How we build a review.

The one rule everything else follows from

We don't publish a recommendation we can't stand behind. Not a summary of ten other listicles. Not a rewritten press release. Each recommendation is built from credible independent evidence — the vendor's own published terms, the most rigorous independent testing we can find, and the consensus of people who have actually lived with the product — cross-checked against each other and reasoned through in the open.

That single rule is why CuratorBits publishes fewer reviews than the listicle factories — and why the ones we publish hold up. Everything below is just the discipline that makes that rule enforceable.

How we build a review

Every review — hosting, VPN, or a freelance marketplace — runs through the same five stages before a single number is assigned.

1. 1 — We start from the real question

   Before gathering anything, we frame the decision a reader is actually trying to make: who is this product for, what would make it the right call, and what would make it the wrong one. That question decides which evidence matters and keeps the review anchored to the buyer, not the marketing page.

2. 2 — We gather credible evidence

   We pull from three independent kinds of source and weigh them against each other: the vendor's own published terms (pricing tiers, SLAs, feature gating, jurisdiction), the most credible independent testing we can find (established benchmarks and audits from reputable outlets), and the consensus of long-term users (durable owner reports about what the product is like to live with over months, not the launch-day hype).

3. 3 — We cross-check and attribute every figure

   Numbers only earn a place in a review if we can point to where they came from. We never present a figure we can't source. A vendor's published claim is labelled as the vendor's claim; a performance result is attributed to the independent testing it came from. Where the sources disagree, we say so rather than pretending there's a single clean answer.

4. 4 — We compare against the named alternative

   A score in isolation is meaningless. Every product is judged against the closest alternatives, carried through the same framework — so "faster," "cheaper," or "easier" always means faster, cheaper, or easier than a specific named thing, not than a vague ideal.

5. 5 — We write the honest verdict, including the parts that don't flatter it

   Pros, cons, who it's genuinely for, and who should walk away. If the renewal price doubles, the support has a poor reputation, or a headline feature is gated behind a tier the marketing didn't mention, that goes in the review — even when we earn a commission on the product.

What decides a score

The dimensions that decide a score change with the category — a host lives or dies on performance and honest pricing; a marketplace on the quality and reliability of the people you hire. A representative view of what carries weight, and where the evidence comes from:

If we can't source a dimension honestly, we say so in the review rather than scoring it on vibes.

How the score works

Every review ends in a single verdict score out of 5. It is not a star-average scraped from other sites and it is not a marketing aggregate. It is a considered judgement built from the dimensions above and the evidence behind them, weighted by what actually matters for that category — for a host, real-world performance and pricing honesty carry more weight than a feature checkbox; for a marketplace, delivery reliability outweighs a slick dashboard.

We deliberately don't hide behind a false-precision formula that pretends a 4.35 is meaningfully different from a 4.4. The number is the honest summary; the review body is where the reasoning lives. Read the "What works / What doesn't" on any review and you'll see exactly what moved the score — nothing is asserted that the review doesn't show.

Independence — and how we're funded

CuratorBits is reader-supported. When you buy through one of our links, the brand pays us a commission at no extra cost to you. That's the whole business model, stated plainly. Here's the wall we keep around it:

• We evaluate first, then add the link. The evaluation and the score are finished before an affiliate link is ever attached. The commission cannot reach back and change a verdict that already exists.
• Commissions never influence a score. We have published — and will keep publishing — critical reviews of products whose affiliate programmes we're part of. A worse product does not get a better score because it pays a higher rate.
• No pay-for-placement, ever. No brand can buy a review, a ranking, a higher score, or a spot on a "best of" list. Sponsored content, if we ever run it, is labelled as such at the top of the page in plain language — never disguised as editorial.
• We'll drop a pick that lets you down. If a product we recommend suffers a serious regression — a security incident, a support collapse, a pricing bait-and-switch — we pause or pull the recommendation and explain why, regardless of the commission attached to it. A partner relationship is not a reason to keep steering you toward something that stopped being the right call.
• Every commercial relationship is disclosed up front, at the top of the relevant page, not buried in a footer.

When something changes — updates & corrections

Software doesn't hold still, so neither do our reviews.

• We revisit every review at least once a year, and immediately when a product ships a major change — a redesign, a pricing shift, a removed feature we relied on. The "updated" date on a review is real.
• We fix mistakes in the open. If we get a fact wrong and it changes the takeaway, we correct it and note what changed — we don't quietly edit and pretend it was always right.
• You can hold us to it. Spot an error, a stale price, or a claim that no longer holds? Email [email protected] and we'll verify and update.

What we don't pretend to be

Trust is easier to give when the limits are stated out loud.

• We're not a 40-person lab. CuratorBits is a small, senior operation — which is exactly why every review carries the judgement of someone who ships production software daily, not the output of an anonymous content farm paid by the word.
• We don't cover everything. We'd rather publish ten reviews we stand behind than a hundred we skimmed. If we haven't researched it properly, you won't find it here.
• We don't chase the algorithm at the expense of the reader. No weekly "refreshed" listicles rebuilt from the same marketing copy. A review changes when the product changes — not when the calendar says it's time to re-rank.

How we use AI — and where humans decide

CuratorBits is a small operation, so we use AI tools to work faster: to gather background research, draft sections, and keep our reviews structured and consistent. We're telling you this plainly because a review site that hides it hasn't earned the trust it's asking for.

What AI does not do here is decide. Every recommendation, every score, and every number that gets published is reviewed and signed off by a human before it goes live — nothing is auto-published unread. And where a review cites a figure — uptime, performance, a price — that figure comes from a source you can verify, never a number a model invented. If we can't stand behind a claim, it doesn't ship.

Put simply: AI helps us move faster; a human is accountable for every word you read.

The person behind the score

See it in practice

The methodology is only worth as much as the reviews it produces. Read one and check our work:

• Cloudways review — how the independent performance evidence and the true multi-cloud cost stack up.
• Hostinger review — entry-tier hosting weighed on published pricing, independent testing, and owner reports.
• All reviews →