How Stack Overflow’s Review Queues Keep Moderation Scalable

The problem: moderation overload on a growing site

Stack Overflow receives thousands of new questions and answers every day. Left unchecked, low‑quality posts, off‑topic questions, and duplicate answers would accumulate, making the site harder to use. The community relies on privileged users to review this influx, but asking every high‑reputation member to judge every item would quickly lead to fatigue and inconsistent decisions.

Thesis: a set of purpose‑built queues, reputation thresholds, daily limits, audits, and aging rules together spread the workload while catching inattentive reviewers

The review system splits the moderation task into six queues, each focused on a specific type of content. Access to a queue is gated by reputation, and each queue imposes a daily cap so that no single user can dominate the process. Audits—hidden known‑good or known‑bad posts—test attention, and items that sit too long automatically age out, preventing stale backlogs.

Overview of the six queues

• First Questions – reviews a user’s very first question (threshold 500 rep).
• Late Answers – reviews answers posted long after a question was asked (threshold 500 rep).
• Close Votes – evaluates whether a question should be closed for being off‑topic, duplicate, etc. (threshold 3 000 rep).
• Reopen Votes – reviews requests to reopen a previously closed question (threshold 3 000 rep).
• Low Quality Posts – spots answers or questions that may need deletion or improvement (threshold 500 rep).
• Suggested Edits – reviews proposed edits from users with less than 2 000 rep (threshold 2 000 rep).

Reputation thresholds and daily limits

Each queue shows its current daily limit in the review interface. For example, on Stack Overflow a user with 3 000 reputation can perform up to 40 Close Votes reviews, 20 First Questions reviews, and 20 Late Answers reviews in a single day. The limits are per‑queue, not a global cap, which means a highly active reviewer can contribute across several queues without hitting a single overall ceiling.

These numbers are specific to Stack Overflow; other Stack Exchange sites often set lower thresholds because their communities are smaller.

Audits and aging out

After you submit a review, the system may silently replace the item with an audit. If you choose the wrong action (e.g., marking an obviously good post as “Looks OK” when it is an audit designed to be bad), the system counts it as a failed audit. Accumulating too many failures triggers a temporary review ban, an automatic guard against robo‑reviewing that requires no moderator intervention.

Items also age out when they receive no action:

• Close Vote tasks expire after 14 days.
• Reopen Votes expire after 7 days.
• First Questions and Late Answers expire after 30 days.

This prevents indefinite backlog growth while still giving the community time to reach consensus.

Worked example: handling a Close Vote task

Suppose you have 3 200 reputation and navigate to the Close Votes queue (https://stackoverflow.com/review/close). You see a question titled “How do I sort a list in Python?” with a close reason flagged as “Duplicate of How to sort a list in Python?”.

1. Click the question to open it in a new tab.
2. Verify that the alleged duplicate indeed covers the same core solution (built‑in sorted or list.sort).
3. If you agree, select “Leave Open” → “No, it is not a duplicate” or, if you think the duplicate is correct, choose “Close Question” → “Duplicate of …”.
4. Submit your review. The system records your vote; three “Leave Open” votes remove the item from the queue, while three close votes trigger the actual closure.
5. After submission, you may see an audit banner if the item was an audit; otherwise you return to the queue.

If you are unsure, the “Skip” button lets you pass without penalty; the item returns to the queue for another reviewer, and a high skip rate on a particular post often signals ambiguity that may spark a meta discussion.

Trade‑offs and limitations

While the queue design spreads work, it concentrates power in a small subset of users who regularly hit their daily limits across multiple queues. This can lead to uneven application of community norms if those reviewers share similar biases. Additionally, the audit mechanism is opaque by design; reviewers cannot tell which items are audits, making it impossible to calibrate difficulty objectively. Finally, the aging‑out rules mean that some content may linger longer than ideal if the community is slow to act, especially in low‑traffic tags.

Practical verification: visit https://stackoverflow.com/review/close/stats to see real‑time queue size, tasks reviewed today, and your personal audit history. On a Stack Overflow for Teams instance, go to Admin → Settings → Review queues to inspect configurable thresholds and toggle audits.

Actionable closing

If you have the required reputation, start by reviewing one queue per day and pay attention to the skip button—using it when you are truly uncertain helps keep the system healthy. Regularly check your audit results via the stats page to avoid accidental bans, and consider discussing ambiguous posts on Meta when you notice a pattern of skips or conflicting votes.