Learn about Centmin Mod LEMP Stack today
Register Now

PageSpeed Google PageSpeed Insights v5 Update

Discussion in 'Web Development & Web Performance' started by eva2000, Nov 13, 2018.

  1. eva2000

    eva2000 Administrator Staff Member

    38,037
    8,356
    113
    May 24, 2014
    Brisbane, Australia
    Ratings:
    +12,872
    Local Time:
    6:07 PM
    Nginx 1.15.x
    MariaDB 5.5/10.x
    Google PageSpeed Insights has released version 5 update which now incorporates Google Lighthouse Report (for desktop and mobile 3G Slow throttle profile) + Chrome User Experience Reports (CrUX). Google PageSpeed Insights now changes the focus of particular metrics for the First Contentful Paint (FCP) and the First Input Delay (FID) metric data gathered from Real World Field Data (CrUX) and from performance tests from Google Lighthouse Report.
    Release Notes Release Notes | PageSpeed Insights | Google Developers

    Desktop

    psi-cmm-forum-desktop-131118-01.png

    mobile 3G Slow down throttle profile

    psi-cmm-forum-mobile-131118-01.png
     
  2. eva2000

    eva2000 Administrator Staff Member

    38,037
    8,356
    113
    May 24, 2014
    Brisbane, Australia
    Ratings:
    +12,872
    Local Time:
    6:07 PM
    Nginx 1.15.x
    MariaDB 5.5/10.x
    Companion article for new field data vs Chrome User Experience (CrUX) data and what it means at How to use Chrome User Experience Report to improve your site's performance - Search Engine Land

     
  3. eva2000

    eva2000 Administrator Staff Member

    38,037
    8,356
    113
    May 24, 2014
    Brisbane, Australia
    Ratings:
    +12,872
    Local Time:
    6:07 PM
    Nginx 1.15.x
    MariaDB 5.5/10.x
    Google PageSpeed Insights v5 doesn't include Google Lighthouse PWA score tests so doesn't take into account sites using Service Workers for PWA. But Chrome browser Audit Google Lighthouse has PWA test and also allows you to control whether or not you clear browser cache storage (for PWA/service workers) so you can test the effects of PWA on 3G slow pagespeeds.

    Since this forum has deployed and converted site to PWA via service worker pre-caching and route level caching, thought I'd test Chrome Google Lighthouse Audit with simulated 3G with storage cleared for non-PWA speeds and with storage cache intact for PWA speeds

    audit-throttling.png

    non-PWA 3G simulated

    audit-lighthouse-simulated-01.png

    versus PWA 3G simulated with storage cache intact (repeat visitor page loads)

    audit-lighthouse-simulated--dont-clear-storage-01.png

    As you can see with PWA service worker caching, 3G slow speeds for repeat visits is much faster due to the cached objects never going over the visitor's network but instead come from vistor's own browser caching.

    The PWA Scores

    audit-lighthouse-simulated--dont-clear-storage-02.png
     
  4. eva2000

    eva2000 Administrator Staff Member

    38,037
    8,356
    113
    May 24, 2014
    Brisbane, Australia
    Ratings:
    +12,872
    Local Time:
    6:07 PM
    Nginx 1.15.x
    MariaDB 5.5/10.x
    Interesting comparison, my gitools.sh script still uses old Google PageSpeed Insights v4 API for testing and sending reports to my Slack channel :)

    gitools-pagespeed-insights-v4.png

    and to compare with gitools.sh webpagetest.org with Google Lighthouse Metrics

    gitools-wpt-california-cable-01.png
     
    Last edited: Nov 14, 2018
  5. eva2000

    eva2000 Administrator Staff Member

    38,037
    8,356
    113
    May 24, 2014
    Brisbane, Australia
    Ratings:
    +12,872
    Local Time:
    6:07 PM
    Nginx 1.15.x
    MariaDB 5.5/10.x
    From testing Google PageSpeed Insights v5 so far, seems scores are tied to Google LightHouse Reports now and metrics there for mobile are tested on slow 3G mobile emulated device with 4x cpu throttle slow down so if you site has alot of javascript then scores maybe lower as the score weighting for performance category seems to be heavily weighted towards Time to Interactive and Speed Index followed by First Contentful Paint. Lower cpu powered mobile devices parse and process javascript much slower than high end cpu powered mobile and desktop device/clients so that's why you get higher scores in Google PageSpeed Insights v5 desktop tests versus mobile.

