One formula. Every tool.

The criteria

• Target fit (25%): whether the tool's published scope covers this website and entity type.
• Field coverage (30%): the share of fields this use case needs that the tool's published output schema carries as a key of its own (a "rating" key or "totalScore", not "ratingHistogram"; nested keys such as location.lat count; keys about another party, such as reviewerName, do not stand in for the record's own name). A tool with no published schema receives a low prior of 0.2 because coverage cannot be verified; the same prior applies to every such tool.
• Price (20%): published price per 1,000 results, normalised against the other tools on the same page. A tool with no published per-result price receives a neutral 0.4; it is not treated as free.
• Reliability (13%): 30-day public success rate when the platform publishes it and at least 20 public runs stand behind it. A tool with fewer recent public runs receives a neutral 0.5, so a perfect rate from a handful of runs cannot lift a tool.
• Adoption (7%): public user count, log-scaled, relative to the tools on the same page.
• Freshness (5%): months since the published build was last updated. Missing update dates receive a low prior of 0.2.
• The same formula runs for every tool. Ties break by public user count, then lower published price, then title. There is no owner preference in the formula.

Target fit

We score whether the tool's public description and taxonomy say it covers the target website and the entity type on the page. A tool that covers LinkedIn companies is not scored as a LinkedIn Jobs tool just because both are on linkedin.com.

Field coverage

For each target page we list the fields the use case needs. The tool's published output schema is checked for those fields. A tool with no public schema receives a low prior of 0.2 because coverage cannot be verified; the same prior applies to every tool without a schema.

Price

Price is the published primary-result rate per 1,000 records. On a page with several priced tools, the cheapest gets the highest price score and the most expensive gets the lowest. A tool with no published per-result price receives a neutral 0.4; it is not treated as free.

Reliability

Reliability is the public 30-day success rate on Apify, calculated from the platform's public run statistics. A tool with no recent public runs receives a neutral 0.5. User reviews are not used in the score.

Adoption and freshness

Adoption is the public user count, log-scaled against the other tools on the same page so one giant tool does not erase the rest. Freshness is based on the last public build or update date.

What the recommendation box means

The full comparison table ranks every tool we index for a target. The recommendation box and outbound Store links on a target page are limited to the catalogue, which is finely curated by our research team after detailed manual testing. Every catalogue Actor is manually tested on a daily basis. A catalogue recommendation is a practical route for this use case; it is not a claim that every other tool was tested or that no other tool can work.

Missing data uses published priors

Scrapers differ in what they publish. Treating an unpublished schema as perfect would reward opacity; treating it as zero would be too harsh. We use fixed priors, the same for every tool: field coverage 0.2 when no schema is published, price 0.4 when no per-result price is listed, reliability 0.5 when no 30-day rate is published, and freshness 0.2 when no update date is available. The page's table shows the underlying price, field count, success rate and update date so you can inspect the inputs.

What we do not use

• Payment, sponsorship or affiliate revenue. Store links are plain links; we do not use affiliate links.
• Which organisation publishes a tool. The formula does not read that field.
• Hand-written "editor's choice" boosts. There is no editor override in the sort.
• Unpublished test results as ranking inputs. The daily catalogue test programme is separate from the published score, which still uses the public inputs shown on each page.

Limits you should know

A published schema can list a field that is empty on many records. A public success rate measures platform runs, not whether the data was correct. A high user count can reflect longevity rather than current quality. Price data excludes Actor start events, storage and plan discounts. The ranking is a starting point for your own checks, not a guarantee.

Rankings are recomputed from the catalogue when data changes. The date at the bottom of each page is the real check date, not a cosmetic bump.