In Q3 2026 (July 1 to September 30, 2026), StatusBird's independent 2-minute monitoring recorded 34 major or critical incidents across the 88 e-commerce services it tracks, adding up to 67.5 hours of combined downtime and a fleet-wide availability of 99.959%.
The full Q3 2026 report, with the ranked per-service table, the category breakdown, quarter over quarter deltas, incident timing, and the raw data as JSON and CSV, is at https://statusbird.io/reports/ecommerce-infrastructure-q3-2026. Every number on it is free to cite with a link.
How Q3 2026 compared with Q2 2026
Incident count was up from 32 to 34, a worsening of 2 quarter over quarter. Combined downtime was down from 181.0 hours to 67.5 hours, an improvement of 113.5 hours quarter over quarter. Fleet availability was up from 99.872% to 99.959%, an improvement of 0.087% quarter over quarter.
Reliability by category
Payments was the least available category of e-commerce infrastructure StatusBird measured in Q3 2026, at 99.878% availability across 11 services and 6 incidents. AI model vendors was the most available, at 100.0%. The full ranked table for all 17 categories is in the report.
The ten services with the most downtime in Q3 2026
| Service | Availability | Major/critical incidents | Downtime (minutes) |
|---|---|---|---|
| Plaid | 98.73% | 4 | 1,548 |
| Shippo | 98.9% | 10 | 1,344 |
| ActiveCampaign | 99.83% | 3 | 202 |
| BigCommerce | 99.85% | 2 | 180 |
| Cloudflare | 99.88% | 2 | 142 |
| Deliverr | 99.9% | 2 | 124 |
| Shopify | 99.9% | 2 | 124 |
| Zapier | 99.9% | 2 | 118 |
| Vertex | 99.95% | 1 | 64 |
| Sezzle | 99.95% | 1 | 62 |
The single longest outage of the quarter hit Plaid, running about 22.0 hours from August 02 at 14:52 UTC. Every StatusBird user monitoring that service was alerted within two minutes of the first failed check.
When the stack breaks
Across the 34 incidents in Q3 2026, the most common start time was the 18:00 UTC hour (5 incidents) and the most common day was Wednesday (12 incidents). Incident start times are recorded in UTC, so a store owner reading this in another timezone should shift accordingly before drawing a conclusion about their own peak hours.
The layer underneath
Grouping downtime by the infrastructure layer each service is known or widely reported to run on, the services sitting on AWS accounted for 60.1 hours of major or critical downtime in Q3 2026 across 59 monitored services. Of AWS's own 0 incidents in the quarter, 0 were followed within 30 minutes by an incident on a service that runs on it. That is a correlation, not a proven causal chain, and the report says so plainly.
AI model vendors
StatusBird began publishing availability for AI model vendors on September 23, 2026, so this category's window is far shorter than the rest of the report and the figures are not yet comparable with services measured for the full quarter. For the record: 4 vendors measured, 100.0% availability, 0 incidents.
How these numbers are measured
StatusBird polls every monitored service's official status endpoint every 2 minutes, around the clock, and writes one row per check. Q3 2026 rests on 4,899,898 such checks across 88 services. An incident is one continuous window of major or critical readings; consecutive major and critical readings without a recovery in between count as one incident, not several. Availability counts major and critical outages only, because several vendors report minor degradation almost continuously and folding that in would produce availability figures that describe their status-page style rather than their service. Degraded windows are reported separately: there were 409 degraded episodes in Q3 2026. Durations are accurate to within the 2-minute polling interval. The full methodology has the rest.
Why we publish this
Vendors publish their own uptime, on their own status pages, using their own definitions. Nobody independent publishes what actually happened across the whole stack a store depends on at once. That is the gap this quarterly report fills, and it is built from exactly the same checks that trigger StatusBird customers' alerts: there is no separate, prettier dataset behind the scenes. Start monitoring your own stack free, or take the numbers and cite them.