance. One bad month with a carrier, three annoyed buyers and one badly handled return are enough to put at risk an account that went two years without a single issue.

This post walks through exactly what makes up the ODR, which other indicators form account health, which thresholds are publicly documented in 2026, and what happens when you cross them. One warning before starting, and it applies to everything that follows: Amazon adjusts these thresholds, their measurement windows and their consequences fairly often, and not always identically across marketplaces. Everything here should be confirmed in Seller Central, under Performance and then Account Health, for the store you actually sell in.

what the order defect rate is and what goes into it

The Order Defect Rate is the percentage of your orders that ended with something Amazon counts as a defect in the customer experience. It does not measure returns, it does not measure product reviews, and it does not measure buyer-initiated cancellations. It measures three specific things, according to Amazon’s public seller documentation:

• Negative seller feedback: one- and two-star ratings. Mind the distinction, because it gets confused constantly: this is feedback about the seller, not a product review. A one-star product review does not enter the ODR.
• A-to-z Guarantee claims that were not denied: if the buyer files the claim and Amazon does not resolve it in your favor, it counts. A claim you resolve directly with the buyer before it escalates is a claim that never enters.
• Chargebacks: when the buyer disputes the charge directly with their bank or card issuer.

A single order with two defects counts as one defective order, not two. The denominator is the orders in the period. And that period, in the standard documentation, is a rolling sixty-day window. That rolling window is what hurts most in practice: by the time you find out you are in trouble, the numerator already contains orders from seven weeks ago, and the only way down is waiting for them to age out while you accumulate clean orders to fatten the denominator.

One thing worth underlining because it is almost always misread: ODR is not an FBM-only metric. An A-to-z claim or a chargeback on an FBA order counts too. What FBA does change is that Amazon absorbs the shipping metrics, not the order-experience ones.

the 1% threshold and why it is tighter than it looks

The documented threshold is to keep ODR below 1%. Written that way it sounds generous. Translated into orders it stops sounding generous.

If you did 300 orders in the last sixty days, 1% is three orders. Three one- or two-star seller ratings, three A-to-z claims or three chargebacks — or any combination of the three — and you are at the threshold. To stay on the safe side you have to be at two or fewer.

If you did 120 orders, a single defect puts you at 0.83% and two put you at 1.67%, well over the limit. That means a small seller is literally one claim away from the problem, all the time, and that their ODR will be far more volatile than a large seller’s. Not because they operate worse, but because percentage arithmetic with small denominators is that abrupt.

With 1,500 orders in the window, the same 1% is fifteen defects. There the number behaves like a trend and not like a lottery. This is why comparing your ODR to another seller’s without comparing volumes says nothing, and why a small seller needs a much earlier internal alarm than a large one.

Glossary: Buy Box, the buy box that decides who sells →

the other account health metrics and their thresholds

ODR is the central customer-experience metric, but account health is a set. These are the performance indicators that specialized guides and Amazon’s public documentation report in 2026, with the most commonly cited thresholds:

• Pre-fulfillment cancellation rate: below 2.5%. These are orders you cancel after receiving them and before shipping, usually because the stock you were showing did not exist.
• Late shipment rate: below 4%. Applies to orders you ship yourself, not to FBA.
• Valid tracking rate: above 95%. Also seller-fulfilled only: the share of shipments carrying a valid, trackable tracking number.
• On-time delivery rate: above 90%, again seller-fulfilled only.
• Invoice defect rate: below 5%, relevant if you sell to Amazon Business customers.

Three things are worth understanding about that list before memorizing it. The first is that three of those five indicators disappear if you run everything through FBA, because Amazon controls shipping and delivery. That does not make FBA automatically better — FBA fees carry their own cost, as the FBA versus FBM comparison shows — but it does shrink your account-risk surface.