    To get a higher PageSpeed Insight v5 mobile score, you need to optimise for Time to Interactive and Speed Index followed by First Contentful Paint for 3G mobile speeds and slow Nexus5 or Moto G4 equivalent powered mobile device :)

    Latest tests for this forum :D

    psi-cmm-forum-desktop-151118-01.png psi-cmm-forum-mobile-151118-01.png
     
  6. eva2000

    eva2000 Administrator Staff Member

    38,037
    8,356
    113
    May 24, 2014
    Brisbane, Australia
    Ratings:
    +12,872
    Local Time:
    6:07 PM
    Nginx 1.15.x
    MariaDB 5.5/10.x
    Updated my gitools script with gitools_v5.sh version to support querying Google PageSpeed Insights v5 API :)

    Example at centminmod/google-insights-api-tools for mobile

    Code (Text):
    ./gitools_v5.sh mobile https://community.centminmod.com
    
    --------------------------------------------------------------------------------
    curl -4s https://www.googleapis.com/pagespeedonline/v5/runPagespeed?url=https%3A%2F%2Fcommunity.centminmod.com&strategy=mobile&key=YOUR_GOOGLE_API_KEY
    mobile CrUX Rating: SLOW
    Test url: https://community.centminmod.com
    FCP: 3046ms (SLOW) FID: 184ms (AVERAGE)
    30.10% pages fast FCP (<1000ms)
    50.00% pages average FCP (<2500ms)
    19.90% pages slow FCP (>2500ms)
    83.20% pages fast FID (<50ms)
    13.50% pages average FID (<250ms)
    3.40% pages slow FID (>250ms)
    
    PageSpeed Insights v5 Score: 79 (average)
    First-Contentful-Paint: 2895
    First-Meaningful-Paint: 3362
    Speed-Index: 2895
    First-CPU-Idle: 5126
    Time-to-Interactive: 5289
    Estimated-Input-Latency: 70
    
    ok
    

    and sending results to my custom Slack channel :)

    gitools_v5-mobile-slack-151118-01.png

    For desktop

    Code (Text):
    ./gitools_v5.sh desktop https://community.centminmod.com
    
    --------------------------------------------------------------------------------
    curl -4s https://www.googleapis.com/pagespeedonline/v5/runPagespeed?url=https%3A%2F%2Fcommunity.centminmod.com&strategy=desktop&key=YOUR_GOOGLE_API_KEY
    desktop CrUX Rating: SLOW
    Test url: https://community.centminmod.com
    FCP: 2998ms (SLOW) FID: 63ms (AVERAGE)
    29.60% pages fast FCP (<1000ms)
    52.50% pages average FCP (<2500ms)
    17.90% pages slow FCP (>2500ms)
    94.00% pages fast FID (<50ms)
    4.20% pages average FID (<250ms)
    1.70% pages slow FID (>250ms)
    
    PageSpeed Insights v5 Score: 99 (fast)
    First-Contentful-Paint: 660
    First-Meaningful-Paint: 680
    Speed-Index: 2778
    First-CPU-Idle: 1331
    Time-to-Interactive: 1510
    Estimated-Input-Latency: 14


    gitools_v5-desktop-slack-151118-01.png
     
  7. eva2000

    eva2000 Administrator Staff Member

    38,037
    8,356
    113
    May 24, 2014
    Brisbane, Australia
    Ratings:
    +12,872
    Local Time:
    6:07 PM
    Nginx 1.15.x
    MariaDB 5.5/10.x
    Asked a question about FCP origin metric differences in Chrome User Experience origin reports (CrUX) and got an answer on twitter due to FCP measurements changing the reported percentile values Rick Viscomi on Twitter

    v4
    v5
     
  8. eva2000

    eva2000 Administrator Staff Member

    38,037
    8,356
    113
    May 24, 2014
    Brisbane, Australia
    Ratings:
    +12,872
    Local Time:
    6:07 PM
    Nginx 1.15.x
    MariaDB 5.5/10.x
    Latest gitools_v5.sh update to add Lighthouse TTFB metric as well as javascript execution times as they are now an important factor (highest weight) for Time To Interactive for calculating PageSpeed Insight v5 Speed Score

    Code (Text):
    ./gitools_v5.sh desktop https://community.centminmod.com
    
    --------------------------------------------------------------------------------
    curl -4s https://www.googleapis.com/pagespeedonline/v5/runPagespeed?url=https%3A%2F%2Fcommunity.centminmod.com&strategy=desktop&key=YOUR_GOOGLE_API_KEY
    desktop CrUX Rating: SLOW
    Test url: https://community.centminmod.com
    FCP: 2998ms (SLOW) FID: 63ms (AVERAGE)
    29.60% pages fast FCP (<1000ms)
    52.50% pages average FCP (<2500ms)
    17.90% pages slow FCP (>2500ms)
    94.00% pages fast FID (<50ms)
    4.20% pages average FID (<250ms)
    1.70% pages slow FID (>250ms)
    
