Queues

queues are lists where cases wait for review by human moderators and are the backbone of nima’s human moderation workflow as described in rule configuration docid\ hfx29ae9o5ktdj2 dokxb , cases can be sent to queues by either rule configuration docid\ hfx29ae9o5ktdj2 dokxb – when content is flagged as 🟠 and requires human review after ai detection, or rule configuration docid\ hfx29ae9o5ktdj2 dokxb – when content is directly routed for manual review (e g , user reports, lea requests, trusted flagger notices) all the rules routing cases to a queue are displayed on the left hand side of the page additionally, each queue page displays which teams can access it and what policies and workflow actions are available how to create and set up a queue team leads or administrators can create queues in configure > moderation & teams > queues add new queue – start by creating a queue queue name – choose a clear, descriptive name (e g , harassment reports eu ) sla configuration – define the maximum handling time to set a service level agreement (sla) for timely moderation enter a value and choose the unit – hours or minutes this is what moderators and team leads are measured against day to day the queue's sla target appears on the queues page next to that queue's pending caseload, together with the number of pending cases that have already breached it a queue left without an sla shows no target and never reports a breach, so its backlog can age unnoticed you can also set an auto release threshold (in minutes) if a moderator picks up a content item but does not complete the review within that time, the item is automatically released back to the queue so another moderator can action it this prevents content from staying locked indefinitely when a moderator is unavailable to set it up, reach out to the tremau support team assign content moderators to this queue – select which teams are responsible for this queue; moderators of assigned teams will have access to queues in their dashboard assign policies to this queue – limit or focus the policies available in this queue to ensure precision in decision making assign decision actions to this queue – define what moderators can do when they action a case in this queue no policy violation is always available; the rest of the list is the set of decision actions configured for your platform (for example, approve content ) assign appeal actions to this queue – select which actions moderators can apply when reviewing appeals raised against cases from this queue configure custom attributes & blur settings – select customize to open the drawer and control what moderators see in this queue select which custom attributes appear for moderators reviewing this queue, controlling information access and compliance with data minimization principles under blur effects , set the default blur and greyscale intensity — each a percentage from 0 to 100 — to protect moderator wellbeing while respecting content sensitivity policies these settings ensure different teams see only the information relevant to their role and expertise when deleting a queue that has rules assigned to it, bulk reassign all associated rules to another queue directly in the deletion flow select the target queue from the dropdown and all rules will be transferred automatically you will also be prompted to select a queue for transferring any existing cases how many queues should you create? start with fewer queues than you think you need a queue is worth creating when a distinct group of moderators works it, or when the work needs its own sla — not merely because the content is different if the same team handles two categories at the same speed, one queue with both policies assigned is easier to staff and easier to measure a few rules of thumb split by team or by sla, not by policy policies already filter what a moderator can select; a separate queue adds value only when the people or the turnaround differ set an sla you would actually act on the target drives the breach count moderators and team leads see on the queues page, so a number nobody enforces makes the whole page less trustworthy watch for queues that never fill a queue receiving a handful of cases a week fragments attention; fold it back in once cases are sitting in a queue, the order they are served to moderators is governed separately — see prioritisation engine docid\ jqgkzzngeonnuj8xbsibw escalation queues escalation queues are specialised queues for handling high risk or complex cases they ensure that elevated risks (e g , threats of violence, child safety) receive additional scrutiny before action is taken they are created in the same place, but behave differently from standard queues in three ways the type is chosen when you create the queue select escalation queue instead of standard queue at creation the choice cannot be changed afterwards, so a standard queue cannot be converted into an escalation queue later they have no sla the sla field does not apply to escalation queues, so they report no target and no breach count on the queues page track them by pending volume instead they are assigned to all managers when you create one, nima assigns every manager to it and asks you to confirm if you need a narrower group of reviewers, use a standard queue with the relevant team assigned