The second is that these thresholds are documented mostly for Amazon.com, and the measurement windows vary by indicator: some are measured over 7 days, others over 10, 30 or 60. On Amazon Mexico the headline figures — the 1% defect rate, the 2.5% cancellation rate and the 4% late shipment rate — appear the same way in Seller Central materials, but the only valid source for your account is your own dashboard.

The third is that these rules move. A concrete 2026 example: as of February 28, Amazon changed how it applies the on-time delivery requirement for seller-fulfilled orders. Previously, falling below the threshold could deactivate all of your seller-fulfilled listings; with the change, deactivation applies to the listings contributing most to the drop, although a severe or repeated shortfall can still reach your whole seller-fulfilled catalog. That is exactly the kind of adjustment that gives any post about thresholds an expiration date, and the reason to read the original notice in Seller Central.

the account health rating: the grade that summarizes everything

Beyond the individual indicators, Amazon condenses account status into a single score, the Account Health Rating. Specialized 2026 guides describe it as a 0 to 1,000 scale that starts at 200 for a new account, rises with the volume of orders fulfilled without incident, and falls with each policy violation, with deductions weighted by severity.

The reported bands are three: 200 or above counts as healthy, between 100 and 199 the account shows as at risk, and below 100 it is exposed to deactivation. A critical violation — authenticity, product safety, intellectual property — can send the score to the floor without passing through the middle bands.

It is worth reading that score for what it is: a summary, not a diagnosis. Two accounts with the same grade can have completely different problems, one from performance metrics and another from policy violations. The grade tells you something is wrong; the indicator list tells you what. And like every aggregate score, it reacts late: by the time it drops a band, the problem has been running for weeks.

what happens when you cross a threshold

There is no single outcome, and that is precisely the part that generates unnecessary panic. The documented consequences form a ladder.

The first rung is usually a notice on the account health dashboard and an email asking for a plan of action. Then come partial restrictions: suspended ability to create new listings in certain categories, or deactivation of specific listings, which is where Amazon moved with the 2026 on-time delivery change. At the end of the ladder sits account deactivation and funds being held, the scenario everyone pictures and the least frequent one when there was an early reaction.

There is a 2026 development that changes the Buy Box conversation and deserves a careful explanation. For years the standard line was that a high ODR automatically removed you from the competition for the featured offer, because a seller-performance eligibility gate decided which offers could even enter the race. In July 2026 Amazon began retiring that gate: the rollout started in the United States at the beginning of the month, continued in Europe and the United Kingdom on July 20, and was announced for the remaining stores in phases. What is being removed is the entry gate, not the competition: Amazon was explicit that the criteria used to select the featured offer do not change, and landed price, delivery speed, fulfillment reliability, inventory availability and seller performance still carry weight there.

Put differently: a bad ODR no longer disqualifies you up front, but it remains one of the variables that loses you the comparison against another offer. And account health was never only about the Buy Box — listing restrictions and deactivation are still fully in place. Like everything else on this topic, the exact rollout status in your store has to be confirmed in the Seller Central announcements.

what a defect costs you before it touches account health

Before the compliance risk there is the cash cost, and it is usually much larger than sellers think because they never add it up in one place.

An example with invented but consistent numbers. You sell a product at 900 pesos, it costs you 600, and after commission, fulfillment fee, shipping and ads it leaves 120 pesos of net margin. One order ends in an A-to-z claim resolved in the buyer’s favor, with a full refund and no resellable product coming back. You lost the 120 of margin you had already counted, plus the 600 of goods cost, plus what you paid in shipping, say 80. That defect cost 680 pesos net. At 120 pesos of margin per order, it takes almost six clean orders to make it back.

If that order also came from a paid click, there is ad spend burned on a sale that unwound. And if the product did come back but as unsellable, it lands in the same bucket of hidden cost described in returns that kill your margin.

