Editorial methodology

The evidence states

Six states appear across this site, thousands of times. Each renders as colour, a glyph and a word — never colour alone — and the two absence states also carry a pattern, so a table is readable three ways.

Verified We loaded the vendor’s own page and read that fact there, on the date shown.

Vendor claim The vendor asserts it. We have not independently confirmed it. Counts at half weight in any score.

Not offered The vendor states it does not do this. Their words, not our inference — and the most useful kind of fact we hold.

Not found We searched the vendor’s site and did not find it. Absence of evidence, not evidence of absence.

Not applicable Out of scope for this product. Excluded from the denominator, never scored as zero.

Could not read We could not read the surface at all. This says nothing about the product.

“Not found” and “could not read” are different, and we never collapse them. One is a fact about the product; the other is a fact about our access. Collapsing them is how a directory publishes false facts. 10 products in this dataset have at least one surface we could not read, and their pages say so.

What counts as a source

The vendor’s own site is the only source for pricing and features. Roundups, review sites and AI summaries are never a source for either.
Fetch, don’t recall. If a URL was not loaded during the pass, it was not verified.
Record the negative. “Searched the vendor site on 28 September 2026, no pricing page exists” is a finding. Silence is not.
Never infer ownership from a shared address, a logo, or a logo filename.
Run a control before trusting any negative. A “no price” result is worthless unless the method is known to return prices when they exist.

The review system

Three kinds of “review” exist and they never touch each other:

Our verified reviews. Collected and moderated by us, identity checked. These are the only ones that can affect a score or appear in structured data.
Third-party platform ratings (G2, Capterra, TrustRadius). Cited in prose only, always attributed and dated. Never in our markup. Google requires first-party ratings and prohibits aggregating other sites’, so mixing them in would forfeit eligibility we actually have.
Vendor testimonials. Quoted as evidence of vendor positioning, labelled as marketing. Not a review. Not counted anywhere.

Every published review will show its verification route, the reviewer’s company size and role, how long they used the product, and whether they were invited and by whom.

Scoring, and why nothing is scored yet

A score is not an average of stars. It has four published components:

Job coverage — 40%. What the software verifiably does, verified rows only.
Verified reviewer experience — 30%. Structured answers from identity-checked reviewers.
Evidence quality — 20%. How much of the profile is verified rather than vendor-claimed. This rewards vendors who publish.
Pricing transparency — 10%. Whether a buyer can find out what it costs.

Gate G1: no score until a product carries eight verified reviews. Gate G2: no category ranking until five of its products are scored. Today zero products are scored, so there are no scores and no “best” lists anywhere on this site. Scores are per category — a design tool and a financing platform do not share a scale.

What we refuse to publish

716 candidate URLs were evaluated and 341 were refused. Comparison pages are the clearest case: 202 published, 281 refused.

G-data — 200 refusals. A vendor surface could not be read, or a record was below our completeness threshold.
G-intent — 141 refusals. The software is bundled with hardware, so the “comparison” would really be a hardware buying decision.
G-unique — 52 refusals. The data holds no meaningful difference, so the page would be filler.
G-legal — 43 refusals. Consumer-credit products, where a ranking is close to regulated advice — and the publisher’s own products.

We also never compare lenders head to head, never compare hardware-bundled monitoring tools as if they were software purchases, and never publish a “best” page for a category that cannot reach five scored products.

Freshness

Every fact carries the date it was read. A page verified more than 90 days ago is removed from our sitemap automatically; beyond 180 days it carries a visible staleness warning. A stale price is a false statement, and the whole claim of this site is that its facts carry dates.

Corrections

Anyone, including a vendor, can submit a correction. We re-verify against the vendor’s own page. If we were wrong we correct it, date the correction and log it publicly. If we were right we reply with the URL and the date we read it. See the corrections log, which opens with seven corrections to our own earlier research.