Comparison methodology

Cohort before score

Every comparison starts with a user, task, environment, date, and evaluated version. We classify each entity before comparing it. A consultancy, collector, database, managed platform, and composed stack may be related, but they are not automatically direct substitutes.

Evidence labels

• Documented: supported by a primary specification, repository, or current documentation.
• Measured: observed under a published method, environment, configuration, data set, and limitations.
• Vendor-stated: claimed by the supplier and not independently verified.
• Community-reported: attributed practitioner evidence whose scope may not generalize.
• Unknown: evidence is absent or insufficient. Unknown is never converted to an average score.

Quality characteristics must be defined for the decision context; standards such as ISO/IEC 25010 provide useful models but do not supply product rankings [1].

Compact graph scale

Page graphs use a 0–100 directional scale only when the dimension has an explicit meaning. The cohort, date, values, and method link remain visible. These graphs are navigation aids; the full matrix and sources control the conclusion.

Conflicts and corrections

Sponsorship, donated access, affiliation, and reviewer conflicts must be disclosed. Reputable disagreement remains visible with its version and scope. Material corrections update Markdown, the review history, and—where a conclusion changed—a correction note.