# Execution Provider Comparison Methodology

Version: 1.0
Last reviewed: 2026-07-27

Compare products only when they solve the same layer. A local execution client,
a private transaction route, a block engine, an RPC provider, and a hosted
interface are not interchangeable categories.

## Required evidence

For every candidate, record:

- current official product and API documentation;
- signing, custody, permission, and revocation model;
- supported chains, transaction types, and documented exclusions;
- privacy, retention, fallback, cancellation, and status behavior;
- complete fee basis and externally imposed network costs;
- release provenance, update process, support, and incident procedures;
- source URL, publisher, review date, and the exact statement supported.

## Fair test rules

Use the same chain, transaction shape, device class, region, observation window,
failure scenarios, and definitions. Publish raw observations and aggregation
code before making a measured performance comparison. Do not infer profit from
latency, simulation, or inclusion.

## Publication rules

Separate documented facts from measurements and opinions. Do not assign a
"best" ranking without a reproducible method. State limitations, avoid expected
returns, and recheck provider documentation on the decision date.