    PageSpeed Insights v5 Score: 100 (fast)
    First-Contentful-Paint: 790
    First-Meaningful-Paint: 790
    Speed-Index: 1854
    First-CPU-Idle: 1167
    Time-to-Interactive: 1202
    Estimated-Input-Latency: 13
    Time-To-First-Byte: 560 ms
    
    JavaScript-execution-time: 0.5 s
    URL  Total  Script-Evaluation  Script-Parse
    https://community.centminmod.com/js/jquery/jquery-1.11.0.min.js                       206.91  144.15  7.40
    https://community.centminmod.com/                                                     185.70  41.73   1.71
    https://pagead2.googlesyndication.com/pagead/js/r20181107/r20180604/show_ads_impl.js  89.61   62.74   19.31
    https://pagead2.googlesyndication.com/pagead/js/r20181107/r20100101/osd.js            56.60   48.29   7.88
    


    gitools_v5-desktop-slack-161118-01.png
     
  9. eva2000

    eva2000 Administrator Staff Member

    38,037
    8,356
    113
    May 24, 2014
    Brisbane, Australia
    Ratings:
    +12,872
    Local Time:
    6:07 PM
    Nginx 1.15.x
    MariaDB 5.5/10.x
    latest gitools_v5.sh update adds render blocking resource listing as it's important for Speedindex, First Contentful/Meaningful paints which in turn affects Time To Interactive metrics

    Code (Text):
    ./gitools_v5.sh mobile https://community.centminmod.com
    '
    --------------------------------------------------------------------------------
    curl -4s https://www.googleapis.com/pagespeedonline/v5/runPagespeed?url=https%3A%2F%2Fcommunity.centminmod.com&strategy=mobile&key=YOUR_GOOGLE_API_KEY
    mobile CrUX Rating: SLOW
    Test url: https://community.centminmod.com
    FCP: 3094ms (SLOW) FID: 169ms (AVERAGE)
    29.80% pages fast FCP (<1000ms)
    50.00% pages average FCP (<2500ms)
    20.20% pages slow FCP (>2500ms)
    83.80% pages fast FID (<50ms)
    13.10% pages average FID (<250ms)
    3.10% pages slow FID (>250ms)
    
    PageSpeed Insights v5 Score: 81 (average)
    Lighthouse Version: 4.0.0-alpha.0
    Total-Page-Size: 449 KB
    First-Contentful-Paint: 2505
    First-Meaningful-Paint: 3486
    Speed-Index: 3352
    First-CPU-Idle: 4778
    Time-to-Interactive: 5010
    Estimated-Input-Latency: 35
    Time-To-First-Byte: 580 ms
    
    JavaScript-execution-time: 1.5 s
    URL  Total  Script-Evaluation  Script-Parse
    https://community.centminmod.com/                                                     464.35  128.00  3.94
    https://pagead2.googlesyndication.com/pagead/js/r20181107/r20180604/show_ads_impl.js  395.86  198.99  51.20
    https://community.centminmod.com/js/jquery/jquery-1.11.0.min.js                       324.85  272.15  26.26
    https://pagead2.googlesyndication.com/pagead/js/r20181107/r20100101/osd.js            140.17  117.08  21.70
    https://pagead2.googlesyndication.com/pagead/js/adsbygoogle.js                        77.88   59.16   18.72
    https://community.centminmod.com/js/xenforo/xenforo.js?_v=874b23ae                    76.99   35.54   36.27
    
    Eliminate Render Blocking Resource Potential Savings: 180 ms
    URL  Size   Potential-Savings
    https://community.centminmod.com/styles/xenbase/font-awesome/css/font-awesome.min.css  7349  180
    
    okokok
    


    gitools_v5-mobile-slack-171118-01.png
     
  10. negative

    negative Active Member

    262
    26
    28
    Apr 11, 2015
    Ratings:
    +59
    Local Time:
    10:07 AM
    1.9.10
    10.1.11
    It looks hard to get perfect score on mobile 3g network anymore :)
     
    • Agree Agree x 1
  11. Revenge

    Revenge Active Member

    439
    92
    28
    Feb 21, 2016
    Portugal
    Ratings:
    +330
    Local Time:
    8:07 AM
    1.9.x
    10.1.x
    My result with IPS 4.3:

    [​IMG]
     
