> ## Documentation Index > Fetch the complete documentation index at: https://developers.perform.ai/llms.txt > Use this file to discover all available pages before exploring further. # AI Performance Alerts > Scheduled threshold alerts on transit time, delivery success, issues, volume and notification engagement, with a root cause analysis for each breach. **Available to all AI Decision Intelligence customers.** ## Summary AI Performance Alerts lets you set thresholds on the delivery metrics that matter to you, such as transit time, delivery success, shipment issues, volume swings and notification engagement, filtered by carrier, market or service. Alerts run on a schedule; when a threshold is crossed your team is emailed, with an AI root cause analysis attached, instead of finding out from a customer. ## Who it’s for Logistics and operations teams who find out about carrier problems from customer complaints, and customer service teams who need to know when notification engagement drops. The people who own a metric write the alerts themselves and change them whenever the operation changes, without a ticket or a development request. ## Key benefits * **Monitoring from day one.** Pre-built alerts cover the common scenarios, and a new alert is a single form that runs at the next scheduled run. * **It explains the breach, not just flags it.** A breached alert carries the contributing carriers, destinations and patterns, what has been ruled out and a recommended action. It also says when the threshold, not the operation, is the problem. * **The critical breaches come first.** Every alert carries a priority, and long-running breaches are ranked by how many days they have been broken, so chronic problems stay visible. * **It can watch your own definition of late.** A dedicated metric alerts on the measured transit time of a Performance Aspirations category, on your working days and cut-offs. * **Hundreds of alerts are manageable.** Create, edit and export alerts by CSV, and mute one for a set window while you retune it without losing its history. ## How it works ### Alert configuration Each alert is one form, filled top to bottom, with no draft state. | Field | Rule | | - | - | | Alert name | Required. Make it specific enough to tell alerts apart in the list and in emails. | | Metric | Required. One of the 18 metrics below. | | Metric type | The variant of the metric, for instance average, median, P90 or P95 for transit time, a specific issue type for issue metrics, a shipment status, or a transit day band. | | Transit time from and to | For transit time metrics, the two milestones the time is measured between. Any pair from the standard milestones, plus custom milestones configured for your organisation. | | Shipment event type | Which shipment date the alert reads: date shipped, date added or date delivered. | | Range in days | How far back the alert looks: yesterday, or the last 7, 14, 21 or 28 days. | | Trailing in days | 0 to 5 days. Shifts the window back to leave out the most recent shipments, whose tracking may still be incomplete. | | Filters | Optional. See below. | | Breach condition | Required. An operator (greater than, greater than or equal to, less than, less than or equal to, equal to, not equal to) and a value in the metric’s unit: days for transit time, a percentage for rates, a count for volume. | | Escalation message | Optional, up to 500 characters. Sent with the email so the recipient knows what the team expects them to do. | | Priority | High, medium or low. | **Trailing days, worked example.** An alert that runs on Tuesday with a 7-day range and 1 trailing day evaluates shipments from last Tuesday to Monday. With 0 trailing days it would evaluate last Wednesday to Tuesday. ### Metrics | Group | Metrics | | - | - | | Shipment volume | Number of shipments · Shipment volume increase (%) period over period · Shipment volume decline (%) period over period | | Transit time | Transit time · Transit time (days) with working days Monday to Friday · Transit time (days) with working days Monday to Saturday · Transit time (Performance Aspirations) · Transit time change (days) period over period · Percentage of shipments with less than x transit days (%) · Decline in on-time deliveries (%) period over period | | Delivery success | Percentage of shipments without a successful first delivery attempt (%) · Increase in failed first delivery attempts (%) period over period · Percentage of shipments without a selected milestone (%) | | Shipment issues and status | Percentage of shipments with issues (%) · Percentage of shipments with a specific issue (%) · Number of shipments with a specific issue · Percentage of shipments with a specific status (%) | | Notification engagement | Notification open rate (24h) · Notification click rate (24h) | ### Filters An alert can be narrowed on 15 shipment fields: carrier, linked carrier, shipping service, tag, origin country, destination country, sender country, recipient country, shipment status, issue state (no issue, past issue, current issue), issue type, domestic or international, channel, tracking page reference, and the shipment event date set above. Issue-level metrics and the issue type filter use the issue types grouped by the party responsible, such as carrier, sender, recipient and warehouse. ### Schedule and notification Alerts are evaluated on a schedule, not in real time. One schedule applies to every alert in your organisation: you set the frequency (for instance daily or weekly), the time, and the email recipients. Email is the notification channel. Each run sends an email with the breached alerts, their current values, the escalation messages and links into the portal. Muted alerts are left out of the email for the muted period. ### Alert lifecycle and statuses 1. **Created.** The alert shows **Pending** with no result until the first scheduled run. 2. **Run.** Every active alert is evaluated and becomes **OK** or **Breached**. A root cause analysis is generated for breached alerts, the AI Alert Summary is written, and the email goes out. 3. **Muted.