The RELIASTRA research agenda

We publish research for one reason: a reliability record is only useful in a commercial argument if the other side can inspect how it was produced. This page states what we intend to publish, what we will not, and the rules we hold ourselves to.

What we will publish

• Methodology. How checks are scheduled, origination, retries, quorum rules, and the failure taxonomy. This is the minimum required for anyone to evaluate our numbers, so it is published first and kept current.
• Aggregate reliability observations. Posture for vendors whose data has been made public, with the window and sample size stated alongside every figure.
• Negative results and corrections. When a measurement turns out to be wrong or misleading, the correction is published in place with the original claim still visible.

What we will not publish

• Anything that identifies a customer, their endpoints, or their configured checks. Public material shows aggregated posture only.
• Vendor rankings presented without their sample size, window and methodology. A number without those is marketing.
• Causal claims. We report correlated timelines and leave causation to the engineers reading them.
• Benchmarks we cannot reproduce from the method we published.

Standing rules

• Every published figure states its window, its origin set and how outcomes were classified.
• Where we lack the data to support a claim, we say so explicitly instead of estimating.
• Evidence reports remain the property of the organisation that generated them, and public verification confirms only that a report exists and matches its checksum.

Current status

The methodology is published and complete for the check pipeline as it exists today. Aggregate vendor reporting is limited to vendors whose posture has been made public; we have not published a cross-vendor comparison, and we will not until the sample and window behind it can be stated honestly.