Web Performance 101 [Chrome Dev Summit 2020]

Tammy Everts

December 07, 2020

Tweet Share

Transcript

1. Web Performance 101: What is web performance and why should

   I care? @tameverts #ChromeDevSummit ¯\_(ツ)_/¯
2. None

3. @tameverts

4. None
5. None

6. speedcurve.com/benchmarks/

7. What is “web performance”? Why should I care about it?

   How do I measure it? How can I get other people in my company to care about it?

8. Slow websites suck.

9. the average web user believes they waste two days a

   year waiting for pages to load

10. “web stress” When apps or sites are slow, we have

    to concentrate up to 50% harder to stay on task. @tameverts

11. 11 When do users start to interact with a page?

12. 12 Source: Jakob Nielsen

13. 13 Source: Jakob Nielsen

14. 14

15. 15 “We want you to be able to flick from

    one page to another as quickly as you can flick a page on a book. So, we’re really aiming very, very high here… at something like 100 milliseconds.” Urs Hölzle SVP Engineering, Google

16. fast slow @tameverts

17. Slow pages affect people’s perception of three things completely unrelated

    to time: 1. Content “boring” 2. Visual design “tacky” “confusing” 3. Ease of navigation “frustrating” “hard-to-navigate”

18. User experience and web performance are predictable indicators of business

    outcomes.

19. Rebuilding Pinterest pages for performance resulted in 40% decrease in

    wait time, 15% increase in SEO traffic, and 15% increase in signup conversion rate. Ancestry.com saw a 7% increase in conversions after improving render time by 68%, page weight by 46% and load time by 64%. Staples reduced median page load time by 1 second and 98th percentile load time by 6 seconds, resulting in a 10% conversion rate increase. @tameverts
20. None

21. @tameverts

22. @tameverts

23. optimal load times for peak conversions @tameverts

24. even 100ms delays matter @tameverts

25. real user data + machine learning

26. collected 1M+ beacons of real user data across 93 attributes,

    including… • top-level – domain, timestamp, SSL • session – start time, length (in pages), total load time • user agent – browser, OS, mobile ISP • geo – country, city, organization, ISP, network speed • bandwidth • timers – base, custom, user-defined • custom metrics • HTTP headers

27. 1sessions that convert have fewer images

28. @tameverts

29. 2pages with more scripts are less likely to convert

30. 160+ scripts… uh-oh @tameverts

31. speakerdeck.com/csswizardry/its-my-third-party-and-ill-cry-if-i-want-to

32. How fast is fast enough?

33. “The real thing we are after is to create a

    user experience that people love and they feel is fast… and so we might be front-end engineers, we might be dev, we might be ops, but what we really are is perception brokers.” Steve Souders

34. But measuring perception is hard. It’s even harder to scale.

35. None

36. What tools do we use? Synthetic (lab) Consistent baseline Mimics

    network & browser conditions No installation Compare any sites Detailed analysis Waterfall charts Filmstrips and videos Limited URLs Real user monitoring (field) Requires JavaScript installation Large sample size (up to 100%) Real network & browser conditions Geographic spread Correlation with other metrics (bounce rate) No detailed analysis Only measure your own site
37. None
38. None
39. None
40. None

41. @tameverts

42. @tameverts

43. @tameverts

44. Free tools to explore Synthetic webpagetest.org developers.google.com/speed/ pagespeed/insights/ Real user

    monitoring github.com/bluesmoon/boomerang developers.google.com/web/tools/ chrome-user-experience-report

45. webpagetest.org

46. developers.google.com/speed/pagespeed/insights/

47. A brief history of performance metrics

48. None

49. TTFB DNS TCP TTI FCP FMP FID OMG WTF

50. ❑ Correlates to what users actually see in the browser

    ❑ Is easy to use and accessible right out of the box ❑ Recognizes that not all pixels and page elements are equal ❑ Allows us to customize what we measure on specific pages The best UX metric…
51. None

52. Is it happening? Is it useful? Is it usable? Is

    it delightful? developers.google.com/web/fundamentals/ performance/user-centric-performance-metrics

53. Load Time The time from the start of the initial

    navigation until the beginning of the window load event
54. None
55. None

56. Start Render The time from the start of the initial

    navigation until the first non-white content is painted
57. None

58. start render repeat visits

59. wow!

