Upgrade to Pro — share decks privately, control downloads, hide ads and more …

From Milliseconds
 to Millions: A Look at the Numbers Powering Web Performance

Harry Roberts
November 08, 2019
Technology
1
2.1k

From Milliseconds
 to Millions: A Look at the Numbers Powering Web Performance

We all know performance is big business, but how big? Let’s take a look at some of the numbers powering the web performance industry from both sides of the table. What do performance improvements mean for my clients, and how do we translate that into a working relationship?

Harry Roberts

November 08, 2019
Tweet

More Decks by Harry Roberts

See All by Harry Roberts
cache rules everything
csswizardry
1
2.1k
My Website Is Slow!
 Where Do I Start?
csswizardry
2
250
Optimising Largest Contentful Paint
csswizardry
8
2.3k
Get Your Head Straight
csswizardry
13
17k
More Than You Ever Wanted to Know About Resource Hints
csswizardry
6
8.5k
It’s My (Third) Party, and I’ll Cry if I Want To
csswizardry
13
5.1k
FaCSSt: CSS & Performance
csswizardry
26
3.7k
Vim for Front-end Developers
csswizardry
12
5.9k
Why Fast Matters
csswizardry
24
8.5k

Other Decks in Technology

See All in Technology
On Your Data を超えていく！
hirotomotaguchi
2
650
オーナーシップを持つ領域を明確にする
konifar
13
3.1k
サーバー間 GraphQL と webmock-graphql の話 / server-to-server graphql and webmock-graphql
qsona
2
180
反実仮想機械学習とは何か
usaito
PRO
8
3k
Java EE/Jakarta EEの現状と将来 ―クラウドネイティブ時代にJava EEは対応できるのか？―
takakiyo
1
130
Azure Container Apps + Bicep 〜 こんな感じで運用しています
kaz29
2
440
検証を通して見えてきたTiDBの性能特性
lycorptech_jp
PRO
6
3.7k
最近たまに見かけるTiDBってなんだ？ - Findy
pingcap0315
2
760
家族アルバム みてねにおけるGrafana活用術 / Grafana Meetup Japan Vol.1 LT
isaoshimizu
0
130
推しは推せるときに推せ！ プロダクトにフィードバックしていこう
nakasho
0
280
Terraformあれやこれ/terraform-this-and-that
emiki
8
1.3k
Compose Compiler Metricsを使った実践的なコードレビュー
tomorrowkey
1
220

Featured

See All Featured
Adopting Sorbet at Scale
ufuk
68
8.6k
From Idea to $5000 a Month in 5 Months
shpigford
377
45k
Practical Orchestrator
shlominoach
182
9.7k
BBQ
matthewcrist
80
8.8k
Building Your Own Lightsaber
phodgson
99
5.7k
Creatively Recalculating Your Daily Design Routine
revolveconf
210
11k
Docker and Python
trallard
34
2.7k
Build your cross-platform service in a week with App Engine
jlugia
225
17k
Design by the Numbers
sachag
274
18k
Unsuck your backbone
ammeep
663
57k
We Have a Design System, Now What?
morganepeng
43
6.7k
Designing for humans not robots
tammielis
248
25k

Transcript

  1. From Milliseconds to Millions A Look at the Numbers Powering

    Web Performance

  2. Hi, I’m Harry Consultant Performance Engineer Leeds, UK @csswizardry

  3. None

  4. From clearfixes to waterfalls.

  5. Ask the Right Questions

  6. I don’t have all of the answers.

  7. Ask the Right Questions How do you know the site

    is slow? What key areas of the site should I look at? What will it being faster mean for the business? How do you intend to measure this?

  8. Nobody Wants a Faster Website

  9. They want the most effective website.

  10. Motives Increase revenues Improve engagement Drive sign-ups Raise conversions Solicit

    donations Deliver content

  11. Case Study

  12. We want to be faster but still run the same

    number of ads. Publishing Client “

  13. 5HYHQXH $PRXQWRI$GVDQG7UDFNLQJ (QJDJHPHQW $PRXQWRI$GVDQG7UDFNLQJDQG5HYHQXH Ads Engagement Revenue

  14. How much can we get away with?

  15. None

  16. You Can’t Fix What You Can’t Measure

  17. Gather Data Spend time gathering statistics Then spend some more

  18. None

  19. Google Analytics Average data means means—we want percentiles Only samples

    1% of sessions—insufficient for many sites Far too coarse—aggregates every device, browser, page, region Focuses on load times—considered a legacy metric

