Severity tuning: making Civora stricter or more lenient

Civora rates each potential violation on a 1–5 severity scale. Each category has its own threshold and its own action when that threshold is crossed. This two-step design — threshold plus action — is what lets the same template be strict for one category and lenient for another.

The 1–5 severity scale#

The scale runs from 1 (very lenient) to 5 (very strict). The threshold value you pick is the level at and above which Civora applies the action.

A low threshold (1–2) means Civora acts on almost anything that touches the category. A high threshold (4–5) means only clearly severe content triggers action. The Moderation tab shows example messages at each level so you can calibrate by example rather than by abstract description.

The six moderation categories are: Toxicity, Harassment, Spam, NSFW, Hate speech, and Self-harm. Each can be enabled or disabled independently.

The five actions#

When a message crosses the threshold, Civora applies the action you chose for that category:

Ignore — Civora does nothing visible. Use this to effectively disable enforcement for the category without disabling detection.
Flag for review — the message stays up; an entry appears in the review queue for a human decision.
Delete message — Civora deletes the message immediately.
Delete + Timeout — Civora deletes the message and times the user out. The timeout duration is configurable per category (Discord allows 1 to 40320 minutes, up to 28 days).
Delete + Ban — Civora deletes the message and bans the user from the server. Reserved for the most serious violations.

Where the controls are#

Open the server from the Servers page in the dashboard.
Switch to the Moderation tab.
Each of the six categories has its own card: an enabled/disabled toggle, a severity slider (1–5), an action selector, and a timeout-duration input that appears when the action is Delete + Timeout.

Changes are batched — a sticky save bar appears at the bottom of the screen when you have unsaved changes. Save to apply.

Custom tuning is a Pro / Business feature#

On Free, the path forward when the defaults don't fit is to try a different template. See Choosing the right template.

Tuning workflow: start lenient, tighten later#

Counterintuitively, starting too lenient is usually safer than starting too strict:

Too lenient → you see borderline content that should have been actioned. You lower the threshold. Cost: maybe an hour of chat looked worse than it should have.
Too strict → false positives. Members feel like the bot is in their face. Trust in moderation drops. Cost: weeks of slow community attrition while you figure out what to relax.

A reasonable cadence:

Day 1. Pick a template, change nothing.
Days 2–7. Watch the Stats page and the audit log. Look for two patterns: messages Civora actioned that were actually fine, and messages that weren't actioned and you wish had been.
End of week 1. Make at most a few adjustments. Pick the biggest pain point and tune that one category. Live with the new setting for another week.
End of week 4. Most communities settle into a calibration after a handful of adjustments. The defaults are reasonable starting points.

When NOT to tune#

Some patterns look like a tuning problem but aren't:

A specific channel always over-moderates → exclude the channel from the Excluded channels tab rather than lowering the global threshold.
A specific user always trips Civora → it's usually a user problem, not a tuning problem. Have a mod conversation, or take direct action from the review queue or audit log.
A trusted role keeps getting flagged → exclude the role from the Excluded roles tab.

Tuning is for systematic mismatches between what Civora thinks is severe and what your community thinks is severe. If the mismatch is one channel, one user, or one role, fix that specific thing.

What's next#

Reviewing flagged messages — your workflow on the review queue.
Excluded channels and roles — when to take Civora out of a part of your server entirely.