60. First Paint First Contentful Paint First Meaningful Paint

61. First Paint (FP) Pixels first start to render

62. First Contentful Paint (FCP) Text and graphics start to render…

    BUT often catches non-meaningful paints (e.g. headers, nav bars)

63. First Meaningful Paint (FMP) The paint after which the biggest

    ATF layout change has happened and web fonts have loaded

64. speedcurve.com/blog/an-analysis-of-chromiums-paint-timing-metrics/

65. Analysis of 40 top Alexa-ranked sites 95% of FP events

    occur before Start Render 85% of FCP events occur before Start Render 50% of FMP events occur before Start Render speedcurve.com/blog/an-analysis-of-chromiums-paint-timing-metrics/

66. Speed Index Average time at which visible parts of the

    page are in the viewport
67. None

68. ❑ Correlates to what users actually see in the browser

    ❑ Is easy to use and accessible right out of the box ❑ Recognizes that not all pixels and page elements are equal ❑ Allows us to customize what we measure on specific pages The best UX metric…

69. Custom metrics Measure performance with high-precision timestamps Available in both

    synthetic and RUM (yay!) https://www.w3.org/TR/user-timing/ https://speedcurve.com/blog/user-timing-and-custom-metrics/

70. how long does it take to display the main product

    image on my site?

71. Time to First Tweet The time from clicking the link

    to viewing the first tweet on each page’s timeline Pinner Wait Time (PWT) The time from initiating an action (e.g., tapping a pin) until the action is complete (pin close-up view is loaded) Time to Interact (TTI) @tameverts
72. None

73. Lighthouse Scores based on audits run on synthetic tests. Checks

    your page against “rules” for Performance, PWA, Best Practices, and SEO. For each category, you get a score out of 100 and recommendations for what to fix. developers.google.com/web/tools/lighthouse
74. None
75. None

76. matuzo.at/blog/building-the-most-inaccessible-site-possible- with-a-perfect-lighthouse-score/

77. None

78. Core Web Vitals

79. None

80. developers.google.com/search/blog/2020/11/timing-for-page-experience

81. None

82. “Core Web Vitals are the subset of Web Vitals that

    apply to all web pages, should be measured by all site owners, and will be surfaced across all Google tools. “Each of the Core Web Vitals represents a distinct facet of the user experience, is measurable in the field, and reflects the real-world experience of a critical user-centric outcome. “The metrics that make up Core Web Vitals will evolve over time. “The current set for 2020 focuses on three aspects of the user experience — loading, interactivity, and visual stability — and includes the following metrics… web.dev/vitals/

83. Largest Contentful Paint First Input Delay Cumulative Layout Shift loading

    interactivity visual stability

84. 75th percentile of page loads across mobile and desktop

85. Amount of time it takes for the largest visual element

    to render. Available in Chrome and Chromium-based browsers. Measurable via synthetic and RUM.
86. None

87. Amount of time it takes for page to respond to

    user input (e.g. click, tap, key). Only measurable via RUM.

88. FID can seem fast because user interactions take place later

    in the page’s rendering cycle... after CPU-hogging long tasks have completed. speedcurve.com/blog/first-input-delay-google-core-web-vitals/

89. No correlation when looking at all sessions speedcurve.com/blog/first-input-delay-google-core-web-vitals/

90. Stronger correlation at 75th percentile speedcurve.com/blog/first-input-delay-google-core-web-vitals/

91. Long Tasks have a high correlation across the board speedcurve.com/blog/first-input-delay-google-core-web-vitals/

92. Long Tasks Measures JavaScript functions that take 50ms or longer.

    Long or excessive JS tasks can delay rendering, as well as cause page “jank”. Measurable via synthetic and RUM.

93. Score that reflects how much page elements shift during rendering.

    Available in Chrome and Chromium-based browsers. Measurable via synthetic and RUM.
94. None

95. Size of the shifting element matters speedcurve.com/blog/visualising-cls-layout-shifts/

96. Image carousels can generate false positives speedcurve.com/blog/visualising-cls-layout-shifts/

97. Web fonts & opacity changes can cause issues speedcurve.com/blog/visualising-cls-layout-shifts/

98. Bounce rate gets worse as CLS degrades Bounce rate improves

    as CLS degrades Bounce rate stays the same as CLS degrades @tameverts
99. None

100. Look at your own data.

101. How do these metrics correlate with my business goals? How

     fast should they be? How do we stay on track?
102. None

103. cnet.com/news/appliance-science-the-well-done-physics-chemistry-of-the-toaster/

104. None

105. How to create a culture of performance

106. “The largest hurdle to creating and maintaining stellar site performance

     is the culture of your organization. Lara Hogan designingforperformance.com

107. “No matter the size or type of team, it can

     be a challenge to educate, incentivize, and empower those around you. “Performance more often comes down to a cultural challenge, rather than simply a technical one.” Lara Hogan designingforperformance.com

108. Educate Incentivize Empower

109. None
110. None

111. 2009 Improved average load time from 6s à 1.2s 7-12%

     increase in conversion rate + 25% increase in PVs Average load time degraded to 5s User feedback: “I will not come back to this site again.” Re-focused on performance 0.4% increase in conversion rate 2010 2011 @tameverts

112. 1. No front-end measurement 2. Constant feature development 3. Badly

     implemented third-parties 4. Waiting too long to tackle performance problems 5. Relying on performance sprints

113. 1. You need a plan

114. Making it up as you go is not always a

     good idea. (Actual photo taken yesterday of my family’s gingerbread village.)