  20. Any data is better than none.

  21. Case Study

  22. Do you know about your Venezuela problem? “

  23. Do you know about your Venezuela problem? “

  24. None

  25. Venezuela Economy is tanked Government minted Petro Electricity is virtually

    free
  26. None
  27. None
  28. None
  29. None
  30. None
  31. None

  32. Follow the Numbers

  33. Numbers seldom lie.

  34. Case Study

  35. None

  36. <head> <script>...</script> <meta name="viewport" content="width=device-width, minimum-scale=1.0" /> <link rel="stylesheet" href="/app.css"

    /> <script src="/app.js"></script> <meta charset="UTF-8" /> <title>...</title> </head>

  37. <head> <meta charset="UTF-8" /> <meta name="viewport" content="width=device-width, minimum-scale=1.0" /> <title>...</title>

    <script src="/app.js"></script> <script>...</script> <link rel="stylesheet" href="/app.css" /> </head>

  38. It got slower.

  39. Move Slow to Move Fast

  40. There is probably no rush.

  41. Take your time to measure.

  42. Release incrementally.

  43. One thematic release per day.

  44. What Has Most Impact? Know where to look next time

    Know what to avoid next time

  45. Maximise the Work Not Done

  46. It’s just as important to know what not to do.

  47. Case Study

  48. None

  49. Most conversions happen at 2s but our median is 2.9s.

  50. Case Study

  51. None

  52.          

      &RPSDQ\      9LWDPL[ %UHYLOOH 1LQMD %OHQGWHF .LWFKHQ$LG 1XWUL%XOOHW /RDG7LPH 6WDUW5HQGHU /RDG7LPHDQG6WDUW5HQGHU±%HIRUH

  53. There is no prize beyond first place.

  54. Let’s Ignore Load Already the best on desktop and mobile

    It’s a legacy metric anyway We’re definitely suffering with Start Render Everyone hitting the site wants to see stuff Start Render is our focus

  55. Load is subresources; Start Render is <head>.

  56. A week in their <head>.

  57.          

      &RPSDQ\      9LWDPL[ %UHYLOOH %OHQGWHF 1LQMD 1XWUL%XOOHW .LWFKHQ$LG /RDG7LPH 6WDUW5HQGHU /RDG7LPHDQG6WDUW5HQGHU±$IWHU
  58. None

  59. 4EKI 8MQI W     ,SQI 'EXIKSV] 4VSHYGX

    4(4 764 0SEH8MQI 88*& 7XEVX6IRHIV *MVWX'SRXIRXJYP4EMRX 7TIIH-RHI\ 0EWX4EMRXIH,IVS *MVWX'49-HPI %TTPI1MPIWXSRI8MQMRKWƁ1SFMPI

  60. Pricing Performance

  61. Don’t do it for the money, but never do it

    for no money. Oliver Reichenstein “

  62. Performance directly affects the bottom line.

  63. This is not a good time to be shy.

  64. Case Study

  65. None

  66. 03.s average [reduction in] latency across our funnel equals £8m

    a year in additional revenue. csswz.it/2FQHt2S “

  67. Value-Based Pricing.

  68. My fee represents my contribution to the project with a

    dramatic ROI for you and equitable compensation for me. Alan Weiss “

  69. How much are we likely to make from this project?

    “

  70. How have you calculated the value of this project?

  71. I am not your Business Analyst.

  72. But I can be for a little bit.

  73. Phase Zero.

  74. Phase Zero A period of pre-engagement Let’s work out the

    value of the project together Then we know how much it’s worth to you… …and I know what to charge you.
  75. None

  76. Customer Data and A/B Tests.

  77. 3× more conversions if we start rendering under 3s compared

    to after 6s
  78. None

  79. 62.1% 87.5% Bounce Rate With and Without Cloud.typography

  80. Now we can put numbers against it.

  81. If we can achieve a start render of 0.9s, we

    stand to increase conversions by 8%. Across one year, this equates to increased revenues of £1–1.2m. I’ll take X%. “

  82. Keep on Top of Things

  83. Performance budgets, pragmatically.

  84. How should we set performance budgets?

  85. Set to your worst point in the last two weeks.

  86. 0.8s under budget: new budget set to 2.2s

  87. Only just staying in budget: leave as-is but don’t slip

    further

  88. Yikes! Lots of work to do here: budget is overshot

    daily

  89. Normalise performance.

  90. Normalise Performance Talk about it like it’s always been there

    Mention it in passing Have dashboards for individual teams
  91. None

  92. Ultimately blur the lines.

  93. Performance is a proxy for business success.

  94. Know When to Stop

  95. Good enough is good enough.

  96. Thank You Very Much

  97. Thank You Understand the situation fully before you begin Maximise

    the work not done Calculate the value of the project Know when to stop

  98. harry.is/for-hire