What your Google Lighthouse score hides from you - BrightonSEO

In this BrightonSEO talk, Aymen will tell us about the challenges he encounters in web performance when we rely heavily on synthetic tools and their scoring systems. These tools are great for debugging and testing web performance but can't represent your website users' real life. Join him on his retrospective on web performance measurement methodology. He will share his findings on how to fix these challenges and better understand users' experience thanks to his workflow and the Real User Monitoring tool (Speetals) he built.

You can watch this talk recording for free by subscribing to BrightonSEO live event: https://april2023.brightonseo.com/talks/what-your-google-lighthouse-score-hides-from-you/

Transcript

1. What Google Lighthouse score hides from you SpeakerDeck.Com/AymenLoukil @LoukilAymen Aymen Loukil Speetals Founder
2. @LoukilAymen #BrightonSEO Back to 2017, my webperf consulting journey…
3. @LoukilAymen #BrightonSEO And this was my #webperf workflow Audit in Lab Write user stories Implement Test in Lab
4. @LoukilAymen #BrightonSEO Awesome!
5. @LoukilAymen #BrightonSEO Me excited Ship it!
6. @LoukilAymen #BrightonSEO Now what?
7. @LoukilAymen #BrightonSEO We need to wait for the next CrUX dataset (in 28 days) [Every second Tuesday]
8. @LoukilAymen #BrightonSEO In 1 month many things happen..
9. @LoukilAymen #BrightonSEO Users faced improvement deterioration! Feb 2020 Mar 2020 73% => 70.8%
10. @LoukilAymen #BrightonSEO Feb 2020 Mar 2020 Me disappointed
11. @LoukilAymen #BrightonSEO A shot in the dark!
12. @LoukilAymen #BrightonSEO As a consultant, This feels bad and frustrating.
13. @LoukilAymen #BrightonSEO Reflection was needed to understand what happened?
14. @LoukilAymen #BrightonSEO Gabriela Palai - Pexels What’s wrong with Lighthouse score…?
15. @LoukilAymen #BrightonSEO Where does it fall in our users? We have 100 users LCP 4s to 2.5s on mobile
16. @LoukilAymen #BrightonSEO Does it represent someone? We have 100 users LCP 4s to 2.5s on mobile
17. @LoukilAymen #BrightonSEO One LH run may or not represent one of your users
18. @LoukilAymen #BrightonSEO Emulation isn’t real life..
19. @LoukilAymen #BrightonSEO
20. @LoukilAymen #BrightonSEO I’ve also been part of this!
21. @LoukilAymen #BrightonSEO Webperf score obsession isn’t new!
22. @LoukilAymen #BrightonSEO We have always been assessed with grades..
23. @LoukilAymen #BrightonSEO Do they even correlate in some way?
24. @LoukilAymen #BrightonSEO 50% of perfect Lighthouse score (90+) pages fail at Web Vitals Brendan Kenny @brendankenny
25. @LoukilAymen #BrightonSEO With a score of 50/100, many pages passes the Web Vitals Brendan Kenny @brendankenny
26. @LoukilAymen #BrightonSEO There is a blind spot!
27. @LoukilAymen #BrightonSEO What do your users and Lighthouse have in common?
28. @LoukilAymen #BrightonSEO CPU Memory Screen size Location Internet speed Device Context flaticon.com
29. @LoukilAymen #BrightonSEO Users interact, but Lighthouse doesn’t!
30. @LoukilAymen #BrightonSEO Are they at least talking the same metrics?
31. @LoukilAymen #BrightonSEO Ultimately they don’t share as many metrics!
32. @LoukilAymen #BrightonSEO Ultimately they don’t share as many metrics! Above the fold only CLS
33. @LoukilAymen #BrightonSEO Ultimately they don’t share as many metrics! Above the fold only CLS Depends on conditions
34. @LoukilAymen #BrightonSEO Variability issues 3 consecutive runs on the same page and on same conditions
