For channelsOpen in TelegramOpen

How We Calculate the Rating

Effective October 26, 2026

This is a reference page, not a contract. The rules that apply to everyone are set out in the "Terms of Service", the "Review Guidelines" and "Disputes and Complaints".

In short

  • A project's rating runs from 1 to 5. It is a weighted Bayesian average of the scores in its reviews.
  • A verified review has a weight of 1. A review marked "as told by the author" has a weight of 0.5.
  • With fewer than 10 reviews, the rating is preliminary and shown with an ellipsis: "3.8…". From 10 reviews on, it is settled and shown with a period: "4.3.".
  • The rating is calculated by code. It cannot be bought or adjusted by hand.

1. What goes into the rating

  • Only the scores from reviews published on Rated count, together with the weight of each review. Nothing else.
  • Reviews deleted by their author or by moderation are not counted. Manipulated reviews are not counted either: we remove manipulation in full.
  • There is one formula for every project: channels, bots, services and exchangers. Nobody gets special terms.

2. Why verified reviews carry more weight

A verified review has a checkmark and a weight of 1 in the rating. A review without proof is published with the label "as told by the author" and has a weight of 0.5. The reason is simple: an experience we could check is more reliable than a story nobody can check.

These count as proof:

  • a link to the ad post: through Telegram we check the date, the views and whether the post was deleted early;
  • a transaction hash: on the public blockchain we check the amount, the time and the recipient;
  • screenshots reviewed by a moderator. Only moderators see them.

Proof confirms a fact: the post went live or the payment went through. It does not make the author's opinion more correct, but it gives the review full weight.

If the author adds proof later, the review's weight goes up and the rating is recalculated. A review that accuses a project of fraud, non-payment or non-delivery is not published without proof, so it never reaches the rating. See the "Review Guidelines" for details.

3. Why not a simple average

A simple average works well when there are many reviews. When there are only a few, it depends too much on a single voice. The very first 5-star review would make a new project the best on the list, and the first 1-star review would make it the worst.

That is why we use a Bayesian average. The formula adds a reference value to the real reviews, a kind of starting point.

  • While there are few reviews, the reference value pulls the rating toward the middle. One random review cannot send a project sharply up or down.
  • With each new review, the reference value matters less.
  • The more reviews there are, the more closely the rating reflects the reviews themselves.

This way, new projects do not end up at the top or the bottom by chance, and projects with a long history get a rating that shows people's real experience.

4. The ellipsis and the period

  • While a project has fewer than 10 reviews, its rating is preliminary and shown with an ellipsis: "3.8…". Each new review can move it noticeably.
  • From 10 reviews on, the rating is settled and shown with a period: "4.3.". New reviews still change it, but more gradually.
  • The threshold is based on the number of reviews, not their weight. Reviews marked "as told by the author" count too.

An ellipsis does not mean something is wrong with the project. It only means there are not many reviews yet.

5. The figures next to the rating

A project page shows three statistics blocks next to the rating. A language model picks which ones from a catalog, based on the type of project. For example:

  • ad prices;
  • how many subscribers came in;
  • the exchange rate;
  • common reasons under "What went wrong".

The figures themselves are calculated by code. It takes them from review fields the authors have confirmed: the model parses the review text into fields, the author confirms them with a tap, and only then do they go into the statistics. The model does not calculate any figures. The statistics help you see the details, but they are not part of the rating.

We also show Telegram data separately: the subscriber count, the SCAM and FAKE labels, and the history of username changes. This data is not part of the rating either. The SCAM and FAKE labels are set by Telegram. Rated only displays them.

6. What does not affect the rating

  • money and paid features, if they ever appear;
  • verifying ownership of a project and the owner's replies to reviews;
  • the subscriber count and other Telegram data;
  • "helpful" marks on reviews;
  • the statistics blocks.

Nobody changes the rating by hand: not the owner, not the Rated team. A review cannot be removed for money. Search positions are not for sale.

7. When the rating changes

The rating is recalculated automatically when:

  • a new review appears;
  • the author changes their score, adds proof or deletes the review;
  • moderators correct or remove a review, including as the outcome of a dispute;
  • manipulation is removed.

While a dispute is open, the review stays on the page with the label "Dispute under review" and counts as usual. How disputes work is described in "Disputes and Complaints".

If an author deletes their account and chooses to keep their reviews without a name, those reviews stay on the page and in the rating.

If the manipulation was ordered by the owner, the project page also gets a notice about it, on top of the recalculation.

8. How to read the rating

The rating brings the authors' opinions together. It is not a guarantee, a recommendation or advice to make a deal.

Before you decide:

  • read the reviews themselves, not just the number;
  • check how many of them are verified;
  • check whether the rating is still preliminary.

Rated does not take part in deals and does not refund money.

9. If the calculation changes

If we change how the rating is calculated, we will update this page with a new date and announce it in the @ratedapp bot and on the website. The principles stay the same:

  • one formula for everyone;
  • verified reviews carry more weight;
  • the rating cannot be bought.

You can ask questions about the calculation in support: @ratedapp → Profile → Support, or support@rated.tg.