Through the Lens of Google CrUX: Dissecting Web Browsing Experience Across Devices and Countries

TL;DR

The paper leverages the Google CrUX dataset to quantify real-world web QoE across three device types and nine European countries, focusing on four key timing metrics: FP, FCP, DCL, and OL. It employs monthly timeframes and segmentations by Effective Connection Type, Device Type, and Country to analyze both cross-sectional and longitudinal patterns. Key findings show a consistent desktop advantage across metrics, with Sweden and Finland achieving the best mobile QoE due to high 4G shares and outperforming other countries by about 25–36% at the $75^{th}$ percentile. The work contributes a reproducible measurement setup, highlights limitations of CrUX aggregation and browser bias, and outlines directions for finer granularity, origin- and browser-specific analyses to guide future research and practical optimization.

Abstract

User quality of experience in the context of Web browsing is being researched widely, with plenty of developments occurring alongside technological advances, not seldom driven by big industry players. With the huge reach and infrastructure of Google, the Chrome User Experience Report (CrUX) provides quantitative real-life measurement data of a vast magnitude. Analysis of this steadily expanding dataset aggregating different user experience metrics, yields tangible insights into actual trends and developments. Hence, this paper is the first to study the CrUX dataset from the viewpoint of relevant metrics by quantitative evaluation of users Web browsing experience across three device types and nine European countries. Analysis of data segmented by connection type in the device dimension shows desktops outperforming other device types for all metrics. Similar analysis in the country dimension, shows North European countries (Sweden, Finland) having maximum 4G connections (85.99%, 81.41% respectively) and steadily performing 25%-36% better at the 75th percentile across all metrics compared to the worst performing country. Such a high-level longitudinal analysis of real-life Web browsing experience provides an extensive base for future research.

Figures (6)

• Figure 1: Growth of the Internet from 2010 to 2023 both in terms of number of URLs (a, top) and median website size in KB (b, bottom) as detected by HTTParchive HTMLarchive2.
• Figure 2: Performance graph generated by Google Chrome's built-in developer tools for https://www.google.com load timings, showcasing its relative temporal relation on a 4G type connection across different metrics.
• Figure 3: Metrics across connection types for the dataset. Higher speeds result in direct improvement for every metric.
• Figure 4: CDFs of all metrics across devices types, for the whole dataset duration. A clear classification is visible - desktop outperforms phones, which in turn outperform tablets - for all metrics.
• Figure 5: Proportions of connection speed type segmented by device. A common trend among all devices is the increasing predominance of 4G type connections.