Editorial standards

Methodology

This page describes exactly how software data on EveryAny.One is gathered, verified and maintained - and where money is and is not involved. If we ever depart from what is written here, treat it as an error and tell us.

Four commitments

1. A person checks every figure. Nothing is published from a scrape alone.
2. Every fact carries a date. A price without a checked date is not information.
3. We publish the negative. Every profile states who the product is not for.
4. We say "not disclosed" rather than estimating. Gaps are recorded as gaps.

How a profile gets built

Products are selected on one criterion: are people searching for information about this, and can we do a better job of answering than what currently exists? Submissions are one input; so are search demand, category gaps and reader requests.

Each profile is then assembled in a fixed order:

1. Primary sources first. The vendor's own pricing, documentation and about pages.
2. Company facts. Founding year, headquarters, founders and funding, from company registries, investor pages and reputable press.
3. Hands-on use where we can. On free tiers and trials, we run the core task the product is bought for.
4. Assessment written last. After the facts, never before.
5. Dates stamped. Each source records the date it was read.

What each profile is based on, stated on the profile

"Hands-on use where we can" is an honest description of how this catalogue is built, and a useless one to a reader holding a single page: it tells them what we do in general, not what was done here. So every profile carries a What this profile is based on panel that answers it for that product, one line at a time - whether prices were read from the vendor, whether the screenshots are ours, whether anyone used the product, on which plan and for how long, whether we paid for it, whether we have used the alternatives we compare it to, and whether the vendor was contacted.

Those lines record a "no" as readily as a "yes", and a "no" is never omitted to make a page look better. Careful desk research from primary sources is a legitimate thing to publish. It is not a review, and a profile that has not been tested says so in the summary at the top of the page, before anything it might have persuaded you of.

How we handle pricing

Pricing is the field that decays fastest and the one buyers most need to be accurate. Our rules:

• We record the vendor's own published figure, with its billing basis - per seat, per contact, per site, per usage unit.
• Where a page shows an annual-equivalent price, we say so rather than presenting it as the monthly cost.
• We record seat minimums, because they change the real entry price.
• "Contact us" is recorded as quote-only, not estimated.
• Currency is the vendor's own, normally USD. We do not convert.

How assessments are written

The assessment, pros and cons are editorial opinion formed after the factual work. Three rules govern them:

• Every criticism must be specific enough to act on. "Clunky interface" is not publishable; "seats cannot be shared and concurrent sessions are logged out" is.
• Every profile names situations where the product is the wrong purchase. If we cannot find any, we have not researched it properly.
• Where we score a product, the score is ours as a publication - it is not an aggregate of user reviews, and we do not present it as one.

This is why our structured data marks the verdict as a Review authored by EveryAny.One rather than an aggregateRating. We do not collect user ratings, so we do not imply that we do.

The EveryAny Decision Score

Every software profile carries a set of scored dimensions rather than one overall mark. A single averaged number is worse than useless here: a tool with deep features and terrible portability averages to the same 7.5 as a shallow tool you can leave whenever you like, and the difference between those two products is the entire decision.

Each dimension is labelled derived or editorial on the page. Derived dimensions are calculated from verified fields by the formulas below, so they update themselves when a price or a capability changes. Editorial dimensions require judgement that cannot be computed from a price list, and appear only where someone has actually made that judgement. We do not invent a number to fill a gap: a dimension we cannot justify is absent from the page.

Cost position (derived)

Where a product's cheapest paid tier sits among comparable tools in the same category, from 0 to 10, where 10 is the cheapest. This is explicitly not a value-for-money judgement, because cheapness and value are different things and only one of them can be calculated.

• The entry price is the cheapest tier you could buy as your only purchase, expressed per month. A yearly-billed tier is divided by twelve rather than left out, and the page says when it has been: Canva publishes only annual prices, so its $180-a-year Pro plan is positioned at $15 a month, with the annual figure printed beside it.
• Add-ons, extra seats, credit top-ups, metered overages and volume-discount variants are excluded, because a $8 custom-domain add-on is a surcharge on a plan, not a cheaper way to buy the product. Which rows those are is recorded as data on each profile rather than guessed from the wording of a plan name.
• Seat minimums count. Where a vendor will not sell fewer than a given number of seats, the figure compared is the smallest invoice the product can produce, not the advertised per-seat rate. Monday advertises $9 a seat and sells no fewer than three, so it is positioned on $27. Box advertises $5 a seat on Business Starter under the same three-seat floor, which makes its $10 single-user Personal Pro plan the cheapest way to actually buy it. Both figures are printed, because the per-seat rate is what you compare once you know how many of you there are.
• The score is the share of category peers whose entry price is higher, so a tool cheaper than 8 of 10 peers scores 8.
• Peers must be in the same currency. We record prices in whatever currency the vendor showed us, and we do not convert at rates we have not verified, so a euro price is never positioned against a median built from dollar prices.
• Quote-only products are excluded entirely, and so is any category with fewer than three comparable peers. An unpublished price cannot be positioned.

