 Collecting web performance data in the field has traditionally been a first party experience. You add some tracking code to your website and your monitoring service will show you insightful data about how users experience your website. The key is that you're only seeing data for users on your own website. But what would you do if you could see the real user performance data for your closest competitor, or other websites in your industry? How about 3 million websites? In this video, we're taking a closer look at the Chrome User Experience Report. Let's get started. The Chrome UX Report is a public repository of real user performance data. Chrome users who opt in have their page load performance aggregated at the origin level and published as a queryable dataset. Developers can mine this data to get insights into performance metrics including paint and load times across dimensions like form factor, effective connection type and geography. Now we've long had the ability to analyze page load performance with synthetic tools like web page test. But we've always had to tiptoe around drawn conclusions about the speed of the page because it was under such contrived test conditions. The special thing about the Chrome UX Report is we now have public data from real users over real network conditions. And unlike traditional real user measurements or RUM, we can now use this data to measure representative websites outside of our control like competing news sites, the entire news vertical, or even the web as a whole. So let's dive in and see it in action. To get started, we'll head over to BigQuery, which is basically our database in the cloud. The project consists of country-specific datasets along with a global dataset named ALL. Each dataset is divided into monthly tables each with several color fields. The origin of the website, which is unique to the protocol and domain, the form factor of the user's device broken down by phone, tablet, or desktop, the effective connection type, which represents the user's connection speed in mobile terms like 3G, 4G, or offline. And finally, the performance metrics, which include first paint, first contentful paint, DOM content loaded, and onload. The performance metrics are represented as histograms, where each bin has a start time, an end time, and a corresponding density. The densities for each combination of device and connection type add up to 100%. So for example, we can find out the percentage of page loads that paint in under a second by adding up the densities of all the bins whose start time is less than 1,000 milliseconds. We can run the same query against a different origin to compare which one has a higher density of fast experiences. So what can this data tell us about the state of the web? If we aggregate the densities for all origins where the start time is less than 1,000 milliseconds, that will tell us the approximate density of fast paint experiences on the web. In a few lines of SQL, we learn that it's 36%, meaning a third of first contentful paints occur in less than a second. Now keep in mind that these are not weighted by relative site popularity. This demonstrates that we've still got a long way to go before achieving fast web experiences everywhere. The good news is that the Chrome UX report enables us to track the progress of metrics like these over each monthly release. Out of the millions of origins in the dataset, things get really interesting when you look at competitors head to head. For example, let's compare the FCP distributions of Reddit and Hacker News. We can start with the previous query and simply group by the competing origins. By adding the average and slow densities, we get a more complete picture of the distribution and see that Reddit actually skews slightly slower. The Chrome UX report also has another hidden power beyond performance metrics. It can show the relative distributions of form factors and effective connection types. Building on our competitive analysis, let's see how users interact with these sites on different devices. Surprisingly, the sites have very different usage patterns on desktop and mobile. With Hacker News, 7030 in favor of mobile and Reddit, 7030 in favor of desktop. Both sites have a very small percent of tablet usage. Finally, let's use this data to make the comparison even more palpable. BigQuery enables us to export our results to a spreadsheet where we can easily generate a chart. Click Save to Google Sheets and the results will be transferred automatically. In a single click, we can visualize the raw data into a chart that really brings the form factor differences to life. These are just a few examples of the many super interesting insights to be gleaned from the Chrome UX report. So get in there and start exploring. You can learn more about the project and the docs on developers.google.com or feel free to reach out to me in the comments or on Twitter should you have any questions. Thanks for watching.