    • Informative Informative x 1
  12. eva2000

    eva2000 Administrator Staff Member

    38,037
    8,356
    113
    May 24, 2014
    Brisbane, Australia
    Ratings:
    +12,872
    Local Time:
    6:07 PM
    Nginx 1.15.x
    MariaDB 5.5/10.x
    yes, Google PageSpeed Insights v5 is heavily weighted towards interactive and speedindex and first contentful paint metrics on 3G moto g4/nexus 5 device emulation. So for dynamic and javascript heavy sites, 3G mobile score will be lower. From v5 API this is the weighting used to determine the score
    • interactive 5
    • speed-index 4
    • first-contentful-paint 3
    • first-cpu-idle 2
    • first-meaningful-paint 1
    If you had ads/adsense might need to do a few runs as different ads/javascript requests are loaded each run/page load.
     
  13. eva2000

    eva2000 Administrator Staff Member

    38,037
    8,356
    113
    May 24, 2014
    Brisbane, Australia
    Ratings:
    +12,872
    Local Time:
    6:07 PM
    Nginx 1.15.x
    MariaDB 5.5/10.x
    My Wordpress7 site easier to optimise for PageSpeed Insights v5 as there's no ads :)

    Code (Text):
    ./gitools_v5.sh all https://wordpress7.centminmod.com/
    
    --------------------------------------------------------------------------------
    curl -4s https://www.googleapis.com/pagespeedonline/v5/runPagespeed?url=https%3A%2F%2Fwordpress7.centminmod.com/&strategy=desktop&key=YOUR_GOOGLE_API_KEY
    
    PageSpeed Insights v5 Score: 100 (fast)
    Lighthouse Version: 4.0.0-alpha.0
    Total-Page-Size: 189 KB
    First-Contentful-Paint: 340
    First-Meaningful-Paint: 340
    Speed-Index: 1036
    First-CPU-Idle: 340
    Time-to-Interactive: 340
    Estimated-Input-Latency: 13
    Time-To-First-Byte: 390 ms
    
    JavaScript-execution-time: 0.1 s
    URL  Total  Script-Evaluation  Script-Parse
    https://wordpress7.centminmod.com/wp-content/cache/autoptimize/js/autoptimize_a18c8ac53fbbf3932ee5a971c77726fb.js  83.10  68.39  2.00
    
    Eliminate Render Blocking Resource Potential Savings: 110 ms
    URL  Size   Potential-Savings
    https://wordpress7.centminmod.com/wp-content/cache/autoptimize/css/autoptimize_01dc868fe34c451d7e463c17f57e0acb.css  28351  110
    
    
    --------------------------------------------------------------------------------
    curl -4s https://www.googleapis.com/pagespeedonline/v5/runPagespeed?url=https%3A%2F%2Fwordpress7.centminmod.com/&strategy=mobile&key=YOUR_GOOGLE_API_KEY
    
    PageSpeed Insights v5 Score: 100 (fast)
    Lighthouse Version: 4.0.0-alpha.0
    Total-Page-Size: 195 KB
    First-Contentful-Paint: 1110
    First-Meaningful-Paint: 1110
    Speed-Index: 1326
    First-CPU-Idle: 2240
    Time-to-Interactive: 2073
    Estimated-Input-Latency: 13
    Time-To-First-Byte: 90 ms
    
    JavaScript-execution-time: 0.4 s
    URL  Total  Script-Evaluation  Script-Parse
    https://wordpress7.centminmod.com/wp-content/cache/autoptimize/js/autoptimize_e8d1cf27c9a9ffd634981807e05b9aa5.js  280.58  227.70  13.83
    https://wordpress7.centminmod.com/wp-includes/js/jquery/jquery.js                                                  105.37  72.84   29.88
    
    Eliminate Render Blocking Resource Potential Savings: 330 ms
    URL  Size   Potential-Savings
    https://wordpress7.centminmod.com/wp-content/cache/autoptimize/css/autoptimize_01dc868fe34c451d7e463c17f57e0acb.css  28351  330
    


    psi-wordpress7-desktop-181118-01.png psi-wordpress7-mobile-181118-01.png
     
    Last edited: Nov 18, 2018
  14. eva2000

    eva2000 Administrator Staff Member

    38,037
    8,356
    113
    May 24, 2014
    Brisbane, Australia
    Ratings:
    +12,872
    Local Time:
    6:07 PM
    Nginx 1.15.x
    MariaDB 5.5/10.x
    re-ran my gitools_v5.sh test for Google PageSpeed Insights v5 test and compared it to previous run here. Seems Lighthouse version was updated now from Lighthouse Version: 4.0.0-alpha.0 to 4.1.0 and metrics for mobile seem to have dropped for score wise but great thing is with gitools_v5.sh tests is i can compare individual metrics easier i.e. CRuX metrics have improved for real world Chrome user experience pagespeed.

