Editorial quality 32%
Output quality, product depth, workflow coherence, and performance on the jobs the product claims to do. Set by an editor from primary evidence.
Version 1.1 · Effective Sep 2, 2026
Rankings people can
inspect and challenge.
ToolsRank is an editorial decision system. We combine structured product evaluation with quality-filtered community signals, publish the weights, and maintain a correction path.
The formula
One score, six explicit factors
Overall = 32% editorial + 24% utility + 18% trust + 12% freshness + 8% engagement + 6% momentum
Practical utility 24%
Whether the product removes real friction, fits existing workflows, and delivers value at realistic limits and cost. Set by an editor.
Trust & transparency 18%
Clear claims, security and privacy information, responsible controls, documentation quality, and evidence traceability. Set by an editor.
Freshness 12%
Current review evidence, maintained product surfaces, and time-sensitive price or feature checks. Set by an editor and decays between reviews.
Engagement quality 8%
Measured saves, votes, and visits on a log scale so raw volume cannot dominate. Until a tool has a dated snapshot it holds the neutral baseline of 50.
Momentum 6%
Measured 30-day growth in interest, centred on the neutral baseline of 50 and moving half a point per percent. It is intentionally the smallest factor to reduce hype bias.
Where each number comes from
Editorial quality, utility, trust, and freshness are set by an editor from the primary sources listed on every profile and carry the profile’s review date. Engagement and momentum are never typed by hand: they are computed from measured, dated snapshots. A tool with no snapshot holds the neutral baseline for both, which means the two community factors cannot reorder tools until real measurements exist. Every leaderboard states whether it is showing editorial factors only or a dated snapshot.
How signals are measured (v1.1)
- Visits are unique daily visitors to a dossier (one per visitor, tool, and day, with known crawlers excluded) plus outbound clicks through the official-site button.
- Saves count signed-in accounts that currently keep the tool saved; device-only saves are not counted.
- Votes are one per Google-signed account per tool per ISO week.
- Growth (30d) compares the last 30 days of visits with the 30 days before, capped at ±100%.
- Streak counts consecutive daily snapshots in the overall top ten.
- Movement compares today’s rank with the previous snapshot for the same metric and period.
A daily job writes immutable snapshots for every metric and period, stores each tool’s all-time signals with the snapshot reference, and only then do engagement and momentum leave the neutral baseline. Day, week, and month periods use signals windowed to those spans; all-time uses totals.
Research and publication workflow
- Scope the claim. We define the primary jobs, intended users, and claims that affect purchase or trust.
- Inspect primary evidence. Official product, pricing, documentation, trust, terms, and support sources come first. We do not turn a marketing claim into an unqualified fact.
- Normalize the profile. Pricing, features, audiences, integrations, limitations, and source dates enter one structured model.
- Score independently. Factor values are bounded from 0–100. The public weights generate the overall score; stable tie-breakers use overall score, trust, freshness, then name.
- Render all claims consistently. Detail pages, category lists, comparisons, alternatives, the leaderboard, API, sitemap, and llms files read the same record.
- Review and correct. Material changes trigger a recheck. Every profile exposes its review date and correction link.
Paid placement and conflicts
Money cannot buy organic rank.
Featured placement, sponsor modules, and verified maker profiles must be clearly labeled and stored separately from ranking factors. An advertiser can be removed for misleading claims, but payment cannot increase its score.
Community-signal integrity
Raw counts are not treated as truth. Events are attributed to an authenticated or privacy-preserving actor, rate-limited, deduplicated, filtered for automation, and retained as auditable aggregates. A metric can be frozen during an investigation. Engagement contributes only 8% and momentum only 6%.
Known limits
A score compresses context; it does not decide for every buyer. Vendor facts can change between review and visit. The catalog expands only when a profile can meet the evidence standard. High-stakes use still requires domain-specific due diligence.
Corrections and appeals
Anyone can submit a sourced correction. Makers may challenge factual errors or supply primary documentation, but cannot demand a particular verdict or score. Material corrections are documented in the public change log.