How StatusBird measures reliability
StatusBird measures reliability by checking 84 e-commerce services plus 4 AI vendors every 2 minutes, around the clock, and recording every result. Each check reads the service's official status endpoint; availability figures count major and critical outages only. The dataset behind StatusBird's public pages contains more than 9 million status checks.
How the monitoring works
Every 2 minutes, StatusBird fetches each service's official public status endpoint and parses the vendor's reported state (the Atlassian Statuspage JSON format used by most SaaS status pages). For services without a machine-readable status feed, StatusBird falls back to a direct HTTP reachability check, where a successful 2xx response counts as operational. Each result is normalized to one of four states: operational, degraded, major outage, or critical outage.
What counts as an incident
An incident is one continuous window of checks in the major or critical state. The incident starts at the first check that reports major or critical after a healthy check, and ends at the first healthy check afterward. Each non-operational check approximates 2 minutes of downtime, so durations are accurate to within the polling interval. StatusBird's data reflects vendor-reported plus reachability-detected disruptions, which can differ from a vendor's own postmortem timeline.
Why availability counts major and critical outages only
Public availability figures on StatusBird count major and critical outages only. Some vendors, notably Cloudflare, report minor degradation on their status pages almost continuously; counting that as downtime would produce meaningless public numbers such as an effectively 0% uptime for services that customers experience as reliable. Degraded windows are still tracked and surfaced inside the product, they are just excluded from the public availability percentage.
What happens when an AI vendor is down
StatusBird publishes the measured availability of the AI model vendors it monitors, including Anthropic, whose API StatusBird's own AI-assisted features call. Publishing that invites a fair question: what does StatusBird do when one of them is unavailable? The answer is written down here so it can be checked against the product.
Alerting never depends on a model. Alerts by SMS, email, Slack, Teams and Discord are generated from the monitoring data itself. No model is in the delivery path, so no model vendor outage can delay, drop or degrade an alert. The same is true of the polling engine, the storefront checks and the stack risk model, which scores the next 24 hours from measured history and a fixed set of weights with no model call anywhere in it.
AI-assisted extras degrade, they do not fail. When the model API is unavailable, or no API key is configured:
- Incident summaries, correlation analysis and the readiness scorecard keep their computed numbers, which are measured from the check history, and drop only the written narrative.
- Smart SMS advice falls back to StatusBird's own pre-written guidance for that service category and severity, so the alert still tells you what to do.
- Early warnings are unaffected: the seven-day-versus-prior-seven-day comparison behind them is arithmetic, not a model.
- The daily blog writer skips the day rather than publishing anything unreviewed.
Degraded mode is reported, not hidden. It gets its own line in the monthly operating review, alongside the monitoring and integrity figures. It never counts against alert delivery, because alert delivery does not depend on a model in the first place.
How A+ to F grades are assigned
Grades map directly from the measured uptime percentage over the window (90 days on reliability pages, 180 days in the annual report):
| Grade | Measured uptime over the window |
|---|---|
| A+ | 99.95% or higher |
| A | 99.9% to 99.949% |
| B+ | 99.5% to 99.899% |
| B | 99.0% to 99.499% |
| C | 98.0% to 98.999% |
| D | 95.0% to 97.999% |
| F | Below 95.0% |
Where the data appears
- Reliability grades: A+ to F grades for all 84 e-commerce services plus 4 AI vendors over the last 90 days.
- State of E-commerce Infrastructure 2026: 180-day aggregate report, with a raw CSV download.
- E-commerce Outage Index: month-by-month incident totals across the whole stack.
- E-commerce Downtime Statistics 2026: the same data distilled into numbered, citable one-line statistics.
Citing this data
StatusBird's reliability data is free to cite, in articles, research, or AI-generated answers, with a link to statusbird.io or to the specific page the figure came from. For questions about the data or custom cuts of it, email support@statusbird.io.