Methodology

Last revised 6 August 2026

This page exists so you can tell how much any given number on this site is worth. Some are checked against a named source on a known date. Most are not, and are marked accordingly.

Where the site stands right now

56 products recorded in total.
49 written for this site; 7 are unreviewed research drafts, each labelled on its page.
45 carry no sourced figure at all. Their ratings and prices display with an "unverified" mark and are excluded from structured data.

Those counts are read from the live dataset, not typed in, so this page cannot quietly fall out of step with the site.

What each label means

Checked date, with a source link: Someone opened that page on that date and read the figure off it. The link goes to the source, so you can check the same thing yourself.
Unverified: The figure is in our dataset but has no recorded origin. It may well be right — we cannot show you why it would be. Treat it as a rough indication and confirm before deciding.
Not published: The vendor does not publish this figure; pricing is quote-based.
Checked over 90 days ago: The figure had a source once, but SaaS pricing moves. Past this point it is flagged as possibly out of date rather than presented as current.
Not yet reviewed: The whole record was assembled by the import and research pipeline and no person has read it. These pages are excluded from search indexing.

Four different things, kept apart

Conflating these is how comparison sites end up misleading people, so the site distinguishes them explicitly:

Vendor-published data — prices, plan limits, feature lists. Taken from the vendor's own pages. Accurate about what the vendor claims, which is not the same as accurate about the product.
Third-party ratings — G2, Capterra, ProductHunt. Always attributed to the platform, never restated as ours.
Our editorial view — written analysis, signed and dated. Opinion, argued from the data above, and identifiable as such.
Machine-drafted summaries — descriptions produced by a language model from web search, unreviewed, and marked as drafts.

Why we don't average third-party ratings

A single blended score out of five looks authoritative and is mostly noise. Review platforms have different populations, incentives and moderation; vendors run campaigns on some and not others; volume varies by orders of magnitude between an established product and a recent launch. Averaging those into one number produces a figure with no defensible meaning.

So each platform's score is shown separately, with its name attached. Where a rating exists in our data without a platform behind it, it is shown as unverified rather than laundered into a house score.

One consequence worth being blunt about: ProductHunt reports upvotes, not reviews. Upvotes measure launch-day attention, not satisfaction after six months of use. They appear labelled as upvotes and should not be read as review volume.

Prices

The recorded price is the lowest regularly advertised monthly price for a paid plan, in USD. Three things it deliberately is not:

Not a promotional price. Launch offers, first-year discounts and seasonal deals expire and would make the figure wrong within weeks.
Not an annual-billing price unless noted. Vendors often headline the discounted annual rate; the monthly figure is the honest comparison point.
Not what you will pay. Per-seat multipliers, required add-ons, usage charges and minimum commitments routinely dominate the entry price. Model your own case.

How a verdict is reached

There is no formula, and no weighted score. Saying otherwise would be the easier answer and it would be false — a published set of weights implies the inputs are measured on comparable scales, and ours are not. A rating with no source cannot be weighted against a verified price without inventing a precision nobody has.

So a verdict on this site is an argument, and it is constrained rather than computed:

It must follow from something on the page. Every claim in a verdict traces back to a figure or plan term stated above it, with its source.
It is scoped to a reader, not declared outright. Comparisons recommend for a described situation — team size, budget, who will maintain it — because "better" without a "for whom" is not a finding.
It never comes from the ordering. No page names a winner because one tool sorts above another; that ordering runs on unverified averages.
It names what would change it. Where a conclusion depends on an assumption — seat count, growth rate, whether anyone will administer the tool — the page says so.

The trade-off is honest: this scales badly. One person can write a defensible comparison at a rate of a few per week, which is why most pairings on this site carry a data table and no verdict at all rather than a generated one.

How often this is updated

An automated job runs weekly. It discovers newly launched products and drafts descriptions of them; it cannot publish reviewed content and it cannot add a page to the search index. Those drafts are labelled on the page and stay out of the index until a person reads them against their sources.

Figures are re-checked by hand, not on a schedule, and each carries the date it was last read. Anything older than 90 days is flagged as possibly out of date rather than presented as current. We do not claim a refresh cadence, because a claimed cadence that nothing enforces is the kind of statement this site has been removing.

What money does not influence

There is nothing to influence: no affiliate programmes, no commissions, no sponsored placements, no paid inclusion. The disclosure page states the position in full, including the one exception — six imported products whose only outbound link is a ProductHunt referral redirect, labelled as such wherever it appears.

Because "we would never let it influence us" is unfalsifiable, the practical safeguard is that the mechanics are published instead: ordering rules, the review threshold, what earns a byline, and what gets indexed are all described on this page and enforced in the build.

Reporting an error

Email [email protected] or use the contact form, which has a dedicated option for corrections and pricing updates. Include the page URL and, where you can, a link to the source that contradicts it — that turns a judgement call into something checkable and is the fastest route to a fix.

Corrections to a figure are made by updating the value together with its source and check date, so the change is visible on the page. Substantive corrections to written analysis are noted on the page itself.

When sources disagree

The vendor's own pricing page wins on price and plan limits — they set them. For anything about how a product behaves in practice, a vendor page is a marketing claim, and disagreement is recorded rather than resolved: where two credible sources conflict and neither is clearly better informed, the page says they conflict.

Ordering and ranking

Listings default to ordering by the stored rating, which for most products carries no source. That ordering is a browsing convenience and is labelled as such. It is not a ranking, no page declares a winner from it, and it never feeds structured data. Sorting by price or filtering by free plan and trial availability uses checkable fields and is more useful.

One rule constrains it: a product with fewer than 500 reviews sorts below every product above that threshold, whatever its average. Averages over small samples are noise, and a newly launched tool sitting at 5.0 above an established one at 4.4 tells a reader something false. Within each band the average still applies.

Feature comparisons

Capability tables appear only on comparisons where a person entered each value by hand, and each cell is one of three things: yes, no, or not recorded. That third state exists because the alternative is worse than useless.

An earlier version of this site generated those tables automatically by merging the two products' feature summaries and marking a dash wherever a phrase was missing from the other list. Since no two products described themselves in the same words, the dashes were nearly universal — the HubSpot vs Salesforce table showed Salesforce without contact management, and the Figma vs Canva table showed Canva without templates. Readers reasonably read a dash as "does not have"; the data only ever meant "not in our notes". Those tables have been removed rather than relabelled, because no caption fixes a table that is wrong.

What is outstanding

Known and unresolved, rather than buried:

Ratings and review counts on editorial entries predate this sourcing requirement and have no recoverable origin. They need re-checking against G2 and Capterra directly, or removing.
No price on the site currently carries a check date. Every one needs verifying against the vendor's pricing page.
Ratings on the research drafts were generated by a language model and are not market data.

Until each is resolved, the affected figures stay visibly marked. If you spot one that is wrong, tell us — a link to the correct source is the fastest way to get it fixed.