It is worth running this calculation once with your own numbers, because it changes priorities. An ODR of 0.9% sounds “within limits.” Translated to pesos, in a thin-margin catalog, that 0.9% can be costing more than the ad campaign you do review every week.

your own baseline and your early warning

Since Amazon’s threshold is a limit and not a target, your baseline has to be stricter than theirs. Three practical rules.

Set your own ceiling at half the threshold. If the limit is 1%, your internal alarm fires at 0.5%. That gives you room to react before Amazon writes to you, which is the only real difference between fixing a process and drafting a plan of action.

Translate the percentage into units every month. The concrete question: with the orders I do in sixty days, how many defects fit before I hit my own ceiling? If you do 400 orders and your internal ceiling is 0.5%, two fit. A small whole number gets watched; an abstract percentage does not.

Watch the causes, not the result. ODR is an output indicator: by the time it moves, the damage is done. The indicators that precede it are operational and you have them earlier: orders you were slow to confirm, shipments that left after the promise, orders cancelled because of stock that did not exist, response time to buyer messages. That last cause is unusually cheap to fix and unusually expensive to ignore, because a good share of A-to-z claims begin as an unanswered message.

And on the most common multichannel cause: overselling. Selling a unit that no longer exists because stock did not sync across channels produces either a pre-fulfillment cancellation or a late shipment, and both push the ODR up. It is not a customer service problem, it is a data problem, and it gets solved on the inventory side, not with apology emails. The mechanics are covered in what overselling is and why it happens.

mercadolibre measures something else with the same logic

If you sell on both channels, it helps to know that MercadoLibre runs its own reputation system and it does not translate one to one. MercadoLibre’s developer documentation describes a seller reputation built on claims, cancellations and delayed shipments, represented by a color thermometer that directly affects the visibility of your listings.

The practical differences are two. First, on Meli the weight of delayed shipments is central, while on Amazon that indicator disappears entirely if you run FBA. Second, the effect of a bad Meli reputation shows up first as lost exposure in search, not as an account-risk notice. The exact thresholds and measurement window for each indicator should be confirmed in MercadoLibre’s help center, because they change and differ by country.

What does carry from one channel to the other is the root cause. A badly synced catalog, stock that does not reflect what is really available, and a mismeasured lead time produce defects on both sides at once, under different names.

Glossary: real-time sync, stock that actually reflects reality →

how this reads in iqseller

iqseller does not replace Amazon’s account health dashboard: that data lives in Seller Central and that is where it should be read. What the panel does is attack the causes that generate defects, which is where a multichannel seller loses the fight.

The Inventory module keeps a single view of stock across Amazon, MercadoLibre, your warehouses and your 3PL, with real available separated from what is reserved and what is in transit. That is the direct lever against pre-fulfillment cancellations and against cross-channel overselling, two of the most frequent causes of an ODR that moves with no obvious explanation.

The Alerts module turns monitoring into something that does not depend on anyone’s discipline: a SKU falling below its critical stock, an inventory discrepancy between channels, or an anomaly in a product’s behavior surface on their own, before they turn into an order you cannot fill.

And the Profitability module is what lets you put a price on all of it. With net margin per SKU — calculated from the COGS you loaded and net of Amazon settlement commissions, MercadoLibre order commissions, FBA and Full fees, shipping, advertising, and the VAT and withholding breakdown — you can run the calculation from the previous section with your own numbers and know how many clean orders each defect costs you. At the Parent → Model → SKU level you can also see whether the defects concentrate in one specific variant, which is a very different diagnosis from “we have a service problem.”

closing thought

The Order Defect Rate is not a growth metric. It is the seatbelt, and it gets reviewed with that logic: not to improve it by a point, but to never be near the limit. A 1% threshold, a rolling sixty-day window and a small denominator make a combination that punishes late reaction far more than the occasional mistake.

The part you actually control sits almost entirely before the defect: synced stock, real availability, shipping promises met, and messages answered. And the part you do not control — the thresholds and the policies — changes often enough that the only reliable reading is your own Seller Central dashboard, checked every week instead of every crisis.