> ## 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 Recommendations > Daily and weekly insights that tell logistics and customer service teams what changed in their shipment data, why, and the next step to take. **Available to all AI Decision Intelligence customers.** ## Summary AI Recommendations reads your shipment data every day and every week and tells your team what changed, what probably caused it and what to do about it, broken down by carrier, destination country and shipment band. The day starts with the picture already assembled instead of with hours of pulling data. ## Who it’s for Logistics and operations teams who spend the first hours of the day working out what happened yesterday, and leadership who only spot trends at the monthly review. Customer service and marketing teams use it for notification engagement, where a slow decline is caught while it is still small. ## Key benefits * **It arrives without being asked.** Insights are generated on a schedule you set, so the daily and weekly read is waiting before the morning stand-up. * **Every insight ends in a next step.** Where the data is clear, the insight recommends an action. Where it is ambiguous, it says so and lists what to check first. * **A number is never shown on its own.** Each metric is read against the previous day, the same day last week, the 7-day average and the 28-day average, above a 56-day trend. * **It names what dominates.** Breakdowns by carrier, country and band size each contributor, so effort goes to the bottleneck rather than to whatever moved most. * **It uses your own definition of late.** The SLA breach rate metric is calculated from your Performance Aspirations rules, not a generic transit time. ## How it works AI Recommendations needs no rules, alerts or reports to be set up first. It reads the shipment data already in Perform.AI and produces insights at the next scheduled run after it is switched on. The one exception is the SLA breach rate metric, which stays empty until Performance Aspirations rules exist. ### Daily and weekly insights | Run | Period analysed | Default trigger | History kept | | - | - | - | - | | Daily | The evaluation day, compared with the same day last week and the last 7 days against the previous 7 | 08:00 in your configured timezone (default UTC) | Last 14 days | | Weekly | The last calendar week (Monday to Sunday), compared week over week and against the last 4 calendar weeks | Monday 05:00 in your configured timezone (default UTC) | Last 4 calendar weeks | Data periods are evaluated in UTC. You can set the trigger time in your local timezone, but a period such as “24 November” refers to the UTC date. ### What an insight contains A metric opens with a plain summary of what moved and what probably caused it, then carries several separate insights. Each insight has a headline stating the finding, the figures beneath it and what to do. The shape follows the evidence rather than a fixed template: * **Finding and analysis.** What moved, and what that movement is consistent with. * **Possible causes.** The competing explanations, where more than one fits the data. * **Key questions to explore.** What the team has to answer before the cause is known. * **Implications.** What follows if the reading holds. * **Recommendation.** The action to take, where the evidence supports one. * **Recommended diagnostic steps.** An ordered checklist to run first, where it does not. Insights can separate a data artefact from a real movement and say which they think it is, and they can tell apart an operational problem from a tracking problem. Nothing happens automatically: every recommendation is a suggestion with the data behind it, and your team decides whether to act. ### Comparisons, trends and breakdowns * **Scorecards.** The evaluation day against the same day last week, and the last 7 days against the previous 7, each with its own change indicator. Some metrics carry a 28-day view as well. * **Historical chart.** A 56-day trend with figures on hover. * **Breakdowns.** By carrier, by destination country with each country’s share of volume, and by the bands that matter to the metric, for instance shipments delivered within one day against those taking more than five. Breakdowns go down to the individual notification configuration or carrier service. ### Metrics covered Metrics are organised into tabs by stage of the e-commerce journey: **All Experiences**, **Post-Purchase**, **Checkout**, **Logistics** and **Returns**. A badge on each tab shows how many insights it holds. Today the metrics are Post-Purchase metrics; the other tabs fill as metrics are added. Hover a metric name for its definition. | Metric | What it measures | | - | - | | Shipments added | Total shipments added in the period | | Shipments dispatched | Shipments dispatched in the period | | Fulfilment time | Time from order placed to pick-up | | Transit time | Time from order dispatched to first delivery attempt | | Shipments with issues | Share of shipments with issues against total shipments delivered | | SLA breach rate | Share of delivered shipments that breached a Performance Aspirations rule | | Failed first delivery attempts | Share of shipments with a failed first delivery attempt against shipments with a first delivery attempt | | On-time deliveries (carrier EDD) | Deliveries made within the carrier’s estimated delivery date | | Notification open rate (24h) | Share of notifications opened within 24 hours against notifications sent | | Notification click rate (24h) | Share of notifications whose tracking link was clicked within 24 hours against notifications sent | A metric only appears when there is enough data for the period. Notification metrics are absent on a day when no notifications were sent, and a thin account may show fewer than ten. ### Homepage summary The portal homepage shows a daily AI summary with up to five critical items, each linking straight to the insight behind it. A date toggle opens the summaries from the last 14 days. ### Feedback Each insight carries thumbs up and thumbs down buttons and an optional text box. Feedback goes directly to the Perform.AI product team. ## Set up in the portal Go to **AI Decision Intelligence** > **AI Recommendations**. Insights appear at the next scheduled run after the feature is switched on; no integration work is needed. 1. Select the **Settings** icon in the top-right corner of the page. 2. Choose the **Timezone** used for scheduling. 3. Set the time for **Daily Insights** and the day and time for **Weekly Insights**. 4. Save. The new schedule applies from the next run. To read past insights, use the date toggle at the top of the page and select **Back to Latest** to return to the current period. Select **View all** to expand the full summary. > [!NOTE] > > Changing the trigger time does not regenerate an insight that has already been produced for that day or week. The change applies to future runs only. ## FAQ ### How often are insights generated? Daily insights are generated once a day at your configured time, and weekly insights once a week on your configured day and time. The defaults are 08:00 daily and Monday 05:00 weekly, in UTC. ### Why don’t I see certain metrics? Metrics are only shown when there is enough data for the evaluation period. Notification metrics do not appear if no notifications were sent that day or week, and the SLA breach rate stays empty until you have active Performance Aspirations rules. ### How far back can I view insights? Daily insights and summaries are kept for the last 14 days, weekly insights and summaries for the last 4 calendar weeks. ### Which timezone is used? The trigger time runs in the timezone you configure. The data itself is analysed in UTC, so the period an insight covers refers to UTC dates. ### Does AI Recommendations take action on its own? No. It suggests, with the figures behind each suggestion, and a person decides whether to act. ## Troubleshooting | Symptom | Cause | Fix | | - | - | - | | A metric is missing from the insights | Not enough data in the evaluation period, for instance no notifications sent that day | Nothing to fix. The metric returns when there is data for the period. | | The SLA breach rate shows nothing | No active Performance Aspirations rules | Create and activate rules in [Performance Aspirations](https://developers.perform.ai/products/ai-decision-intelligence/performance-aspirations.md). The metric fills from the next run. | | I changed the trigger time but today’s insight did not change | A new trigger time applies from the next run and does not regenerate an insight already produced | Wait for the next scheduled run. | | The dates in an insight look one day off | Data periods are evaluated in UTC, not in the scheduling timezone | Read the period as a UTC date. | | A journey tab shows zero insights | Only Post-Purchase metrics are generated today | Use the **Post-Purchase** or **All Experiences** tab. | ## Related * [Performance Aspirations](https://developers.perform.ai/products/ai-decision-intelligence/performance-aspirations.md) * [AI Performance Alerts](https://developers.perform.ai/products/ai-decision-intelligence/ai-performance-alerts.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.