Review integrity policy

Anyone can publish a review site. What makes one useful is whether it can tell you where the reviews came from, and whether it holds that line when a vendor is unhappy. So we wrote the process down in public, before we had to defend it.

This is also our statement under the EU consumer rules that require a platform showing reviews to say whether and how it checks that they come from people who actually used the product. Ours is below in full, and there is no exception we have left out of it.

• Terms of service
• Privacy policy

You need an account, your real name, a work email address, your job title and the company you work for. And you need to have actually used the product — at work, not in a demo you sat through once.

• One review per product per person. You can update it whenever your opinion changes, and the update is dated.
• You cannot review a product built by the company you work for, or by a company you are paid to represent.
• You cannot write a review on somebody else’s behalf, or post one your employer drafted for you.
• If anyone gave you anything for writing — a discount, a voucher, a gift card, an entry into a draw — you have to tell us, and the review says so where it is published.

We do not pay for reviews. If we ever run a review drive together with a vendor, the vendor does not choose the reviewers, does not see anything before publication, and the reviews carry a visible label explaining how they were collected.

Verification is a ladder, not a stamp. A single verified tick would flatten a real difference: knowing that an email address works is not the same as knowing that the person behind it runs finance at a company that actually pays for the software. Every published review shows the rung it reached.

A review never publishes at a rung it did not earn, and a rung can move later. LinkedIn sign-in is not available on every deployment at every moment; when it is not, reviews publish at the email rung with the matching badge and are upgraded afterwards without anyone rewriting a word of them.

A pasted profile URL proves nothing — anyone can paste anyone’s. So we do not ask for one. You sign in with LinkedIn, and LinkedIn tells us who you are directly. Your password never passes through us.

• What we receive: the identity LinkedIn confirms — your name, your profile picture and a verified email address — and, where you grant it, your current position.
• What we do with it: compare it against the job title and employer you stated, to decide whether the review earns the role-verified badge.
• What we publish: your name and role as you entered them, plus the badge. Never your profile link, your picture or your network.
• What we never do: post anything to LinkedIn, contact anyone in your network, use the data for marketing, or pass it to the vendor being reviewed.

The evidence behind a verification is kept for 24 months and then deleted. You can unlink LinkedIn at any time: the review stays up, and the badge falls back to whatever is still provable.

• What we do with LinkedIn data

If the company you work for is the company behind the product, the review is rejected automatically, before a moderator ever sees it. This is the most common way a review site gets gamed, and it is the one check we run without human discretion.

Four things get compared:

• The employer you stated, against the vendor that owns the listing.
• Your verified email domain, against the vendor’s corporate domains.
• Your account, against the members of the vendor’s organisation on SoftSelect.
• The identity LinkedIn confirmed, where you have linked it.

Two companies do sometimes share a name. If the match is wrong, write to us and a person will look at it. But a vendor employee who wants to talk about their own product has a better route than a review: every listing carries a public right of reply, clearly labelled as coming from the vendor.

Reviews from a direct competitor are not rejected automatically — sometimes the person best placed to judge a tool works next door to the people who built it. They are flagged for a human, and published only if they read as an experience rather than a hit piece.

Underneath the review form sits a layer of anti-fraud signals. An honest reviewer never sees any of it working. A review farm trips all of it at once.

• A bot challenge on the form, which most people pass without clicking anything.
• Rate limits per account and per network, so one actor cannot post fifty reviews in an afternoon.
• Duplicate and near-duplicate text detection across every review we hold.
• Behavioural signals, such as how long the form took to fill in.
• The reputation of the network a submission arrived from.

No signal rejects a review on its own. A bad signal raises the review’s priority in the moderation queue, where a person decides.

Everything below the role-verified rung goes into a human queue before it publishes, and anything reported afterwards goes back into it. We aim to clear the queue within two working days.

A moderator can do four things:

• Publish the review.
• Reject it, with the reason in writing.
• Ask you for a detail we cannot verify ourselves.
• Redact something we cannot publish — a named individual, contact details, material that looks confidential — and tell you exactly what was removed.

What a moderator cannot do is rewrite your opinion, soften your wording or change your rating. If a review cannot be published as written, it is rejected and you are told why, rather than quietly edited into something you did not say.

You can reply to a rejection and ask for it to be looked at again. A second person handles the appeal.

• Reviews of a product you have not used, or that describe somebody else’s experience.
• Reviews of your own employer’s product, or of a product you are paid to promote.
• Reviews that were paid for or incentivised where the incentive was not disclosed.
• Personal attacks on named employees, harassment, slurs, or anything unlawful.
• Confidential information, anything covered by an NDA, or another person’s personal data.
• Serious allegations we cannot substantiate — a breach, fraud, a broken law — stated as established fact. Tell us what happened to you and we will publish that. We will not publish a verdict neither of us can support.
• Advertising, referral links, or a review that is really a pitch for a competitor.
• Duplicate submissions, and text copied from another review or from the vendor’s own marketing.

Being critical is not on that list, and never will be. Harsh, specific and negative is exactly what a useful review often looks like.

We publish critical reviews. A vendor cannot buy one back, negotiate it away or threaten it away, and a subscription with us buys precisely nothing on this page. If that were not true, the positive reviews would not be worth reading either.

What a vendor can do is respond in public, underneath the review, with their name on it. Most disputes end there — and a good response does more for a vendor than a deleted review ever would.

When a vendor escalates instead, this is the process:

1. 1The vendor files a complaint. In writing, naming the statement they say is wrong and attaching the evidence for it.
2. 2We separate fact from opinion. A checkable claim — a price, a feature, a contract term, a date — can be right or wrong. An experience cannot. We investigate the first kind only.
3. 3We ask the reviewer. They get to answer, and their identity is never passed to the vendor.
4. 4We decide, and we say why. The review stands, or the reviewer corrects it, or we attach a correction note, or the review comes down because it broke a rule or the facts are demonstrably wrong.

We do not tell a vendor who wrote a review, and we do not hand over a reviewer’s identity unless a court or a law obliges us to. If that ever happens we tell the reviewer, unless we are legally barred from telling them.

Formal legal notices go to {{CONTACT_EMAIL}}. Every one is logged and every one is answered, and an unhappy vendor and an unhappy reviewer go through the same process.

Every review has a report link, and anyone can use it — you do not need an account, and you do not need to be the vendor. Tell us which rule you think the review breaks and what makes you say so.

You get an acknowledgement, a decision, and the reasoning behind it. If you disagree you can appeal once, and a different person handles it. Reports filed in bulk or in bad faith get rate-limited like anything else.

Reports and notices

{{CONTACT_EMAIL}}

Scores are computed from reviews, nightly, by the same formula for every product, and nobody edits one by hand. The formula weighs the number of reviews as well as the average, so a perfect five from three reviews does not outrank a 4.6 from four hundred.

Sponsored placements do exist — they are part of how a free site stays free. They sit in a fixed, labelled slot, they never enter the ranked list, and they never touch a score. If you cannot tell which is which on a page, that is a bug and we want to hear about it.

• Edit or delete it whenever you like. The product score is recomputed on the next run.
• Publish under your full name, or under your initials and role. The badge is the same either way.
• Export everything we hold about you as a machine-readable file.
• When you close your account, you choose: take the reviews with you, or leave them published with your name removed.

• Your data rights in full

The date at the top is the version. We announce material changes on the site before they take effect, and we keep the earlier versions — ask us if you want to read the one that applied when you posted.