From product fit to evidence to ranking.

L1: establish task, audience and industry fit

We identify the individual product and the concrete job it supports, then determine the audience and, for business tasks, the industry. A company category, a possible use or evidence about another product does not establish this product's fit.

Direct workflow evidence determines the scope of the comparison. Adjacent capabilities, partial workflows and insufficient evidence remain qualified. Products are not admitted to every industry simply because they appear in a cross-industry comparison. L1 classification establishes relevance; it is not a final ranking score.

L2: connect evidence to the product and scoring pillars

We attribute evidence to the relevant product, task, industry and source. Supporting passages retain their original links and qualifications. Company-reported outcomes, company-selected testimonials and third-party reviews remain distinguishable, including mixed and negative experiences.

Evidence is assessed for the applicable pillars. Repeated copies of one story are not separate customer outcomes, and evidence for one product or use case does not automatically transfer to another. Missing information remains unknown. Displayed evidence samples are selected examples rather than the full evidence record.

Ranking: combine six pillars into 100 points

The six-pillar scorecard combines task and industry fit with supported customer signals. These are the maximum contributions of each pillar to the total; they are not customer star ratings or benchmark performance measurements.

Read observed evidence, partial support and unknowns separately

The score combines task fit, industry fit and supported customer signals. Task and industry fit account for 65 points. Partial evidence marks a calculated fit score with unconfirmed steps; it is not a fully established assessment. Unknown means there is insufficient evidence to score that pillar; unknown fit uses a neutral value in the total. Evidence samples are selected examples, not a complete evidence record.

A pillar's stored contribution to the total and its observed evidence value are different fields. An operational value used for an unresolved pillar is not an observed measurement. Read the score basis and source qualifications alongside the number. Value signals describe reported benefits; they do not establish comparable value for money. Popularity measures interest rather than product quality or customer satisfaction.

Use the published rank and tie information. The order of two equal-ranked entries does not establish that one is better. Scores compare products within the stated task and industry or audience; they do not establish a universal company ranking or guaranteed business outcome.

Use the method and date of the specific edition

Each ranking shows its own scoring method and update date. The six-pillar method is explained below. Rankings that retain an earlier method, including dated editions, keep their original scorecards and weights.

The current publication and a dated edition answer different questions. Earlier five-criterion rankings and earlier evidence editions retain their recorded scores and methods; the six-pillar weights above must not be applied retrospectively. A local preview is not a published edition. Product scores and a whole-ranking review are separate assessments.

Placement is never paid for. Submitting evidence does not guarantee a higher score. Sources and missing information should remain visible when a ranking is quoted.

• Methodology and publication updates →

  Read the recorded changes and edition history.

Research benchmark results use a different scoring system

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

Business evidence scores and research benchmark metrics have different meanings and must not be compared numerically.

• Research benchmarks →

  Find results by dataset, task and metric.