    diffcheck side by side or (left) before vs (right) now Diff Checker

    page size much larger now at 1,222KB vs before 448KB due to adsense ads hence lower mobile score due to more javascript = lower time to interactive score

    gitools_v5-compared-01.png

    Code (Text):
    ./gitools_v5.sh mobile https://community.centminmod.com
    
    --------------------------------------------------------------------------------
    curl -4s https://www.googleapis.com/pagespeedonline/v5/runPagespeed?url=https%3A%2F%2Fcommunity.centminmod.com&strategy=mobile&key=YOUR_GOOGLE_API_KEY
    mobile CrUX Rating: SLOW
    Test url: https://community.centminmod.com
    FCP: 3097ms (SLOW) FID: 65ms (AVERAGE)
    33.00% pages fast FCP (<1000ms)
    48.80% pages average FCP (<2500ms)
    18.20% pages slow FCP (>2500ms)
    94.00% pages fast FID (<50ms)
    4.90% pages average FID (<250ms)
    1.10% pages slow FID (>250ms)
    
    PageSpeed Insights v5 Score (mobile): 46 (slow)
    Lighthouse Version: 4.1.0
    Total-Page-Size: 1,222 KB
    First-Contentful-Paint: 4080
    First-Meaningful-Paint: 4155
    Speed-Index: 4454
    First-CPU-Idle: 9217
    Time-to-Interactive: 9324
    Estimated-Input-Latency: 63
    Time-To-First-Byte: 590 ms
    
    JavaScript-execution-time: 1.4 s
    URL  Total  Script-Evaluation  Script-Parse
    https://community.centminmod.com/js/jquery/jquery-1.11.0.min.js                       472.20  399.87  28.62
    https://pagead2.googlesyndication.com/pagead/js/r20190128/r20190131/show_ads_impl.js  305.39  235.74  60.42
    https://pagead2.googlesyndication.com/pagead/js/adsbygoogle.js                        256.57  217.65  30.39
    https://www.googletagservices.com/activeview/js/current/osd.js?cb=%2Fr20100101        230.45  196.22  32.72
    https://community.centminmod.com/js/xenforo/xenforo.js?_v=3d32dd93                    105.33  50.68   48.54
    
    Eliminate Render Blocking Resource Potential Savings: null
    URL  Size   Potential-Savings
    
     
    Last edited: Feb 3, 2019
  15. upgrade81

    upgrade81 Premium Member Premium Member

    188
    13
    18
    Sep 5, 2016
    Italy
    Ratings:
    +17
    Local Time:
    9:07 AM
    1.13.8
    10
    the new metrics are embarrassing.

    I have a site with mobile score 50, now it has dropped to 6.
     
  16. frm

    frm New Member

    15
    3
    3
    Dec 17, 2018
    Ratings:
    +4
    Local Time:
    5:07 PM
    1.15.7
    10.3.11
    Have you made any modifications to ngx_pagespeed conf to get a higher score with the update and with 2.1 being a stable release? I remember deferring JS broke the site, so I'm hesitant on that... but there could be other options now?
     
  17. upgrade81

    upgrade81 Premium Member Premium Member

    188
    13
    18
    Sep 5, 2016
    Italy
    Ratings:
    +17
    Local Time:
    9:07 AM
    1.13.8
    10
    Does it still make sense to use ngx_pagespeed?

    now with the Amp pages, and PWA in my opinion does not make any sense.
     
  18. frm

    frm New Member

    15
    3
    3
    Dec 17, 2018
    Ratings:
    +4
    Local Time:
    5:07 PM
    1.15.7
    10.3.11
    Why not? It's just another way of serving content to different devices. I think of it like transcoding video, some devices don't like it–this is where ngx_pagespeed fills the gap, in my opinion.

    While PWA would be best, you have to get an install. The conversion may be low for that, comparative to development costs.

    AMP pages are bland too and may turn users that like flashier things.

    I'd say use all 3 to capture the largest audience and reevaluate in 3-6 months which is used more, and, put tracking codes on ads to see which is used more and which gives you the most profit.

    Edit: If desktop is still 1/3rd your traffic, why not have it installed for them to have a better experience. Same goes for mobile, before you push the PWA install, as they're likely also going to land on a responsive layout.
     
    Last edited: Feb 12, 2019
..