Portability and lock-in (derived)

Portability starts at 3 points and adds: +3 if the software can be self-hosted, +2 for a public API, +1 if it is open source, and +1 where a data export path is documented. Lock-in is reported as 10 minus that score.

Those four signals measure technical portability, and on their own they are biased towards developer tools. A public API does nothing to help a designer move an InDesign layout or keep a licensed typeface. So where a profile documents migration friction item by item, that evidence outranks the generic signals: the mean severity of those items, on a scale of 1 to 3, is subtracted from the portability score. Both the components and the deduction are printed on the page next to the number.

Subscription flexibility (derived)

How easily you can start, stop, or avoid paying. From 2 points: +2 for a free plan, +2 for a free trial, +1 where month-by-month billing exists, plus a pricing-model component of +3 for a one-time purchase, +2 freemium, +1 subscription or usage-based, and 0 for quote-only.

Feature depth, ease of learning, collaboration, AI, adoption

These are editorial and are reviewed when the profile is re-verified. We deliberately do not derive feature depth from the number of features we listed: that would measure how much we wrote, not what the product does.

Because they are judgements, they are published as a band - very low, low, moderate, high, very high - and not as a figure. "Professional adoption: 10/10" reads as the output of a measurement, and a reader is entitled to ask ten out of ten by what reckoning. Where a specific basis is recorded on the profile, the number is shown and the basis is printed underneath it. Where there is not, you get the band, which is the honest width of the claim. The bar beside it is drawn from the underlying figure either way.

The terms rank, and why it is not a popularity score

The software directory publishes an order: a top ten overall, a leader in each category, and a rank on every profile. Discovery sites usually produce that order from votes. We do not collect votes, we do not sell placement, and a vote count would tell you how many people were asked rather than anything about the product - so the order here is arithmetic you can re-run from the profile in front of you.

The terms score is the mean of three derived dimensions above, equally weighted: cost position, portability and subscription flexibility. It measures how a product is sold: what it costs against its peers, how much of your work you can take with you, and how easily you can stop paying. It is not a quality score and the pages that carry it say so. A tool can be the best in its market and rank low here because it is expensive, proprietary and sold annually - that is a real fact about buying it, not a criticism of the software.

Feature depth, ease of learning, collaboration, AI capability and professional adoption are deliberately excluded, even though they are on every profile. They are judgements. A ranked order that mixes in a judgement cannot be reproduced by the reader, and reproducibility is the only thing that makes this worth more than a vote count.

Two rules follow from that, and both remove products from the order rather than fudging one:

• No published price, no rank. A quote-only product, or one with too few same-currency peers to compare against, cannot have its entry price positioned. It is marked unranked with the reason on its own profile, and it stays in the directory table with everything else. An unranked product is not a worse product; it is a less transparently priced one.
• No editorial figure in the order. If any of the three dimensions is ever set by hand on a profile rather than derived, that product leaves the ranking. The editorial number may well be the better one - it is simply not one you can check.

Ties share a position, so the sequence reads 1, 2, 2, 4. Where the published figures are equal, ordering the products against each other would be inventing a distinction the numbers do not contain. A category needs at least four ranked products before it gets a leader: "#1 of 2" is arithmetically true and tells you nothing.

How much of this site references a product

Every profile publishes a count of the other pages here that reference it - a buyer guide that ranked it, an article that priced it, a deal record, another profile that names it as a competitor or as the better tool for a particular job - and lists every one of them as a link.

Category indexes, the directory hub and the sitemap are not counted. Those list every product automatically, so counting them would hand each record several references for merely existing and turn a signal into a constant. Only a reference somebody chose to make counts.

The number is deliberately a measure of our work rather than of the product's popularity, and it is checkable in one click, which is the point. Zero is a normal answer and is printed as one: it means the research on that product has not yet been connected to anything else here.

AI tool categories

Alongside its ordinary category, a product may carry one AI tool category. This is a second, cross-cutting axis rather than a promotion of AI to the top of the taxonomy: AI is one entry beside Finance and HR and stays that way. The axis exists because the primary classification files tools by the department that buys them, which is the wrong grouping for a reader choosing between them - the tools that measure whether AI answers mention a brand are filed under SEO, and the generators under Design.

A product is filed on this axis only where AI is its function rather than a feature it also ships. A great deal of software in the directory now includes AI and appears in none of these groups for exactly that reason; applying the rule loosely would produce a taxonomy in which nearly everything is an AI tool, which is the same as one in which nothing is. A group is only published once at least three researched products qualify for it.

