Q3 2026 compared with Q2 2026

Both quarters are measured the same way, by the same checks, at the same 2-minute interval.

Reliability by category

Payments was the least available category StatusBird measured in Q3 2026, at 99.878% across 11 services. AI model vendors was the most available, at 100.0%.

Bars show the downtime share, which is what varies; each bar is labeled with the availability figure it implies. Longest bar = 0.122% of checks. July 1 to September 30, 2026.

The ten services with the most downtime in Q3 2026

The ten services with the most incidents in Q3 2026

A different list from the one above on purpose: a service can break often and recover fast, or break once and stay down for hours. Both are worth knowing about, and they call for different responses.

Best availability among services that had at least one incident

Services with a perfect quarter are excluded from this table, because a service that never broke tells you nothing about how well it recovers. These are the services that did break and still finished the quarter near the top.

The longest single outage of the quarter

When incidents start

Across the 34 incidents in Q3 2026, the most common start hour was 18:00 UTC (5 incidents) and the most common day was Wednesday (12 incidents). These are counts of incident starts, not of downtime minutes, so one long outage does not outvote twenty short ones. All times are UTC.

Tallest bar = 5 incidents. July 1 to September 30, 2026.

The layer underneath: infrastructure cascade

Most of the services a store depends on are themselves running on a handful of platform layers. This table groups downtime by the layer each service is known or widely reported to run on, and counts how often a layer's own incident was followed within 30 minutes by an incident on a service that sits on it. This is a correlation, not a proven causal chain. The dependency map is approximate and assembled from public reporting, and two services breaking close together can share a cause without either causing the other.

AI model vendors

Across 4 model vendors and 228 checks since September 23, 2026, availability was 100.0% with 0 major or critical incidents.

Studio and the AI-Built Store Index

StatusBird also runs an independent scanner for AI-built storefronts. These figures cover the last 30 days rather than the full quarter, and come from the index host rather than the monitoring database. The current AI-Built Store Index edition (2026-09-21) covers 234 stores, of which 52.6% were empty shells. 1 free store checks were run in the window.

Full results: all 88 services

Every service measured in Q3 2026. Also available as raw CSV and JSON.

Methodology

Sample size. This report rests on 4,899,898 individual status checks across 88 services, recorded between July 01, 2026 and September 23, 2026. Nothing is sampled, modelled or extrapolated: every figure is a count over that set of rows.

How the data is collected. StatusBird polls each monitored service's official status endpoint every 2 minutes, around the clock, and writes one row per check. Where a vendor publishes a machine-readable status page, the reading is that vendor's own declared state; where one does not, the reading falls back to whether the endpoint responded successfully. These are the same checks that trigger StatusBird customers' alerts. There is no separate, tidier dataset behind the scenes.

What counts as an incident. An incident is one continuous window of major or critical readings for a single service. A service that slides from major to critical and back without ever recovering in between counts as one incident, not three. Two separate outages on the same day count as two.

Why availability counts major and critical only. Several vendors, Cloudflare most conspicuously, report minor degradation on their status pages almost continuously. Folding that into availability would produce figures that describe a vendor's status-page habits rather than its service, and would rank the most transparent vendors worst. Degraded readings are therefore excluded from availability and reported separately in the "Degraded" columns above, where 409 degraded episodes were recorded this quarter.

How downtime is calculated. Each non-operational check stands for one 2-minute polling interval, so downtime is the count of major or critical checks times two minutes. Durations are therefore accurate to within the polling interval, and a reported incident length can differ from a vendor's own postmortem.

How availability is averaged. Fleet and category availability are weighted by check count, not by service. A service added partway through the quarter contributes in proportion to how long it was actually measured, so a newly added vendor cannot swing the headline number.

Known limits. Availability measured from a vendor's status page reflects what the vendor chose to declare and when; an outage a vendor never acknowledged is not in this data. Timestamps are UTC throughout. Dependency mapping in the cascade section is approximate. Read the full methodology for the rest.

Frequently asked questions

What period does the Q3 2026 e-commerce infrastructure report cover?

It covers July 1 to September 30, 2026, measured continuously at a 2-minute polling interval across 88 e-commerce services, for a total of 4,899,898 individual status checks. The quarter is still in progress, so the figures update as it fills in.

How many outages hit the e-commerce stack in Q3 2026?

StatusBird recorded 34 major or critical incidents across the 88 services it monitors in Q3 2026, totalling 67.5 hours of combined downtime and leaving fleet-wide availability at 99.959%.

Which category of e-commerce infrastructure was least reliable in Q3 2026?

Payments, at 99.878% availability across 11 services and 6 incidents. The most reliable was AI model vendors at 100.0%.

Why does this report exclude minor degradation from availability?

Because some vendors report minor degradation nearly all the time, and counting it would measure how candid a status page is rather than how reliable the service is. Degraded windows are still tracked and are reported separately: 409 degraded episodes in Q3 2026.

Can I cite or reuse this data?

Yes. The figures are free to quote, chart or redistribute with a link to this page. The full per-service table is published as CSV and JSON under a Creative Commons Attribution 4.0 licence, specifically so the underlying data is citable and not just the prose.

How does this differ from the annual report?

The annual report covers a rolling 180-day window and answers "how is the stack doing lately". This quarterly report covers a fixed calendar quarter and compares it with the quarter before, which is the view you want for tracking a trend or citing a specific period.

Cite this report

Related: the annual State of E-commerce Infrastructure report, the monthly E-commerce Outage Index, per-service history on the reliability index, and every StatusBird report in one place.