A deep dive into Google’s Core Web Vitals Because sometimes fast isn’t fast enough! Bastian Grimm, Peak Ace AG | @basgr

I have a confession to make

For me, there is nothing worse than having to wait for anything to load! I'm really impatient

GDPR is actually pretty cool!

pa.ag @peakaceag 5 USA Today created a super fast GDPR-compliant domain 500 vs 34 requests, 140 vs 0 JS files, 6 vs 1 CSS, 5.01 MB vs 356 KB in size, etc. EU US Start Render 0.300 sec 1.700 sec First Interactive 0.345 sec 3.604 sec Load Time 0.995 sec 19.261 sec Speed Index 443 8,792 Total Requests 34 859 Bytes in 356 KB 5,092 KB

pa.ag @peakaceag 6 Let’s recap your users’ expectations: Slow page loading time is a major factor in page abandonment. According to a Nielsen report, 47% of people expect a website to load within two seconds, and 40% will leave a website if it does not load fully within three seconds.

Fast loading time plays an important role in overall user experience! Performance is about user experience!

pa.ag @peakaceag 8 Revisited: page speed already is a ranking factor Source: http://pa.ag/2iAmA4Y | http://pa.ag/2ERTPYY

pa.ag @peakaceag 9 What most people tend to forget… Performance is a relative concept! Source: https://pa.ag/38jyW6a A site might be fast for one user (on a fast network with a powerful device) but slow for another user (on a slow network with a low-end device). Two sites may finish loading at the exact same time, yet one may seem to load faster (if it loads content progressively rather than waiting until the end). A site might appear to load quickly but then respond slowly (or not at all) to user interaction.

pa.ag @peakaceag 10 Google’s 2020 Core Web Vitals set, at a glance The current set focuses on three aspects of user experience - loading, interactivity, and visual stability - and includes the following metrics (and their respective thresholds): Source: https://pa.ag/3irantb LCP measures loading performance. To provide a good UX, LCP should occur within 2.5 seconds. FID measures interactivity. To provide a good UX, pages should have an FID under 100 milliseconds. CLS measures visual stability. To provide a good UX, pages should maintain a CLS of less than 0.1. i

And how do I measure them accurately? Where do I find these numbers?

pa.ag @peakaceag 12 Performance measurement scoring & metrics, at a glance But what do these numbers actually mean? And how are they different from what we have used before? Source: https://pa.ag/38jyW6a Historically, web performance has been measured with the load event. However, even though load is a well-defined moment […], that moment doesn't necessarily correspond with anything the user cares about.

pa.ag @peakaceag 13 If you’re not familiar with Google Lighthouse… There are several ways to install Google Lighthouse and run it to audit your webpages: Google Chrome Google Chrome Extension Node Command Line Interface (CLI) ▪ Navigate to the webpage you wish to run the audit on. ▪ Right-click on the page and click “inspect element” or press CTRL + SHIFT + I (for dev. tools) ▪ Navigating to the audits tab, find the Lighthouse logo and CTA which says “Perform an audit…” ▪ Simply click on the audit button and it should show you 4 options to run; press “Run Audit”. ▪ Download the extension (https://pa.ag/3cAXxqg) and add it to your Google Chrome browser. ▪ Navigate to the webpage you want to audit. ▪ Hit the Lighthouse Chrome extension icon and let Lighthouse run. ▪ Download and install Google Chrome ▪ Install a current and stable version of Node (>=6) ▪ Run the following command line to install the global Lighthouse npm package: npm install -g lighthouse ▪ Run Lighthouse by typing the following into the CLI: lighthouse https://bastiangrimm.com/

pa.ag @peakaceag 14 To start, visit any webpage you want to test in Chrome Press “CTRL+SHIFT+I“, find the “Lighthouse“ tab, adjust settings & click “Generate report“ #1 #2 #3 #4 Make sure to always run Lighthouse in Chrome Incognito mode! #5 #6

pa.ag @peakaceag 15 Overall scores (range: 0 – 100) depending on which categories you selected Detailed performance measurement breakdown using the most common metrics, e.g. FCP/FMP Film strip view with various browser paint timings Import/export as well as various “pretty print” and JSON data formats

pa.ag @peakaceag 16

pa.ag @peakaceag 17 Zoom in for detailed info for each opportunity

pa.ag @peakaceag 18 Important: metrics are measured in one of two ways: To find yours: Chrome > dev tools > performance > timings In the lab: using tools to simulate a page load in a consistent, controlled environment In the field: on real users actually loading and interacting with the page”

pa.ag @peakaceag 19 Track paint times using PerformanceObserver Or even better: use the official web vitals JS library Get the web-vitals JS library: https://pa.ag/2VlVKy6

pa.ag @peakaceag 20 This is what it looks like in Google Analytics Behaviour > Events > Pages: performance metrics [first-contentful-paint] LCP , FID as well as CLS available in Google Analytics:

Enough theory – let’s get into it, shall we?