True annual cost and exit exposure

An advertised monthly price is frequently a twelve-month commitment billed monthly, which is a different product from a month-by-month subscription. Every profile multiplies its monthly-billed plans out to twelve months, which is arithmetic almost no vendor shows.

The cancellation calculator appears only where a vendor publishes its own termination terms, and it uses the vendor's stated percentage against the remaining obligation. Where those terms are not published, no calculator is shown. We would rather omit the estimate than guess at a fee someone might budget against, and the calculator says plainly that your contract governs rather than our box.

Field-level freshness

A single "verified August 2026" stamp on a page carrying thirty facts is a weaker promise than it appears, because prices move at a different rate from headquarters addresses. Each field group therefore carries its own date. Groups default to the date the profile as a whole was checked, which is accurate because we verify a profile as a unit, and are overridden individually when one fact is re-checked on its own.

Individually sourced claims are printed as claim, source link and check date together, so any single assertion can be followed to its evidence rather than to a bibliography at the foot of the page.

Why these sections are generated, not written

Plan pickers, per-audience verdicts, buy and do-not-buy logic and the question set are generated from the same verified fields as the fact table wherever possible. This is a correctness decision rather than a shortcut. Prose duplicated across a hundred profiles rots silently the moment a price changes, because nobody remembers which paragraphs restated which number. A derived section cannot disagree with the facts above it.

Where a section genuinely needs judgement, such as whether a photographer should buy a particular plan, it is authored per profile and rendered only when it exists. An absent section means we have not made that judgement yet, not that the answer is no.

Statistics we compute from our own catalogue

Where an article needs a market statistic, we prefer to measure it across the products we already verify rather than repeat a figure from elsewhere. Those calculations are reproducible from the published profiles, and the script that produces them lives in the repository next to the checks that gate our builds.

The first of these is the annual billing discount. Of the products we track, those that publish both a monthly price and an annual-equivalent price for the same plan give a set of directly comparable pairs. As of the most recent run that is 44 plan pairs across 20 of 102 products, with a median discount of 17.1%, a mean of 18.5%, and a range from 10% to 36.8%. Twenty-nine of the 44 fall between 15% and 25%.

Two limits on that figure, which we would rather state than bury:

• It counts plans, not revenue. It is not weighted by how many customers sit on each tier, so it describes how vendors price rather than what buyers actually pay in aggregate.
• It excludes introductory pricing, because a first-year promotion is a different mechanism from a billing-term discount. WordPress.com's Personal plan is $4 for the first year against $9 billed monthly, which reads as 56% off for paying annually when the steady-state gap is $8 against $9. Counting promotions as billing discounts inflates the number, and a good deal of published data does exactly that.

Only 20 of 102 products qualify, because most publish one figure or the other rather than both in a comparable form. We report the denominator for that reason: a statistic drawn from a fifth of the catalogue is worth having and is not the same as a survey of the whole market.

How ranked guides are ordered

Each best-of guide publishes its selection criteria before its conclusions. Order reflects our judgement against those criteria for the stated use case - not popularity, not market share, and not commercial relationships. No vendor is told its position before publication, and no vendor reviews a draft.

The tools behind the research

Pricing pages are captured with an automated scraper so we have a dated, quotable snapshot of what the vendor published. The scrape is evidence, not output: a person reads every capture and writes the figures by hand. That distinction matters, and it is not theoretical. On our first pass a plan-name scan missed the fact that Ahrefs sells a $29/month Starter tier, because the pricing page renders the currency symbol and the number on separate lines. The page said one thing; the naive read said another. A human reading the capture caught it.

We record affiliate programme terms on every profile because founders and marketers search for that information. Recording a fact about a product is not participating in it.

How we write for AI answers

A growing share of readers reach software information through AI assistants rather than search results. We write to be quotable by both: short factual statements, tables, explicit prices, named people, dates and sources. We do not pad pages to hit a word count - a thousand words containing twenty verifiable facts is more useful, and more citable, than four thousand words of generic review. The citation rules we ask machines to follow are published at /llms.txt.

Corrections

We publish facts that decay, so we will get some wrong. Email [email protected] with the page and the specific field. What happens next:

• Factual errors - wrong price, wrong founder, wrong date - are corrected and the verification date is updated.
• Material corrections are noted on the page rather than changed silently.
• Disagreement with our assessment is not a correction. We will publish a founder response; we will not remove the criticism.

What we do not do

• Let a vendor approve a draft before publication.
• Publish a price we have not read on the vendor's own page.
• Estimate a figure and present it as verified.
• Publish AI-generated text as editorial without a human verifying every fact in it.

Who writes this, and who is accountable

Profiles and guides are produced by the EveryAny.One editorial team and published under that name unless individually bylined. A house byline is not anonymity: the publication is operated by Alston Antony, Founder & Editor, EveryAny.One, who is answerable for everything on it. Where an individual authored or personally tested a page, that page carries their name instead. More on the publication at about.