Skip to content
GridRouterhome

Search

Search providers, capabilities and pages

How the GridRouter Grid is scored

Two axes, both 0-100, both reproducible from published data. One formula, shared with the leaderboards and the router, lives in packages/quality.

Ability to Execute (vertical)

Measured from routed calls over the trailing 30 days: the same per-endpoint breakdown the leaderboards use, averaged per vendor across its endpoints in the category, weighted by calls.

InputWeightHow it is measured
Reliability25%Wilson lower bound (95%) of successful calls over all calls, rescaled so 95% scores 0 and 100% scores 1.
Fill20%Wilson lower bound of hits over successful calls. A hit follows each capability's published hit definition.
Accuracy20%Share of hits verified correct (benchmarks, verify-after-find, outcome feedback). Dropped and the rest re-weighted when a capability has no ground truth.
Speed10%Fastest p50 in the capability divided by this endpoint's p50.
Value15%Cheapest cost per verified hit in the capability divided by this endpoint's. Customer-plan tools (CRMs) score 1.
Probe recency10%1 when the last successful probe is within 15 minutes, falling linearly to 0 at 24 hours. Stands in for probe uptime until that ratio is published.

Completeness of Coverage (horizontal)

From each vendor's researched profile (catalog/vendors/<slug>.yaml, every fact sourced and dated) plus the capabilities it serves through GridRouter.

InputWeightHow it is measured
Breadth30%Capabilities in the category the vendor offers (documented or routed), over the most any vendor in the category offers.
Data coverage10%Vendor-claimed coverage counts (up to two) 50%, documented data types (up to five) 25%, stated geography 10%, stated refresh cadence 15%. Claims are never audited by GridRouter.
API completeness25%Access (self-serve 1, sales-led 0.6, partner 0.3, none 0) 35%, bulk/async/webhooks/MCP/sandbox 25%, documentation depth (docs, a spec, reference, rate-limit and changelog pages) 20%, official SDKs 10%, published rate limits 10%.
Pricing transparency15%Public list prices 0.8, or 1 with a free tier or a free API. Customer-plan tools 0.5. Contact sales only 0.2. Not found 0.
Compliance posture10%SOC 2 (Type II 1, Type I 0.7, stated 0.5) 45%, stated GDPR 20%, ISO 27001 15%, stated CCPA 10%, a published DPA 10%, as the vendor publishes them.
Integrations10%Native integrations the vendor publishes (CRMs, Zapier, Clay, …), up to six.

Quadrants and thresholds

A vendor needs 5,000 routed calls in the category, and at least one endpoint with 1,000 on its own, to be placed. Everyone else we researched is listed as Not yet rated with the reason, so absence from the chart is never silent.

With three or more rated vendors, each axis line sits at the category median, so a quadrant says where a vendor stands against this field today. With fewer, the lines sit at 50. A score exactly on a line counts as high. Charts zoom each axis to the field; the tables carry the unzoomed numbers.

  • Leaders

    High execution, broad coverage

    Measured results are strong and the vendor covers most of what the category asks for, with an API you can buy and audit.

  • Performers

    High execution, narrower coverage

    Measured results are strong on the jobs they cover. Fewer capabilities, thinner API surface or less transparent pricing keep them left of the line.

  • Visionaries

    Broad coverage, execution still proving out

    Wide capability and API coverage, but routed traffic has not yet shown results at the level of the category's best.

  • Specialists

    Focused niche

    Narrow by design or early in our data. Often the right pick for one job; check the scores for that job before routing to them.

Snapshots and movement

Every report carries its snapshot date. Each publish keeps the previous placements, so a report shows how far every vendor moved on both axes and lists every quadrant change, entry and exit. Scores recompute from the catalog snapshot on every publish; nothing is adjusted by hand. Each category's scores download as CSV from its report.

Disclosures

Sample data. Until live traffic accrues, Ability to Execute is computed from GridRouter's sample catalog. Every report says so at the top while that is true. Coverage inputs are real research with sources and check dates.

No paid placement. No vendor pays for placement, and nothing a vendor buys from Relaygrid changes its inputs. Vendors cannot edit their scores; they can correct a sourced fact through the vendor program, and the correction is dated like every other source.

First-party services. Relaygrid's own utilities (fetch, suppression, do-not-call checks) are never placed on the Grid.

Not a review. The Grid publishes measurements, not ratings. Pages emit dataset structured data only, never review or rating markup.

Unknowns stay unknown. Facts we could not confirm are recorded as unverified and score zero; whether a vendor allows resale and how it stores data stay unknown until a written agreement says otherwise.