Getting started with DebugBear
This guide explains how to set up DebugBear for your website, interpret performance data, and deal with common problems.
Demo Videoβ
See how DebugBear works and see an example of how to investigate a performance regression.
Set up URL monitoringβ
DebugBear runs website tests in a controlled lab-environment. This is also called synthetic testing. DebugBear builds on top of Google's Lighthouse tool, so you also get recommendations for SEO and accessibility.
All you need to do is provide a URL you want to test and configure the test environment:
- Select a simulated test device
- Select a test location
The standard Lighthouse test Mobile and Desktop test devices are usually a good starting point.
Pick a test location that's close to where most of your users are located.
Invite team membersβ
Invite the rest of your team so they can access monitoring data and get alerted to projects.
Add new members in the Teams tab or grant full account access to Admins.
Test result data: an overviewβ
Each DebugBear test result includes a lot of detail you can use to discuss website problems.
The Overview tab shows a high-level summary of all your site metrics.
The Web Vitals tab provides information specifically for improving Core Web Vitals, and shows Google field data from the Chrome User Experience Report (CrUX) along-side the lab results from DebugBear.
The Requests tab provides in-depth data on what happened on the network when loading your website. This includes page size and an in-depth request waterfall showing blocking requests, request prioritization, and rendering progress.
The Performance tab lets you build custom performance charts, for example you can compare image download size to the Largest Contentful Paint. The tab also lets you view custom metrics from the User Timing API.
The Validation tab show HTML validation errors. Most of these can generally be ignored, but occasionally this surfaces invalid CSS or attribute names.
The Console tab shows any errors and logs that were shown in the console while running your tests.
Interpreting site speed dataβ
Start with the Web Vitalsβ
If you don't know where to start, taking a look at the Web Vitals data from the Chrome User Experience Report is a good start. To get the SEO benefits 75% of user experience should be "Good" according to each metric.
For Largest Contentful Paint (LCP) issues, first check what UI element is causing the LCP. Keep in mind that this can be vary for real users, for example based on the size of their device.
Then check out the waterfall view in the Requests tab βΒ find the request causing the LCP element here. Then ask what causes it to occur too late. Are there many render-blocking script tags? If the LCP element is an image, does downloading it take a long time? Are there many preloaded resources competing for bandwidth with the LCP image?
For Cumulative Layout Shift issues check out the layout shifts table to see what content is shifting around. If the DebugBear lab data shows no layout shifts, but the CrUX data shows a lot then that might mean that layout shifts happen after user interaction.
For First Input Delay check the CPU timeline βΒ are there slow third-party scripts, or a large JavaScript app that needs to render? While JavaScript is running the browser won't be able to respond to user input quickly.
Performance trendsβ
The project overview page lists all pages along with historical metric data.
Here you can:
- switch between daily and weekly data
- view different metrics like Lighthouse scores and Web Vitals
- bulk edit and group pages by clicking on the pencil icon
Common website monitoring problemsβ
Here's a list of common website monitoring problems and how to address them. If you run into problems you're also welcome to contact DebugBear support.
Discrepancies between different data sourcesβ
Different performance tools collect data in different test environments and report data in different ways.
Lab data collected with Lighthouse-based tools usually shows slower speeds than field data from real users. This is because the default Lighthouse network settings are very slow. In DebugBear you can select alternative device configurations like Mobile Fast to simulate a more realistic user experience.
Cumulative Layout Shift (CLS) is often higher in field data, as content shifts around as users scroll down the page or otherwise interact with it.
Test results will also vary between different lab environments. For example, PageSpeed Insights throttles the network in a way that can generate less accurate results. There's also variation in where test servers are located and what their CPU speeds are.
Variability and inconsistent test resultsβ
Each speed test will generate slightly different results, depending on server response times and available network bandwidth. To reduce variability you can repeat tests several times and then select the average test result.
A/B tests or other customizations can also cause inconsistent test results. You can configure your website so that it behaves consistently, for example by detecting the DebugBear user agent or by setting a cookie when the page starts loading.
Finally, performance tests can sometimes finish too early, depending on when Lighthouse decided that the page looked complete. You can control this behavior by defining a Page Loaded Condition.
Page tests blocked by bot protectionβ
Sometimes the DebugBear test servers receive 403 Forbidden responses, or a 429 Too Many Requests code. In those cases you can add the IP address of the test server to a list of allowed IP addresses. Contact support to obtain a list of addresses for DebugBear servers.
Pages behind user login or Basic Authβ
If only logged-in users can access a page you can fill out a login form before running the test.
DebugBear also supports testing pages behind Basic Auth.
Noisy alertsβ
You can mute notifications containing specific terms, like "gzip" or "google-analytics". You can also limit notifications to only certain pages, or disable certain types of alerts, for example in the "Performance" category.
Page crashes or NO_FCPβ
Especially if a slow mobile device is emulated, tests can sometimes fail for complex pages as the CPU becomes blocked or the pages takes too long to render. Tests also time out after waiting 15 seconds for the First Contentful Paint.
If you run into this you can either select a faster test device or contact support to increase the timeout.
Book a demoβ
We can help you set up monitoring and give advice on how to optmize your website. Just book a call here.