Skip to content
Robnu
Amazon · Operations

What the Amazon order defect rate actually counts

ODR is the metric most likely to deactivate an account and the one most often misread, because three different events feed a single number and they do not behave alike.

Free during early access · Forever free under 25 orders/day
Updated 11 October 2026

ODR combines negative feedback, A to Z guarantee claims and credit card chargebacks into one rate against orders in a period. A refund on its own does not count, a resolved A to Z usually still counts, and the number moves on a lag because claims arrive weeks after the order. The ceiling commonly quoted is market convention, since no public Amazon.in page states it.

TL;DR
  • Three inputs: negative feedback, A to Z claims, chargebacks. One number.
  • Refunding a buyer does not by itself remove a defect that is already counted.
  • The rate lags, because claims land weeks after the order that caused them.
  • A small seller's rate is volatile: one claim on few orders is a large percentage.
  • The threshold everyone quotes is convention; your Account Health dashboard is authoritative.

The order defect rate is the account-health metric with the most direct path to deactivation, and most of the confusion about it comes from treating it as a single kind of event when it is three.

The three inputs, which behave differently

Negative feedback is a buyer rating the transaction poorly. A to Z guarantee claims are a buyer escalating to Amazon for a refund or resolution. Chargebacks are a buyer disputing the charge with their card issuer. One order can produce more than one of these, and they arrive on very different timelines.

Treating them as one thing is what makes the metric feel arbitrary. Feedback arrives soon after delivery. An A to Z claim typically arrives once a buyer has given up waiting, which is later. A chargeback can arrive much later still.

The lag, and why the number looks wrong

Because claims arrive after the orders that caused them, the rate you are shown reflects a period whose orders are already in the past. A seller who fixed a dispatch problem two weeks ago can watch the metric worsen for a while afterwards, which feels like the fix did not work.

The practical consequence is that you cannot judge a correction by the metric moving immediately. Judge it by whether new defects are being created, which you can see from the dates on the individual claims rather than from the headline rate.

What a refund does and does not do

Refunding a buyer is often the right thing commercially and it does not automatically undo a defect that has already been counted. An A to Z claim granted against you generally still counts even where you then refund, and negative feedback stands unless it is removed under the specific conditions Amazon allows.

So the sequence matters more than the generosity. Resolving a buyer's problem before they escalate prevents a defect; resolving it afterwards mostly prevents the next one.

Small sellers and volatility

The rate is a proportion, so at low order volume a single claim is a large percentage and the metric is inherently jumpy. A seller with few orders a week can cross a threshold on one bad transaction while a large seller absorbs the same event invisibly.

This is worth knowing because it changes the right response. For a low-volume account the useful work is preventing the single claim, since there is no volume to dilute it, and that is mostly about communication before a buyer escalates.

The threshold, and why we will not state it as policy

The ceiling for ODR is quoted confidently almost everywhere and it is not available on any Amazon.in page you can open. Checked on 14 October 2026, both the seller-performance and account-health paths on sell.amazon.in return 404, and the detailed guidance sits inside Seller Central behind a login.

Your own Account Health dashboard shows the target that applies to your account, and that is both authoritative and current for you. Treat any figure stated elsewhere, including the ones repeated most often, as convention.

Where this sits in the rest of your Amazon operation

Robnu is read-only on Amazon today: orders, returns, payouts and inventory sync so you can see and reconcile them, and it cannot accept Amazon orders or make their labels. Order processing is live for AJIO and Meesho. Related reading:

Sources & further reading

  • Amazon India seller performance page checked 14 October 2026 and it returns 404, as does the account-health path. No public Amazon.in page states these thresholds, which is why they are described here as market convention.
  • Amazon India Seller University the public training hub. It serves, and it does not publish the account-health thresholds.
  • Amazon listing and image style guides login-gated. It redirects to Amazon sign-in, so the detailed rules it holds are not quoted here and are described as market convention instead.

Not legal or tax advice. Amazon changes fees, windows and policies without notice, and the detailed style guides sit behind a Seller Central login, so any figure quoted here as convention is not Amazon-published. Check anything you plan to act on against your own Seller Central account or flat-file processing report first.

Questions sellers ask

Three things: negative buyer feedback, A to Z guarantee claims, and credit card chargebacks, combined into one rate against orders in a period. One order can produce more than one of them, and they arrive on very different timelines, which is why the metric feels arbitrary when treated as a single kind of event.

Not by itself. An A to Z claim granted against you generally still counts even if you then refund, and negative feedback stands unless removed under the specific conditions Amazon allows. Resolving a problem before the buyer escalates prevents a defect; resolving it afterwards mostly prevents the next one.

Because the metric lags. Claims arrive weeks after the orders that caused them, so the rate reflects a period already in the past. Judge a correction by whether new defects are still being created, which you can see from the dates on individual claims, rather than by the headline rate moving.

Because it is a proportion and you have low order volume. One claim against few orders is a large percentage, so a small seller can cross a threshold on a single bad transaction that a large seller would absorb invisibly. At low volume the useful work is preventing the single claim rather than diluting it.

It is not stated on any Amazon.in page you can open. Checked on 14 October 2026, the seller-performance and account-health paths both return 404 and the detailed guidance is behind a Seller Central login, so the figure quoted almost everywhere is market convention. Your own Account Health dashboard shows the target that applies to you.

Keep reading

Related seller guides

More on the operations, money and claims that decide whether a marketplace catalogue actually makes money.

Writing an Amazon plan of action that gets read

A plan of action has three parts and most rejected ones are missing the third. The structure Amazon expects, what counts as evidence, and the habits that get an appeal dismissed.

Amazon cancellation rate, only your cancellations count

A buyer cancelling does not hurt you; you cancelling does. Why the metric is almost always a stock accuracy problem, and the two habits that quietly generate it.

Unicommerce Alternatives in India: Honest Options by Seller Shape (2026)

What Unicommerce is genuinely good at, why sellers look elsewhere, and the real alternatives by seller shape. Only three order management products in this market publish a price you can compute.

Meesho manifest and handover mistakes that trigger penalties

Closing the manifest, the handover scan, and the gap between them is where Meesho penalties hide. The mistakes that trigger them, manifested-but-not-handed-over, missing scans, wrong counts, and how to close cleanly.

Meesho OMS: The Honest Buyer's Guide for Sellers

What an order management system for Meesho actually is, what to look for before you commit, the questions that separate real tools from dashboards, and where Robnu fits in as one honest option.

Meesho orders stuck in pending: what it means and how to clear them

Blocked does not mean paused, the dispatch clock usually keeps running. The four causes, which two you can fix yourself, and what to document when you cannot.

AJIO payout doesn't match your orders? Reconcile it

Build the expected line per order, tag every delta, reversals, freight slabs, adjustments, tax lines, separate cycle lag from real mismatches, and raise one ticket that sticks.

How many images should an Amazon listing have

Amazon gives you more image slots than it shows a buyer. What Amazon actually publishes about image counts, what the market treats as settled, and where the honest answer runs out.

build 0ad8ad1c505b1affc99f3e66e10b18b87ab43281 · 2026-10-14T08:38:17Z