softosaurRANKING METHODOLOGY · VERSION 1
How software is ordered
Default category results show Promoted products first, followed by all other software. Promotion comes from Vendor Pro. Within each group, the same data quality formula determines order. Payment adds no points and does not change reviews or software matching.
One score per product and category
Ranking Score ranges from 0 to 100. Different categories have different applicable fields, so a product may have different scores in each category.
| Component | Weight | What it measures |
|---|---|---|
| Completeness | 35% | Approved public answers to applicable, rank-eligible fields |
| Freshness | 20% | Explicit vendor confirmation of the public product data |
| Reviews | 35% | Current published reviews, with recency, content and confidence |
| Evidence | 10% | Supporting sources for facts where evidence is useful |
Completeness
Critical fields weigh 4, high-priority fields 2, normal fields 1 and extended fields 0.5. A known answer, including No or a zero price, earns full credit. Not applicable earns 70%; unknown and not disclosed earn 25%; missing earns none. The percentage uses applicable public fields for this category and the product's published entities, rather than the whole model.
Vendor Checked and freshness
Any verified Free or Pro company can review and confirm its public product data. The “Data checked by vendor” badge lasts 90 days. It records the vendor's statement and is not a Softosaur endorsement. Editing a field does not renew it. Ranking freshness falls smoothly from 100 at confirmation to zero after 365 days: 100 × (1 − age in days / 365), floored at zero. Without confirmation it is zero.
Reviews
Only current, published native reviews with an answered overall rating participate. Hidden, withdrawn and pending reviews do not. Each review's recency weight halves every 365 days. Quality starts at 0.70, with 0.10 each for at least 200 characters of review text, pros and cons of at least 10 characters each, and answered role and usage-duration context. The maximum is 1.00. This is deterministic; AI does not judge review quality.
Effective count is the sum of recency × quality. Rating is weighted by those factors and normalized from 1–5 stars to 0–100. Confidence is effective count / (effective count + 5). The component is 50 + confidence × (normalized rating − 50). No reviews is neutral at 50. A single review has limited influence; older reviews gradually return the signal toward 50.
Evidence
Category facts and selected pricing, deployment, integration and technical sections benefit from evidence. An eligible fact receives credit only when its approved answered assertion has a public HTTP(S) source that supports it, checked within the policy's age limit (365 days initially). Pricing, security, reliability and API require an approved vendor or documentation source. Contradictory links and vocabulary provenance are not support. Field weights also weight this percentage. Where there are no evidence-eligible fields, the component is 100 for everyone.
Order, filters and updates
Filters are applied first. Matching promoted products remain above other matches. Each group sorts by the full score, then freshness, completeness, effective review count, stable alphabetical name and product ID. Pagination walks that complete order: all promoted matches are exhausted before other software begins. Section headings identify the group on each page. Explicit name or evidence-date sorts follow the selected order.
Canonical changes invalidate derived scores. The existing server maintenance job refreshes up to 100 products each hour; daily decay is also calculated on reads for stale entries. Displays round scores to one decimal; sorting uses full precision. Traffic, clicks, advertising spend, company size and subscription price never enter the formula. A methodology change requires a new formula version and recalculation.