# Transparency Review scores use the recorded review and vote history. Operational counts, including reports and private requests, also use service records; those records are not all public vote-log events. The reporting job computes the displayed aggregates. Nobody edits it. The same numbers are published as JSON. The live block shows the reporting period for the accumulated counts and for the last 30 days. The configured launch date is a planned reference date, not a claim that the site has already launched. The time of the last run is printed at the top of the numbers. ## Reviews **Submitted** is every review sent from the composer, whether or not it went further. **Published** is how many of those are on the open web, with the number declined before publication and the number still pending a person's approval beside it. **Counted** is how many published reviews currently count toward the meters under the [eligibility rule](/trust/eligibility), out of all reviews that are published and not notes. **Theological notes** is how many submissions are notes, with the number a moderator moved there beside it. ## Published, not counted, by reason One line per reason the rule can give: account newer than 14 days; body under 200 characters; no REVA reading or study activity yet; declared connection to the maker; insider (works on REVA); earlier review of the same resource replaced. Most reviews on the first line will count when the clock runs out. The activity line is printed even though the activity rule is off at launch, so that you can see it stays at zero. The reason is printed beside each review as well as counted here. ## Removed, by reason and sentiment A table with one row per listed reason (suspected fake; unrelated to the resource; harassment or obscenity; personal information about a private person; a review of a person) and one column per sentiment (positive, negative, mixed), with totals. We never remove a review because it is negative, and this table is where that is checkable. Doctrinal arguments are moved to notes, not removed, and the number moved in the period is printed under the table. Each row of the table is listed by review in the [takedown log](/trust/takedowns). ## Reviews of REVA, our own app Our product, reported on its own line. **Published** reviews of REVA, with how many are counted and how many are waiting on the 14-day clock. **Removed**, with sentiment. **Moved to notes or frozen.** **Insider attempts refused**: how many times the composer refused an insider trying to review REVA, which insiders may not do at all. Compare the removal rate on this line with the table above. REVA writes no editorial content about its own app, and its page carries no rating markup for search engines. The rest is on the [conflicts page](/trust/conflicts). ## Automated screening How many reviews the third-party classifier labeled, and what happened next: how many were clean, how many were labeled, how many of the labeled ones a person published anyway, and how many were declined or removed. **Screening unavailable** counts the reviews the classifier could not answer for; a person reads those as usual. The labels themselves are never published: a classifier can be wrong, and a label is not a finding. ## Takedown proposals A proposed decline, removal or freeze normally needs confirmation by a second organization. This block counts confirmations, rejections, withdrawals, expiries and **operator overrides without a second organization**. A different REVA organization administrator can use the limited override when no partner has an assigned moderator, or after the configured waiting period when the timed route is enabled. **By organization** breaks the same actions down per organization, so one organization proposing everything and another confirming everything would show here. ## Reports How many reports of reviews were received, how many are resolved, and how many are open. Of the resolved, how many led to a removal and how many were kept with a reason recorded. ## Requests Every request we received, by kind. **Corrections** to a fact: applied or declined. **Suggestions** for a listing: listed, declined, or open. **Takedowns**: applied or declined. **Disputes** of a description, under the [disputes process](/trust/disputes). Requests enter the staff inbox. Include a reply email if you want a response. Support and seller-link inquiries are operational requests, not votes or review approvals. ## Appeals How many appeals of a removal, a decline or a move to notes were opened, upheld and declined. The normal route uses an organization that did not take the disputed action and never the same person. Limited, recorded operator overrides are described in the [moderation policy](/trust/moderation). ## Resources How many listings exist, how many are published, and how many have been withdrawn. Withdrawals are named and dated here. We never delist on doctrine; the reasons a listing can be withdrawn are on the [listing policy](/trust/listing). ## Reviewers How many vetted reviewers are admitted, and how many published reviews carry the Insider badge. Insiders abstain until 4 April 2027; after that their reviews are badged and never counted. ## Freezes Every freeze of a listing, with its dates and its public reason. A freeze pauses new reviews and votes on one listing; existing reviews stay visible. ## Scoring The scoring version in force and its constants, printed in full, with the changelog line for every change. The current scoring version and effective date are displayed in the live block. The formula constants include C = 7, share prior 0.70, meter prior 3.0, z = 1.281551565545, rank floor 10, segment floor 5, share floor 3, minimum body 200 characters, minimum account age 14 days, activity rule (10 chapter reads or 3 study artifacts) off, recent window 90 days. The [methodology](/trust/methodology) explains each one. ## Checking these numbers Every count here can be recomputed from the export at `/api/votes/export` with the open scorer. The monthly data dump described on the conflicts page carries the same rows as CSV and JSON. If your recomputation disagrees with this page, tell us, and keep the copy you downloaded.