Yo ho ho and a few billion pageviews of RUM
19 likes6,161 views
The document discusses website performance, emphasizing a shift from backend to frontend bottlenecks and the importance of real user measurement in correlating site performance with business metrics. It presents data from over a billion pageviews, highlighting variations in load times across different states and countries, and the impact of speed on bounce rates and user engagement. Key takeaways include the necessity of accurate measurement for performance improvement and the significance of analyzing one's own data.
Yo ho ho and a few billion pageviews of RUM
- 1. Yo ho ho and a few billion pageviews of RUM Josh Fraser josh@torbit.com © 2012 Torbit 1
- 2. Introductions Josh Fraser Co-founder & CEO of Torbit. Born in Scotland. Computer geek. Addicted to speed. @joshfraser josh@torbit.com Torbit Next generation measurement, analytics and optimization. We’re here to make the web faster. @torbit torbit.com © 2012 Torbit 2
- 3. Performance in 2012 Performance bottlenecks have moved from the backend to the frontend © 2012 Torbit 3
- 4. Problem Limitations of synthetic testing (active monitoring) Limited visibility Expensive Strains the systems you are testing Limited test locations and browsers Backbone connections paint an overly optimistic picture Can be easily gamed by CDNs or other providers © 2012 Torbit 4
- 5. Problem Website performance is viewed as a technical metric instead of a business metric © 2012 Torbit 5
- 6. A better idea Real User Measurement Use JavaScript to measure every visitor on every page and correlate how site performance impacts the business metrics you care about most © 2012 Torbit 6
- 7. A better idea What we’ve learned since launching Torbit Insight © 2012 Torbit 7
- 8. The golden rule of performance Backend 7% Frontend 93% Using TTFB as the marker between frontend/backend © 2012 Torbit 8
- 9. About our data Collected from a sample of 1,000 websites that have installed Torbit Insight Here are a handful of them... © 2012 Torbit 9
- 10. About our data 1,000-site sample of 6.7 billion pageviews Collection via the Episodes library from Steve Souders http://stevesouders.com/episodes2 Global audience © 2012 Torbit 10
- 11. More accurate measurement Navigation Timing Start the clock before the page arrives © 2012 Torbit 11
- 12. Timing data sources Navigation Timing (window.performance) Cookie 5% Inline JS 44% Nav Timing 50% Google Toolbar 1% Supported in Firefox 7, Chrome 6, Internet Explorer 9 and Android 4 © 2012 Torbit 12
- 13. Variables that impact speed Your visitors Your website Browser CDN Geography # of resources on the page Connection type Size of resources on the page Hardware performance Third party JavaScript Virus protection software © 2012 Torbit 13
- 14. Overall stats Geometric mean 2.27s 90th percentile 10.52s Median 2.67s 95th percentile 16.87s Average 5.07s 99th percentile 43.55s © 2012 Torbit 14
- 15. Overall stats (mobile) Geometric mean 3.12s 90th percentile 12.07s Median 3.87s 95th percentile 18.11s Average 6.23s 99th percentile 44.42s © 2012 Torbit 15
- 16. Which states in the USA have the fastest load times? The slowest? © 2012 Torbit 16
- 17. US speed by state Onload © 2012 Torbit 17
- 18. US speed by state Fastest Slowest Massachusetts 3.98s Alaska 6.14s Rhode Island 4.24s Mississippi 6.03s Delaware 4.40s West Virginia 5.68s Oregon 4.52s Idaho 5.60s New Hampshire 4.52s Arkansas 5.50s Washington 4.59s Wyoming 5.47s Virginia 4.61s Montana 5.44s New York 4.63s Kentucky 5.35s Connecticut 4.64s New Mexico 5.30s Minnesota 4.68s Alabama 5.29s © 2012 Torbit 18
