「Yahoo!検索」におけるWebパフォーマンス改善の取り組み / Efforts to Improve Web Performance in "Yahoo! JAPAN Search"

Yahoo!検索のWebパフォーマンスの可視化の取り組み、および可視化したパフォーマンスを用いた改善の取り組みについてご紹介します。

1. Efforts to Improve Performance Using Core Web Vitals in "Yahoo! JAPAN Search" LY Corporation Search Company, Search Headquarters, Development Headquarters Hatsue Okudaira
2. • Yahoo! JAPAN Search • what are Web Vitals • 3 metrics of Core Web Vitals • how to visualize Core Web Vitals • utilize our services • conclusion Agenda
3. Yahoo! JAPAN Search
4. Name • Yahoo! JAPAN Search Service Description • This service allows users to swiftly and accurately find their desired information from the vast amount of data available on the internet. Yahoo! JAPAN Search
5. Service Features • Simultaneous development of diverse functions on a single page • Module display positions vary with each search Yahoo! JAPAN Search
6. What are Web Vitals?
7. Google's initiative to provide a unified guideline on quality signals considered important for delivering an excellent user experience on the web What are Web Vitals？
8. • Core Web Vitals are a set of metrics that measure real user experience in terms of page loading performance, interactivity, and visual stability. • Improving Core Web Vitals can enhance search rankings and provide an overall better user experience. • Core Web Vitals impact UX and SEO. What are Core Web Vitals？
9. Impact of Core Web Vitals on Business Click- through rate Non-Click- through rate impression
10. 3 metrics of Core Web Vitals
11. • Largest Contentful Paint（LCP） • Interaction to Next Paint (INP) • Cumulative Layout Shift（CLS） 3 metrics of Core Web Vitals
12. • This is a metric that measures the time it takes for the largest content to be displayed on a page. A low LCP indicates that the page loads quickly. • A good LCP value is less than 2.5 seconds. Largest Contentful Paint（LCP）
13. • This is a metric that measures the responsiveness of a page to user inputs. • Pages with a low INP indicate to users that the page responds quickly to interactions, enhancing the website's reliability and comfort. • A good INP value is less than 200 ms. Interaction to Next Paint (INP)
14. • This is a metric that monitors unexpected changes in page layout to measure layout stability. • A low CLS indicates that the page layout is stable and does not change unexpectedly when the user tries to interact with it. • A good CLS score is less than 0.1. Cumulative Layout Shift（CLS）
15. how to visualize Core Web Vitals
16. 2 data categories Lab data Field data
17. • Analysis in a specific environment • Measurement of representative queries • Consistent conditions for measurement • Useful for identifying areas for improvement • May deviate from actual user experience Lab data
18. • Collected from actual users' browsers • Reflects real performance conditions • Classification and analysis of queries by domain • Convenient for identifying areas needing improvement • Useful for understanding general trends Field data
19. utilize our services
20. Dashboard Review: • Divide queries by major areas Performance Evaluation: • Identify areas with relatively poor performance compared to others • Check areas that have worsened compared to the previous week Identification of Areas for Improvement: • Determine areas that need improvement Analysis Methods Utilizing Field Data
21. • Log HTML elements that have impacted queries and performance • Narrow down to the relevant query area to identify queries and HTML elements affecting performance Analysis Methods Utilizing Field Data
22. Improvement examples
23. • Check the field data • Confirm that the CLS value in the person information area is worse compared to other areas • Check and aggregate the CLS log information for the person area Case1：Cumulative Layout Shift（CLS）
24. • Identify that layout shifts commonly occur in many queries for elements within the red frame on the screen • After identifying the element, use Lighthouse to confirm the occurrence of layout shifts Case1：Cumulative Layout Shift（CLS）
25. • Check the field data and confirm that the LCP value for queries displaying regional information is worse compared to other areas Case2：Largest Contentful Paint（LCP）
26. • Pick up several queries displaying regional information • Review the Lighthouse results • Identify that the large image displayed at the top of the screen is the bottleneck Case2：Largest Contentful Paint（LCP）
27. conclusion
28. • Share the knowledge and case studies handled • Establish a system to notify when performance metrics deteriorate • Conduct workshops to enable performance analysis and improvement Initiatives for the future
29. ❌Perform Performance Improvements ⭕Build a Culture of Performance Improvement