In the performance world, Lighthouse and Webpagetest are two of the most commonly used tools, which are preferred by the developers to debug the performance issues. By definition, these tools are synthetic tools. This means that they are run either on a dev machine or a lab, on a specific browser, specific url and connection. The limitation of such systems is that they are not always able to detect and reproduce particular issues happening on the actual real user site.
One of the customers on Gemini recently faced a Core Web Vitals issue, where the Gemini tool detected significant degradation in the Cumulative Layout Shift (CLS). Gemini raised Core Web Vital Alerts automatically and also gave preliminary root cause analysis.
This site is an E-commerce site and the Gemini alert algorithm was instantly able to detect which page template faced the issue. In this case, it was the ‘Product Details Page’.
With this information, the developers of the E-commerce site started looking at debugging the issue. However they were not able to reproduce the issue. Whenever the Url was run in the Lighthouse of the developer machine, the CLS score appeared in green.
The Webpagetest of the some random Product Details Page Url also did not show in the degradation. The team was unable to reproduce the problem yet it was happening for the users.
This is where the Scatter Plot module with the Waterfall view from Gemini tool came to rescue. Gemini captures almost all real user data samples for analysis by the developers. By clicking on the point in the below graph, one is able to check the Core Web Vital conditions of each user condition anonymously.
Picking up a sample which had the CLS , gave all the user parameters which lead to the CLS .
Not just the browser and OS version, the user agent and the actual url with various query strings was also captured. It turned out that the E-commerce website was running some campaigns on Facebook and Instagram, where in particular cases the CLS was happening.
Armed with this useful information, the developers were able to reproduce the problem locally and were able to fix it instantly. In this case, the problem was happening on a campaign being served to Facebook and Instagram users. The campaign was triggered in random cases based on the campaign criteria. There was no way for the developers to know about the campaign and the criteria on which it was running. Gemini was able to detect the issue instantaneously and was able to help identify the condition for reproducing it.
For any digital website, we recommend having real user monitoring through Gemini tool. Contact us if you are interested in the free trial.