At a glance
Share of Amazon Prime Shipping consignments that hit any non-success scan event in the period: failed first attempt, customer-not-home, address issue, refused, damage, returned-to-sender. Exception is broader than “Prime promise miss” because some exceptions still deliver on-time after a re-attempt. The card surfaces operational noise that does not always show on the headline OTD but degrades Prime customer experience.
| What it counts | COUNT(DISTINCT shipments WHERE any_event IN ('FAILED_ATTEMPT','REFUSED','RTS','DAMAGED','UNDELIVERABLE')) / COUNT(shipments) over rolling 30 days. Each consignment counts once. |
| What counts as “exception” | Any non-success scan event from the carrier (UPS, USPS, OnTrac, Lasership, Amazon Logistics) in the merchant’s chosen Prime-eligible carrier mix. |
| Customer-fault vs carrier-fault | Pooled. ~60% customer-fault (recipient absent, address wrong, refused), ~30% carrier-fault (mis-route, depot delay), ~10% external (weather, force-majeure, customs for international Prime where applicable). |
| Service level scope | Prime-eligible consignments only. |
| Money-back-on-late interaction | Customer-fault exceptions do not affect SFP-eligibility scoring directly (Amazon excludes them from on-time calculation if properly tagged). Carrier-fault exceptions on Prime promises feed amazon_prime_promise_miss_rate. |
| Buy with Prime cohort | Buy with Prime customers tend toward fewer exceptions (the cohort is Prime-engaged and provides reliable delivery preferences) but disproportionately impacted by exceptions when they occur. |
| 2-day Prime coverage gaps | Rural-ZIP exceptions are over-represented in the headline because rural deliveries hit more access issues, weather delays, and address-issue events than urban deliveries. |
| Time window | 30D vsP |
| Alert trigger | >3% critical, >1% good. Amazon Prime Shipping cohort typically runs 1.0 to 2.5% (lower than Parcelforce / APC because of carrier mix + buyer-cohort selection bias). |
| Roles | owner, operations |
Calculation
Calculated automatically from your Amazon Prime Shipping (SFP) data. See the At a glance summary above for what the metric tracks and the worked example below for a typical reading.Worked example
A US health-and-beauty merchant on SFP shipping from Phoenix: $45 average order, mix of UPS Ground 2-Day to East and OnTrac to West Coast. Reading taken at 09:00 PT on 22 Mar 26 for the trailing 30 days (20 Feb 26 to 21 Mar 26).| Exception reason | Consignments | % of exceptions | % of total volume |
|---|---|---|---|
| Recipient absent (no safe place) | 88 | 41% | 0.7% |
| Address details required | 38 | 18% | 0.3% |
| Refused at door | 22 | 10% | 0.2% |
| Damaged in transit | 18 | 8% | 0.1% |
| Driver mis-route / depot delay | 32 | 15% | 0.3% |
| Weather / force-majeure | 18 | 8% | 0.1% |
| Total exceptions | 216 | 100% | 1.7% |
- Recipient-absent at 41% is below the carrier-network average (~50 to 60% on UK premium tier; less applicable here, USPS/UPS/OnTrac differ but pattern similar). Amazon Prime customers tend to be home or have engaged delivery preferences (apartment-package-rooms, secure lobby drops); the cohort selection helps.
- Damaged-in-transit 0.1% is healthy. Health and beauty packaging tends to be reasonably robust; below 0.3% benchmark.
- Driver mis-route / depot delay at 15% is the carrier-fault cohort that feeds
amazon_prime_promise_miss_rate. Track this slice; rising trend indicates carrier-side performance issue requiring carrier mix review. - Weather / force-majeure is seasonal. Storm-of-the-century events drive cohort spikes; Amazon excludes documented weather-events from SFP eligibility scoring if filed properly via the Seller Central case system.
- 1.7% exception rate against 1.0 to 2.5% benchmark is in spec. No urgent action; track refused-at-door cohort because it correlates with downstream A-to-Z claim rate (Amazon’s customer-protection program). A 1-point refused-at-door rise predicts 0.2 to 0.4 point A-to-Z claim rise at 10 to 14 day lag.