** You silence an alert for a set window. It keeps running and recording its result; only the email stops, and the alert resumes notifying on its own when the window ends. Only active alerts can be muted. The alert list shows each alert’s status, a change indicator (OK to Breached or back), its latest result against the breach condition, its priority, a strip with the outcome of the last several runs, and when it was last changed. Filter the list to breached alerts when it gets long. Selecting an alert opens a panel with its **trigger log** (every run, the shipment date range it covered, the result and the condition), its breach condition, its escalation message and, when breached, the root cause analysis. ### Bulk management The upload dialog carries a downloadable alert template and a link to the guidance on filling it in, and takes CSV files up to 3 MB. To edit many alerts at once, export them, change the rows and re-upload; a row with an alert ID updates that alert, a row without one creates a new alert. The same route duplicates an alert with different filters. There is no hard limit on the number of alerts. ### Alerts Dashboard The dashboard shows the active alert count split across breached, OK and muted, today’s run against yesterday’s, a 14-day breach rate trend, and a ranked list of the longest-running breaches with the number of days each has been broken. ### AI root cause analysis and AI Alert Summary For a breached alert, the root cause analysis lists the contributing factors, the patterns leading to the breach, what has been ruled out and a recommended action. Where the evidence points at a threshold set wrong rather than an operation going wrong, it says so and recommends reconfiguring the alert. The AI Alert Summary is a written summary of the day’s breaches, each with a headline naming the finding, the figures behind it, what has been ruled out and a recommended action. You can step back through previous days. It needs at least five active alerts before it produces anything; until then it shows **Insights Collection Ongoing**. Nothing happens automatically. The analysis recommends; your team decides what to change. ### Alerts on Performance Aspirations The **Transit Time (Performance Aspirations)** metric watches the measured transit time of one Performance Aspirations category, as an average, median, P90 or P95, so you are told when a warehouse or carrier slips against your own standard. It needs active rules in [Performance Aspirations](https://developers.perform.ai/products/ai-decision-intelligence/performance-aspirations.md); without them, use the standard transit time metrics. ## Set up in the portal Go to **AI Decision Intelligence** > **AI Performance Alerts**. 1. Select the **Settings** icon and set the schedule frequency, the execution time and the email addresses that receive the alert email. The schedule applies to every alert in your organisation. 2. Select **Create Alert**. Name the alert, choose the metric and its type, and for transit time metrics the milestones it is measured between. 3. Add shipment filters, choose the shipment event type, the range in days and the trailing days. 4. Set the breach condition as an operator and a value, write an escalation message and choose a priority, then save. The alert runs at the next scheduled run. 5. To load many alerts, select the upload action, download the CSV template, fill it in and upload the file. To change existing alerts in bulk, export them, edit the rows and re-upload. 6. To pause the email for an alert while you retune it, open its three-dot menu, select **Mute Alert** and choose a duration. > [!TIP] > > Set a trailing day or two on transit time and issue alerts so that shipments whose tracking is still updating do not trigger a breach. ## FAQ ### Are alerts real time? No. Alerts run at the scheduled time you configure, for instance daily at 09:00 or weekly on Mondays, and check their conditions then. They do not fire the moment a threshold is crossed. ### Can I set different schedules for different alerts? No. One schedule and one set of recipients apply to every alert in your organisation. ### How many alerts can I create? There is no hard limit. Brands commonly run dozens to hundreds, managed by CSV. ### Why is my dashboard empty? Either no alerts exist yet, or the first scheduled run has not completed. Alerts show **Pending** until the first run, and an organisation with no alerts shows an empty dashboard regardless of shipment volume. ### What does “Insights Collection Ongoing” mean? The AI Alert Summary has nothing to summarise yet: no run has completed, or fewer than five alerts are active. ### Will we end up with alert fatigue? Every alert carries a priority, long-running breaches are tracked separately from new ones, and a root cause analysis flags thresholds that are set wrong so a false positive gets fixed rather than tolerated. ## Troubleshooting | Symptom | Cause | Fix | | - | - | - | | An alert stays **Pending** for days | Filters too restrictive, or no shipments match the criteria in the window | Widen the carrier, country or tag filters and check that matching shipments exist. | | The dashboard is empty | No alerts are configured, or the first scheduled run has not happened | Create alerts, then wait for the next scheduled run. Check the trigger time in **Settings**. | | The AI Alert Summary shows **Insights Collection Ongoing** | Fewer than five active alerts, or no completed run | Activate at least five alerts and wait for the next run. | | A breached alert did not appear in the email | The alert is muted | Muted alerts are left out of the email until the mute window ends. Unmute it if the disruption is over. | | Transit time alerts breach on shipments that are still moving | Trailing days set to 0, so incomplete recent shipments are evaluated | Set trailing days to 1 or more. | ## Related * [Performance Aspirations](https://developers.perform.ai/products/ai-decision-intelligence/performance-aspirations.md) * [AI Recommendations](https://developers.perform.ai/products/ai-decision-intelligence/ai-recommendations.md) * [Business Intelligence](https://developers.perform.ai/products/ai-decision-intelligence/business-intelligence.md) * [AI Decision Intelligence overview](https://developers.perform.ai/products/ai-decision-intelligence/overview.md) This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.