Products
Tides
Operational- 30 days
- 100.00%
- 90 days
- 100.00%
- 24h p95, synthetic probe (1/min)
- 57 ms
- 24h p95, customer traffic
- 77 ms
Waters
Operational- 30 days
- 100.00%
- 90 days
- 100.00%
- 24h p95, synthetic probe (1/min)
- 125 ms
- 24h p95, customer traffic
- 10018 ms
Heavens
Operational- 30 days
- 100.00%
- 90 days
- 100.00%
- 24h p95, synthetic probe (1/min)
- 116 ms
- 24h p95, customer traffic
- 16260 ms
at or above 99.9% 99 to 99.9% below 99% no data
Open incidents
No open incidents.
Scheduled maintenance
No maintenance scheduled.
Past incidents (30 days)
No incidents in the last 30 days.
How we measure
A synthetic probe queries every product once a minute with a fixed request against the same serving path customers use. The current state of a product is read from the last 5 minutes of probes and from any open incident recorded against it.
A minute counts as available for a product when the synthetic probe returned a value in that minute, or, when there was real traffic, when the product served at least one non-5xx response. A minute with neither a probe nor traffic is not counted at all. Availability is available minutes divided by counted minutes.
Latency percentiles are taken from the probe. Where the product also served real traffic in the window, the p95 of the measured handler time for those requests is shown separately and labelled as customer traffic. Percentiles are never mixed between the two sources.
Daily cells and windows longer than seven days come from a per-day rollup of the same measurements; the last 24 hours and 7 days are computed directly from the raw samples. Days without a rollup show as no data.
Targets
| Product | Availability target | Latency target (p95) | Measured path |
|---|---|---|---|
| Tides | 99.9% | 3000 ms | Tides synthesis path (cache miss) |
| Waters | 99.9% | 800 ms | Waters grid lookup |
| Heavens | 99.9% | 800 ms | Heavens grid lookup |
Targets are the levels the service is operated against: the availability target from our monitoring, and the latency level a single probe may not exceed without being logged as an anomaly.
Service commitments
What we commit to while you hold an active plan, on a reasonable-efforts basis. There is no credited service-level agreement: no service credits, no penalties, and no binding uptime, availability or response-time guarantee. The binding text is the Terms (Section 14, Service commitments and warranty disclaimer); this is a summary of it.
- Availability
- We aim for high availability of our serving infrastructure. Any availability figure we publish, including the targets on this page, is a target, not a guarantee, and carries no service credits or penalties.
- Freshness
- For each live product we re-compute and publish on the cadence described in the documentation, provided the upstream sources are themselves publishing normally.
- Honesty and provenance
- We do not fabricate values: where a source does not cover a location or time the value is returned as null, each response reports its data age and the sources behind it, and we serve only data we are licensed to redistribute.
- Interface stability
- The API is versioned. We do not make a breaking change to a served version without reasonable prior notice and, where practicable, a migration path.
By product
- Tides (data we own and run)
- Synthesized from a tidal model we hold and run on our own infrastructure, with no per-request dependency on a live third-party feed. We additionally commit to continuity, reproducibility (the same request returns the same result over time) and validated quality against independent tide-gauge observations, published in the documentation.
- Waters, Heavens and Lands (data we source)
- Derived from third-party open data. We do not guarantee that any source remains available, nor the accuracy, completeness or continuity of sourced data. A source becoming unavailable, changing or being removed is an availability matter, not a breach; where a product blends several sources, the loss of one typically degrades resolution rather than removing the product.
Retiring a live product or the Service is our own decision and comes with at least 90 days notice and a pro-rata refund of prepaid fees. A temporary interruption of a sourced product is an availability matter shown on this page, not a discontinuation.
This page is also available as JSON at /api/status.