PulseHawk Measurement Framework

Which rules apply to every report?

• Marked values. Every value is marked counted, source or computed. Counted is our own count, source is a value we take over and name, computed is arithmetic from counted or source values.
• Window and n. Every value states its window in UTC and the n it rests on.
• Raw data. Every report publishes its data as a CSV, with the sha256 of the file and the method version.
• Fixed ID. Every report has a fixed ID, such as PH-UR-2026-001, and a permanent address under pulsehawk.io/reports.
• How to cite. Every report carries a citation in plain text and in BibTeX, with ID, title, date, address and method version.
• Neutral values. A report shows measured values only, with no grades and no words like good or bad.

What is checked before a report is published?

One shared set of checks runs for every report type. If a check fails, no page is built and we are alerted.

• Every value carries its marking, its window and its n.
• Every value on the page is computed again from the CSV and must match exactly.
• Percentiles rise in order, counts are never negative, and the window fits the event.
• Reference values such as a target interval come from two independent sources, a protocol constant and a release.
• No IP address, hostname, peer address or key appears on the page or in the CSV.
• The text follows our writing rules.
• A self test feeds the checks wrong data on purpose, and every check has to fire.

Every report is also reviewed before it is merged.

How do independence and corrections work?

The rules on our independence page apply to every report. Corrections are public and dated in the correction log.

Which report types are there?

• Upgrade Report. Measures a network upgrade on a PoW or DAG chain, from 24 hours before to 72 hours after the activation height, with seven fixed values.
• Confirmation Depth Report. Measures how deep reorgs reach on a chain, as a basis for how many confirmations cover them.
• Network Baseline. Measures block intervals, gaps, reorgs and peer software of a chain over a period without an upgrade, as the reference for later reports.
• Load Test Report. Measures how block intervals and inclusion times of a chain behave while a known transaction load runs on a test network.
• Anchor Verification Report. Checks independently whether data anchored on a chain is there, unchanged, at the stated height and transaction.

How do versions work?

Each specification has its own version. Every change to how a value is computed raises it. Old versions stay online at their own address, such as /standard/upgrade-report/v1-0, and a report names the version it follows and keeps it.

The framework changelog has one entry, dated 2 November 2026, the first version with five report types, one of them active.