- 19. Which countries have the fastest load times? The slowest? © 2012 Torbit 19
- 20. Global speeds by country Fastest Slowest Slovenia 3.41s Tonga 31.49s Sweden 4.02s Cuba 30.59s Denmark 4.27s Vanuatu 28.65s Canada 4.27s Niger 27.05s Switzerland 4.32s Burkina Faso 26.83s Netherlands 4.34s Burma 24.87s Belgium 4.39s Liberia 23.78s Norway 4.45s Sierra Leone 23.30s Aaland Islands 4.48s Gambia 22.13s Iceland 4.58s Micronesia 22.12s © 2012 Torbit 20
- 21. The United States? #22 © 2012 Torbit 21
- 22. Antarctica? © 2012 Torbit 22
- 23. Global speeds by continent North America 5.08s Europe 5.22s Oceania 7.22s Asia 8.11s Africa 10.57s South America 10.90s Antarctica * 22.42s * Not a statistically significant sample ** Data could be skewed by the geography of our customer base (mostly US) © 2012 Torbit 23
- 24. Fastest desktop browser? © 2012 Torbit 24
- 25. Fastest desktop browser Fastest onload time Safari 4.47 Firefox 4.79 Chrome 5.53 IE 8 6.1 Opera 6.44 IE 9 6.89 IE 7 7.39 IE 6 9.63 0 3 5 8 10 © 2012 Torbit 25
- 26. Fastest mobile browser? © 2012 Torbit 26
- 27. Fastest mobile browser © 2012 Torbit 27
- 28. Fastest mobile browser Opera Mini 4.68 Chrome iOS 5.82 Safari on iPhone 6.6 Android browser 7.16 Chrome on Android 8.7 0 2 5 7 9 © 2012 Torbit 28
- 29. What really matters... How does speed affect your business metrics? © 2012 Torbit 29
- 30. How speed affects bounce rate © 2012 Torbit 30
- 31. How speed affects bounce rate Error pages © 2012 Torbit 30
- 32. How speed affects bounce rate y = 0.6517x + 33.682 R² = 0.91103 Error pages © 2012 Torbit 30
- 33. Every second = 0.65 increase in bounce rate © 2012 Torbit 31
- 34. Where to focus © 2012 Torbit 32
- 35. Where to focus y = 2.4x + 27.4 R² = 0.97297 © 2012 Torbit 32
- 36. Want to reduce your bounce rate by 10%? Make your site load in 2 seconds instead of 6 © 2012 Torbit 33
- 37. How speed affects bounce rate (mobile) © 2012 Torbit 34
- 38. How speed affects bounce rate (mobile) The only pages that load in 1 second on a mobile device are error pages © 2012 Torbit 34
- 39. How speed affects user engagement © 2012 Torbit 35
- 40. Want to double your user engagement? Make your site load in 2 seconds instead of 6 © 2012 Torbit 36
- 41. How speed affects user engagement (mobile) © 2012 Torbit 37
- 42. How speed affects user engagement (mobile) Again, a 1 second load time on a mobile device is rarely a good thing © 2012 Torbit 37
- 43. What’s next? Resource timing Detect which resource is taking too long Coming soon to a browser near you © 2012 Torbit 38
- 44. Takeaways Every second counts If you want accurate data, use Real User Measurement It’s important to really dig into your data One second load times aren’t always a good thing You need measurement first before you can improve Nothing beats looking at your own data © 2012 Torbit 39
- 45. Measure your own speed Get your free RUM at torbit.com © 2012 Torbit 40
- 46. Thank you! Josh Fraser josh@torbit.com © 2012 Torbit 41
Editor's Notes
- #2: \n
- #3: \n
- #4: \n
- #5: \n
- #6: \n
- #7: \n
- #8: \n
- #9: \n
- #10: \n
- #11: \n
- #12: \n
- #13: \n
- #14: \n
- #15: \n
- #16: \n
- #17: \n
- #18: \n
- #19: \n
- #20: \n
- #21: \n
- #22: \n
- #23: \n
- #24: \n
- #25: \n
- #26: \n
- #27: \n
- #28: \n
- #29: \n
- #30: \n
- #31: \n
- #32: \n
- #33: \n
- #34: \n
- #35: \n
- #36: \n
- #37: \n
- #38: \n
- #39: \n
- #40: \n
- #41: \n
- #42: \n
- #43: \n
- #44: \n
- #45: \n
- #46: \n
- #47: \n
- #48: \n