Amazon’s Account Health dashboard is a scoreboard with consequences: offers suppressed, features gated, and, sustained badly enough, the account itself paused. The good news is mechanical: each metric is fed by specific, fixable operational failures, and none of them require heroics, just processes that do not leak.
The four that matter most
Order defect rate (ODR), target under 1 percent. Negative feedback, A-to-z claims, and chargebacks over total orders in the window. The feeders: item-not-as-described (listing honesty, data quality), item-never-arrived (tracking and carrier choice), and disputes escalating because messages went unanswered. ODR is the composite the others flow into.
Pre-fulfillment cancel rate, target under 2.5 percent. Seller-caused cancellations before shipment, overwhelmingly stock you promised and did not have. The feeder is stale counts: a unit sold elsewhere while Amazon’s number lagged. This is the metric real-time sync exists to protect, and where buffers on thin SKUs buy insurance.
Late shipment rate, target under 4 percent. Orders ship-confirmed after the expected ship date. Feeders: promise inflation (handling times set to ambition, not reality) and queue failures, orders found late ship late.
Valid tracking rate, target 95 percent+. Shipments with carrier-scanned tracking. Feeder: manual fulfillment that skips or delays tracking upload, automation per channel makes this metric free.
Around them: policy compliance (IP complaints, authenticity, restricted-category violations) scored on the same dashboard, where the currency is documentation, real supply chains produce real paper.
Reading the dashboard like an operator
- Metrics are trailing windows (10-60 days depending on the metric): today’s fix shows up as the window rolls. Do not confuse lag with failure.
- One incident in low volume looks like a crisis: at 50 orders, a single defect is 2 percent. Watch the absolute counts alongside the rates while small, and grow the denominator, volume is the best dilution.
- Every red flag has a case log: work items name the order and the reason, the same reason-code discipline as inventory adjustments. Fix the process the incident names, not the incident.
The operational spine under green metrics
The same short list keeps every marketplace scorecard healthy, eBay’s version is the same story with different thresholds:
- Stock truth in seconds across every channel, so cancellations from phantom availability stop existing.
- Promises set to measured reality: handling time you hit on an ordinary Tuesday.
- Tracking uploaded automatically at dispatch.
- Messages answered inside 24 hours, disputes are usually escalation failures.
- Listings that tell the truth, the returns log’s reason codes will tell you where they do not, read them.
Common questions
Which metric gets sellers suspended fastest?
Sustained ODR above threshold and policy violations (authenticity complaints) move fastest. Cancellation and lateness usually degrade features and Buy Box share before they threaten the account.
Do FBA orders count against my metrics?
FBA shifts fulfillment-driven metrics (lateness, tracking) to Amazon’s side, one of FBA’s real trades. Listing honesty and policy compliance remain yours everywhere.
How fast can a bad metric recover?
As fast as its window rolls: stop the feeder today, and a 30-day metric is clean in 30 days of good orders. Volume accelerates the math.
Is there any appeal for wrong strikes?
Yes, defects and violations have dispute paths, and documented cases win them: tracking scans, message logs, invoices. The operator’s habit is keeping the evidence before needing it.
Protect the scoreboard at the source
Real-time stock truth, buffers, one deadline-sorted queue, and the sync trail behind every count, from $49/month with unlimited orders. See pricing.