Sibling cards merchants should reference together
Exception rate is the broad operational-noise gauge. Pair with these:| Card | Why pair it with Exception Rate | What the combination tells you |
|---|---|---|
| Failed Deliveries | Subset specifically carrier-fault first-attempts. | Subtract failed-deliveries from exceptions for the customs/damage/access-issue bucket. |
| Prime 1-Day / 2-Day Promise Miss Rate | Carrier-fault subset that affects SFP eligibility. | Most carrier-fault exceptions feed promise-miss. |
| Returned to Sender | Terminal exception subset. | High exception + rising RTS = re-attempt logic failing. |
| On-Time Delivery Rate | Some exceptions still deliver on time after re-attempt. | High exception + high OTD = re-attempts working. |
Cross-connector: amazon.az_claims | A-to-Z claims downstream impact. | Refused-at-door + damage exceptions drive A-to-Z claims at 10 to 14 day lag. |
Cross-connector: shopify.refund_rate (for Buy with Prime) | Direct downstream for Buy-with-Prime cohort. | Prime-cohort refund rate is sharper-responding than non-Prime. |
Cross-connector: shipbob.sb_exception_rate | Peer 3PL benchmark. | Different orders entirely; comparable cohort behaviour for benchmarking. |
Reconciling against the vendor’s own dashboard
Where to look: Exception rate is not a native single Amazon-dashboard metric. The closest views: Seller Central → Performance → Account Health → Customer Service Performance for refused / claim events, and Performance → Shipping Performance → Late / Failed Shipment Rate for delivery-side exceptions. Carrier-side (UPS, USPS) portals expose their own exception dashboards which are the source-of-truth for delivery-attempt events. Why our number may legitimately differ:| Reason | Direction | Why |
|---|---|---|
| Distinct vs event-count | Ours lower | Carrier portals count each scan; card de-duplicates per consignment. |
| Customer-fault inclusion | Ours higher | Amazon’s eligibility-scoring excludes properly-documented customer-fault; the card includes all. |
| Time zone | Boundary | PST (Amazon) vs connected-timezone (card). |
| Card | Expected relationship | What causes legitimate divergence |
|---|---|---|
amazon.az_claims | Downstream. | A-to-Z claims involve more than late delivery. |
shipbob.sb_exception_rate | Peer 3PL. | Different orders. |
Known limitations / merchant FAQs
Why is the Prime Shipping exception rate lower than Parcelforce / APC? Selection bias plus carrier mix. Amazon Prime customers are more delivery-engaged (more likely to set apartment-package-room preferences, secure-lobby drops); the cohort generates fewer recipient-absent exceptions. Carrier mix (UPS, OnTrac, Amazon Logistics) favours large-network operators with mature exception-handling infrastructure. Customer-fault exceptions: do they count against SFP eligibility? Not when properly documented. Amazon’s SFP eligibility scoring excludes weather, force-majeure, customer-fault delivery refusals if the seller files a Plan-of-Action via Seller Central case management. The card includes all because all affect operations and customer perception; the SFP-eligibility-specific card (amazon_prime_sfp_otd_sla) tracks the Amazon-scored slice.
Damaged-in-transit on Prime drives A-to-Z claims; how do we mitigate?
Two paths. (1) Packaging review: Prime customers often have heightened expectations because Amazon’s branded packaging is well-engineered; merchant-side packaging that arrives bruised triggers refusal disproportionately. (2) Carrier choice: UPS Ground generally handles parcels more gently than USPS Priority for small-package volumes.
Refused-at-door climbing: what is the lead indicator?
Often an Amazon listing-side issue: stale product photos, sizing variance, customer expectation gap. A 0.3-point refused-at-door rise typically follows a SKU launch where listing copy did not match received product. Cross-reference with Amazon’s recent return-reason analysis for the affected ASINs.
Weather / force-majeure: how do we get them excluded from SFP scoring?
File a case in Seller Central within 48 hours of the event with documented context (NWS storm warnings, carrier-published service alerts, FAA advisories for air-freight). Amazon excludes these from the eligibility floor when accepted; appeals run 5 to 10 days.
Buy with Prime exceptions: are they pooled?
Yes the card pools both unless the merchant only has one product line connected. Buy with Prime customer-cohort tends toward fewer exceptions but disproportionately impacted by bad ones (Prime-trust signal is sharper-broken).
Q4 peak: what is realistic?
Q4 typically lifts exception rate 1.5 to 2.5 percentage points across all networks. Plan capacity in October; consider pre-emptive Prime-suspension during November-December if exception rate threatens SFP eligibility floor.
A single bad UPS depot: how do we identify?
Pair this card with ama_route_otd for geographic split. If exceptions cluster in 2 to 3 ZIP code prefixes that share a UPS hub-and-spoke route, the issue is depot-level. Account-team conversation with UPS or shift to Amazon Logistics for those ZIPs if available.
Compared to FBA-level exception rate, where does SFP land?
FBA typically runs lower exception rate (1 to 2%) because Amazon’s fulfilment centres are highly automated and use Amazon Logistics for most last-mile. SFP at 2 to 3% is comparable but consistently slightly higher; the gap is the merchant-side warehouse vs Amazon-side warehouse efficiency.
Single-warehouse SFP: are exceptions higher than 2-warehouse setup?
Yes structurally. Single-warehouse SFP has longer transit distances on average (more zone-distance per shipment), more weather and route-variability exposure, more chance of an exception per shipment. Multi-warehouse setups reduce exceptions by 0.5 to 1.5 percentage points.