115. Tools ≠ Enough

116. 2. Have a champion higher up

117. None

118. 3. Then build a cross-disciplinary team

119. None

120. Everyone who touches a page should care about the performance

     of that page.

121. Embrace performance from the ground up. Embed engineers into other

     teams. Enlist performance ambassadors. Teach people how to use (or at least understand) the monitoring tools you use.

122. 4. Set shared goals

123. It’s perilously easy to accidentally become a gatekeeper.

124. We first went to the engineering leaders, and then we

     went to our product leader. Our pitch was totally different... Reefath Rajali // PayPal chasingwaterfalls.io/episodes/episode-two-with-reefath-rajali/

125. “When we went to our product leaders, we spoke more

     about the business numbers and the business benefits. “When we spoke to our engineering leaders, it was more about our consumer delight.” Reefath Rajali // PayPal chasingwaterfalls.io/episodes/episode-two-with-reefath-rajali/

126. Find out what people care about

127. ❑ bounce rate ❑ cart size ❑ conversions ❑ revenue

     ❑ time on site ❑ page views ❑ SEO ❑ user happiness ❑ user retention ❑ competitors

128. If they care about business metrics…

129. None
130. None
131. None
132. None

133. If they care about user engagement…

134. None
135. None

136. If they care about SEO…

137. None

138. If they care about third parties…

139. None
140. None

141. Who they are What they care about What to show

     them Executives Competition Business impact Benchmarks (filmstrips and videos) Correlation charts (perf + KPIs) Marketing Third parties Traffic + engagement SEO Content Third-party performance Correlation charts (perf + bounce rate) Lighthouse SEO audits Image size Devs / engineers Well, lots of stuff, probably Consult with perf team

142. 5. Make everyone accountable

143. Performance budgets FTW!

144. Thresholds YOU create for metrics that are meaningful for YOUR

     site addyosmani.com/blog/performance-budgets/ Milestone timings (e.g. start render) Quantity-based (e.g. image weight) Rules-based (e.g. Lighthouse scores)
145. None

146. A good performance budget should show you… What your budget

     is When you go out of bounds How long you’re out of bounds When you’re back within budget

147. zillow.com/tech/bigger-faster-more-engaging-budget/

148. zillow.com/tech/bigger-faster-more-engaging-budget/

149. Super important! Look at your own data Monitor your competitors

     No sandbagging allowed Take a step-by-step approach if necessary Use synthetic and RUM (numbers may will vary)

150. Pro tips Create budgets for your popular and regularly changing

     pages Review violations early and always Compare before and after releases Update budgets accordingly zillow.com/engineering/bigger-faster-more-engaging-budget/

151. Who What Metric Ops Back-end issues TTFB Marketing Most important

     content Third parties SEO Largest Contentful Paint JS Long Tasks Lighthouse SEO score & audits Devs / engineers How well pages are built Performance issues Start Render, Web Vitals Lighthouse Performance audits

152. Give people ownership

153. “One of the original directives of the performance team was

     we weren’t going to set ourselves up to be performance cops.” Dan Chilton, Vox Media responsivewebdesign.com/podcast/vox-media-performance/

154. “We weren’t going to go around slapping people on the

     wrist, saying, ‘You built an article that broke the page size budget! You have to take that down or change that immediately!’ “Our goal setting out was to set up best practices, make recommendations, and be a resource within the company that people can turn to when they have to make performance-related decisions.” Dan Chilton, Vox Media responsivewebdesign.com/podcast/vox-media-performance/

155. 6. Communicate

156. “We, as engineers, should learn how to show the impact

     on anything we do.” Malek Hakim // Priceline chasingwaterfalls.io/episodes/episode-one-with-malek-hakim/
157. None
158. None
159. None
160. None
161. None
162. None

163. How often is often enough? Wall monitors and dashboards 24/7

     Alerts (to people who can make fixes) in realtime Reports no more than 1X week Meetups, hackathons, etc. monthly (if possible)

164. 7. Don’t forget to celebrate!

165. None
166. None
167. None

168. !!!

169. medium.com/the-telegraph-engineering

170. Score some easy wins

171. “The dull boring stuff” ~Andy Davies Scripts (especially third parties)

     Images Extraneous code Defer assets where possible

172. Shaved 15KB off logo Ran A/B test Increased bookings chasingwaterfalls.io/episodes/episode-one-with-malek-hakim/

173. In summary…

174. There’s no magic. Show up with a plan. Do the

     work. (Be patient.)

175. Thanks! @tameverts speedcurve.com/blog