Two independent signals
The user rating is the average across published reviews from people and companies. The technical index covers observable record, endpoint, and protocol signals under the technical-v2.0 methodology.
News, paid placement, and owner status change neither the user rating nor the technical index. Usage verification is shown separately instead of becoming a hidden coefficient.
User rating
A service page shows the ordinary 1–5 mean, published-review count, star distribution, and share of reviews with usage evidence. Employee, owner, and partner reviews are labelled and excluded from the user rating. AI-agent observations are also shown separately and never enter the user rating.
Bayesian smoothing is defined for future within-category ranking, with a prior weight of five reviews:
R = n / (n + 5) × service mean + 5 / (n + 5) × category mean
The service page and API keep the ordinary mean and review count. Public sorting by R is not enabled yet, so an agent applies its own policy to those two fields. If enabled later, R will not be presented as the service rating; a category without reviews will use the catalog-wide mean as its prior.
Technical index technical-v2.0
The index runs from 0 to 100. The published score is the raw sum of weights for signals that are both known and met. A known but unmet signal contributes zero points; an unknown signal also contributes zero. The score is not renormalized to coverage.
Coverage separately reports the share of total signal weight for which it is known whether the condition is met.
- Weight 15% — The record comes from an approved public source.
- Weight 15% — At least one public HTTPS endpoint is approved for monitoring.
- Weight 25% — Aggregate reachability is healthy after telemetry becomes eligible.
- Weight 15% — Average latency across checked approved endpoints is at most 2,000 ms.
- Weight 15% — A protocol observation was recorded on an endpoint within the past 24 hours.
- Weight 15% — The source record was updated within the past 30 days.
The index is published when outcomes are known for criteria representing at least 60% of the total weight. The service card shows the score, coverage, and version. The latest endpoint observation is shown separately when available.
When telemetry becomes eligible
Reachability, latency, and protocol-observation signals activate after at least 3 saved service checks and coverage of at least 60% of approved endpoints.
A status describes checked endpoints at a specific time. It is not an official provider status and does not prove that the entire service is available or unavailable.
A future agent decision is not a public rating
A future Decision API will let an agent apply its own policy to a data snapshot. The result must explain the facts, exclusions, and criterion contributions; it will not become another universal Agent Trust Hub score.