Analytics & Quality Control

our analytics dashboard provides an end to end view of your trust & safety operations you can access the analytics from moderation workspace > interactive analytics navigating your data every dashboard includes a set of filters date custom select specific timeframes to compare performance over days, weeks, or months select label(s) isolate data for specific content categories (e g , "spam" or "harassment") to see how different violation types impact your workflow select policy(s) narrow your view to specific rules within your playbook to evaluate how individual policies are being triggered and enforced across the platform to help you navigate the dashboard, here is a summary for each tab this dashboard provides a clear picture of today’s performance compared to historical benchmarks what you can find offers a high level snapshot of today's performance, from the total volume of incoming cases and user reports to the current backlog and overall moderation throughput cases inflow & management helps you visualise real time volume spikes and compare current inflow patterns against seven day averages it helps you identify whether emerging content patterns are deviating from weekly norms policy & actions tracks the application of platform guidelines and the resulting enforcement actions it provides visibility into violation trends, the volume of cases cleared as non violative, the performance of automated rules moderators performance monitors the balance between human vs automated moderation, providing granular data on throughput, queue pressure, and the specific policies being handled by each moderation type it is designed to identify bottlenecks and balance the load between ai and humans appeals tracks the volume and status of user challenges against moderation decisions it provides data on the specific policies most frequently contested by users, and the daily ratio of maintained versus overturned decisions to monitor enforcement accuracy and user friction docid\ xhbqkyol19f7fzber3vdt this section offers a deep dive into how content enters the system it analyzes detection sources, precision levels, and current queue distribution this tab provides an overview of policy violations and rate, moderation actions, rule performance, and related reporting, with filters and exports to support operational tracking and transparency use the date and policy filters to evaluate how specific rules are performing or to audit enforcement consistency across different violation types what you can find policy violations policy metrics on violations, detections, and trends give a snapshot of trust & safety activity it uses flow mapping to visualize the journey from initial detection or input labeling to the final policy classification, highlighting which policies are handled by ai versus human moderators actions these analytics provide a view of enforcement volume—such as bans, warnings, or removals—and detail how these actions are distributed across different moderation types mapping detection sources to specific actions helps you understand how different entry points result in different enforcement levels rules provides summaries of auto actions, human review, and no action rates; shows which rules are working, misfiring, or need tuning dsa transparency reporting these charts map proactive detection and user reports directly to digital services act (dsa) categories, ensuring the platform meets legal transparency and reporting obligations this page is an internal dashboard for monitoring moderator and queue performance it highlights performance management metrics like average resolution time and case volumes, provides moderator level analytics, and analyzes queues the date filter allows you to monitor performance trends and ensure the team is meeting operational targets what you can find performance management these metrics track average resolution time (current and trend) and total case volumes, helping teams spot throughput changes and identify days or periods where backlogs or slowdowns occur moderator analysis per moderator time spent, cases handled, policies applied, case statuses, escalations, and appeals outcomes, providing visibility to coach individuals, balance workloads, and recognize high performers queue analysis analyzes queue pressure over time, average resolution time and average time spent per queue, plus actions taken by queue, enabling targeted tuning of processes where delays or misclassifications concentrate it also cross references speed against accuracy and policy application, providing a view of how different categories of violations are being processed and where the highest latency or escalation rates occur this section provides detailed oversight of the challenge process and decision accuracy, offering the transparency needed to evaluate the "second look" workflow by analyzing the frequency and success of user challenges, teams can effectively close the loop between initial enforcement and policy refinement by filtering by date , you can assess the health of the appeals pipeline and identify if specific timeframes saw a rise in user disputes what you can find overview tracks high level performance metrics, including the frequency with which users challenge decisions and the overall success rate of those appeals it offers a macro view of decision integrity by comparing maintained versus overturned actions and monitoring how these trends evolve over time policy analysis detailed breakdowns of appeal reasons by policies reveal the specific "why" behind users appeals by identifying this, you can pinpoint areas where policy language or enforcement criteria may require refinement for better consistency queues and human moderation tracks the distribution of appeals across various queues and identifies the volume of pending challenges in the backlog, facilitating timely responses to users which is vital for meeting regulatory timelines for dispute resolution th e anomaly detection tab acts as an automated early warning system by utilizing z scores, this dashboard identifies statistical deviations from historical norms, allowing trust & safety teams to respond to emerging crises before they escalate understanding the alert system the system categorizes deviations by severity to help your team prioritize their response red alerts (extreme outliers) high priority alerts indicating a z score of 2 0 or higher, signifying a massive departure from standard volume or behavior orange & yellow alerts (significant deviations) indicators of notable shifts (z score 1 0 2 0) that may require investigation before they become critical active alerts table a live, centralized list of specific queues or policies currently triggering these statistical warnings identifying the "what, where, and when" spike origins use the top policy violations and detection source anomalies charts to determine if a spike is caused by a coordinated violation campaigns or a sudden viral trend triggering a specific policy queue health the top queue issues chart allows leads to instantly see which moderation teams are operating outside their normal parameters, often indicating an unexpected surge in workload temporal patterns the "when do anomalies happen?" and hourly trend visualizations help identify if outliers are occurring at specific times of day, which can be useful for identifying bot attacks or shift related operational gaps why this is essential for t\&s rapid response instantly detect "raids" or coordinated harassment campaigns that might otherwise be buried in global volume totals system health monitoring identify technical bugs, such as an automation rule "over firing" and sending thousands of incorrect cases to a queue strategic staffing understand temporal patterns of instability to ensure high tier responders are active during peak "anomaly" hours csv download the csv download utility allows users to export raw data for custom reporting or offline analysis this ensures that the insights gathered within the platform can be integrated with external business intelligence tools or shared with stakeholders who require granular spreadsheet access