35. @LoukilAymen #BrightonSEO It is like Rock paper scissors
36. @LoukilAymen #BrightonSEO You can easily trick it! Barry Pollard @tunetheweb
37. @LoukilAymen #BrightonSEO My customer’s Web agency
38. @LoukilAymen #BrightonSEO You can even hack it!
39. @LoukilAymen #BrightonSEO If the page is loaded with Lighthouse, Gtmetrix, or other tool => Only load HTML, no styles, no JavaScript, no images
40. @LoukilAymen #BrightonSEO Do you mainly care about your users or Lighthouse
41. @LoukilAymen #BrightonSEO Your users are different
42. @LoukilAymen #BrightonSEO CPUs are getting faster every year… 4G and 5G are common… Our audience top device is iPhone.. But..we often assume
43. @LoukilAymen #BrightonSEO You are just making assumptions!
44. @LoukilAymen #BrightonSEO iPhone isn’t real life…(Alex Russel) Your users potentially have 2019 smartphones Android iOS
45. @LoukilAymen #BrightonSEO A user does not care about your website Lighthouse score, your technical stack, or even if it is AI-powered..
46. @LoukilAymen #BrightonSEO So..Rely on your Users’ data, And here is how👉
47. @LoukilAymen #BrightonSEO Remember my failed #webperf workflow? Audit in Lab Write user stories Implement Test in Lab https://www.vectorstock.com/ https://knowyourmeme.com/
48. @LoukilAymen #BrightonSEO I transformed it into this: Audit in field Prioritize Implement Test in Lab Fast validate on users Iterate
49. @LoukilAymen #BrightonSEO Audit in field Prioritize Implement Test in Lab Fast validate on users Hear your users experience, validate your effort with them Iterate Now you can use Lighthouse..
50. Audit in the field (Hear your users experience) STEP 1
51. @LoukilAymen #BrightonSEO Identify improvement points for each metric/ device
52. @LoukilAymen #BrightonSEO Identify improvement points for each metric/ device Detect distribution gaps between devices
53. @LoukilAymen #BrightonSEO See how your audience LCP distributes
54. @LoukilAymen #BrightonSEO See how your audience distributes 75% of your users are facing a 6.1s LCP loading!!
55. @LoukilAymen #BrightonSEO Compare against competition’ audience experiences
56. @LoukilAymen #BrightonSEO How my geo-audience is experiencing it
57. Prioritize. (Do the right thing first) STEP 2
58. @LoukilAymen #BrightonSEO Sort your top page-types per metric Domain data Page-level data
59. @LoukilAymen #BrightonSEO Identify low-hanging fruits Start with this listing page template
60. Implement & Audit in lab. (Implement, test) STEP 3
61. @LoukilAymen #BrightonSEO Audit in field Prioritize optimizations Implement Test in Lab Fast validate on users Iterate Now you can use Google Lighthouse!
62. Validate the outcomes. (Don’t wait 28 days..) STEP 4
63. @LoukilAymen #BrightonSEO 34%+ Desktop CLS increase in 10 days - PDP Validated Deployed
64. @LoukilAymen #BrightonSEO +7% Mobile LCP improvement in one week - PLP
65. @LoukilAymen #BrightonSEO +7% Mobile LCP improvement in one week - PLP Validated Deployed
66. @LoukilAymen #BrightonSEO LCP P75th percentile chart going down! 4.2s => 2.3s
67. @LoukilAymen #BrightonSEO Feb 2020 Mar 2020 Remember this frustration?
68. @LoukilAymen #BrightonSEO Fast detect regressions - CLS on PDP 😞
69. @LoukilAymen #BrightonSEO Set up regression Slack alerts
70. @LoukilAymen #BrightonSEO Validate monthly to confirm the outcomes Domain-level LCP on Mobile
71. Thank you! Aymen Loukil @LoukilAymen https://speetals.com International SEO and Web Performance Consultant