Skip to main content
The optimization page uses real user traffic — Core Web Vitals and Flashduty’s custom loading-time metrics — to find, from the user’s point of view, why pages load slowly. Metrics can be broken down by browser, region, app version, and more, so you can track trends and find the most affected user groups.

Diagnosis and optimization workflow

Open the optimization page from the Performance Monitoring menu, or jump to it from performance data in an analytics dashboard.
1

Metric Selection

After selecting a page and metric, you can start performance data analysis:
  1. Click on a metric card to select the metric to analyze
  2. Select a group in “Show Filter Breakdown”
  3. Evaluate metrics at different percentiles (for example, p75 represents the 75th percentile value)
The metric cards adapt to the application’s platform, so they always show the metrics that platform actually reports:
The platform follows the application type; when you switch applications, the metric cards and the tree map (metric view) dropdown refresh to match. For mini programs, onload_to_onshow and onshow_to_onready are the two page-lifecycle segments and help diagnose slow cold starts.
2

Data Viewing

Change the aggregation or filters to compare how the metric behaves under different conditions.
Data Viewing
3

Problem Diagnosis

In the problem diagnosis section, you can see resources and errors that users encountered on the page that may affect metric performance.
Problem Diagnosis
The page also shows:
  • Diagnosis: the main source slowing down the current metric, judged from the data below. Click View evidence to jump to the matching section.
  • LCP elements / Interaction targets / Shifted elements: correspond to LCP, INP and CLS, with each element’s share of views and P75.
  • Scripts blocking the main thread: the longest-running scripts across the latest views with long tasks. Only Chromium-based browsers report script attribution.
For example, for Largest Contentful Paint (LCP), you can view resources loaded before LCP was triggered. LCP measures how long the largest element on the page takes to load. Diagnose it from two angles: resource loading and errors.
4

View Event Samples

If you cannot locate key information more accurately during the resource analysis phase, you can select appropriate event samples in the event samples section for more detailed contextual analysis.
Event Samples

Optimization directions

The resource loading analysis panel displays detailed loading information for all resources under the current view, helping you identify performance bottlenecks.You can switch between two tabs to control the viewing scope:
  • All Resources: Shows all resources loaded in the view
  • Blocking Resources: Shows only resources marked as blocking render-blocking resources, which delay the page’s first render
The resource list contains the following fields:Hover over the duration bar to see each phase’s timing. When one phase takes 50% or more of the time, the card marks it as the dominant phase and suggests optimizations. This tells you whether the connection, server response, or resource size is the bottleneck. TLS counts as part of connection setup, so it’s never the dominant phase on its own.
If Link Integration - Built-in Tracing is configured and the resource carries a trace_id, you can drill down to the corresponding trace. When the browser does not expose phase timings (commonly caused by cross-origin responses missing the Timing-Allow-Origin header), the bar area shows phase timings unavailable.