Before we start optimising, here is a brief overview of which metrics are influenced by which components: What impacts what?

pa.ag @peakaceag 23 Optimising for Core Web Vitals such as LCP, FID and CLS? An overview of the most common issues and respective fixes: LCP is primarily affected by: ▪ Slow server response time ▪ Render blocking JS/CSS ▪ Resource load times ▪ Client-side rendering FID is primarily affected by: ▪ Third-party code ▪ JS execution time ▪ Main thread work/business ▪ Request count & transfer size CLS is primarily affected by: ▪ Images without dimensions ▪ Ads, embeds and iframes without dimensions ▪ Web fonts (FOIT/FOUT) Optimizing for LCP: ▪ Server response times & routing ▪ CDNs, caching & compression ▪ Optimise critical rendering path ▪ Reduce blocking times (CSS, JS, fonts) ▪ Images (format, compression, etc.) ▪ Preloading & pre-rendering ▪ Instant loading based on PRPL Optimising for FID: ▪ Reduce JS execution (defer/async) ▪ Code-split large JS bundles ▪ Break up JS long tasks (>50ms) ▪ Minimise unused polyfills ▪ Use web workers to run JS on a non-critical background thread Optimising for CLS: ▪ Always include size attributes on images, video, iframes, etc. ▪ Reserve required spaces in advance ▪ Reduce dynamic injections

The code and resources required to render the initial view of a web page Critical rendering path

pa.ag @peakaceag 25 Critical rendering path optimisation Initial view (critical) Below the fold (not critical)

A brief, technical overview:

pa.ag @peakaceag 27 CSSOM: the CSS Object Model ▪ The CSSOM is a “map” of the CSS styles found on a web page. ▪ It’s much like the DOM (Document Object Model), but for CSS rather than HTML. ▪ The CSSOM combined with the DOM is used by browsers to display web pages. body font-size:18px; h1 font-size:22px; a font-size:12px; div font-size:16px; p font-size:12px; p font-size:16px;

pa.ag @peakaceag 28 Web browsers use the CSSOM to render a page If this is external CSS, the browser needs to wait for the download.

pa.ag @peakaceag 29 Google doesn’t make a single CSS GET request! Because requesting external CSS is more expensive than inlining everything.

pa.ag @peakaceag 30 How to know which CSS is critically required “Critical” renders in multiple resolutions and builds a combined/compressed CRP CSS: Critical & criticalCSS on GitHub: http://pa.ag/2wJTZAu & http://pa.ag/2wT1ST9 ▪ Minimum: a snapshot of CSS rules to render a default desktop resolution (e.g. 1280x1024). ▪ Better: various snapshots for mobile phones, pad/s & desktop/s – manually, that’d be a lot of work!

pa.ag @peakaceag 31 Putting it all together Fit the HTML, CSS & JS that are necessary for “Start Render” into the first 14kB round trip! CRP loading demo h1 { colour: green; } *! loadCSS. [c]2017 Filament Group, Inc. MIT License */ (function(){ ... } ()); /*! loadCSS rel=preload polyfill. [c] 2017 Filament Group, Inc. MIT License */ (function(){ ... } ()); h1 { colour: green; } *! loadCSS. [c]2017 Filament Group, Inc. MIT License */ (function(){ ... } ()); /*! loadCSS rel=preload polyfill. [c] 2017 Filament Group, Inc. MIT License */ (function(){ ... } ()); Inline your critical CSS. 1 Loading non-critical CSS async using rel=“preload“. 2 Apply the CSS once it has finished loading via “onload“. 3 Fallback for non-JS requests. 4 Implement loadCSS script for older browsers. 5

pa.ag @peakaceag 32 Almost brand new: content-visibility, a new CSS property content-visibility enables the user agent to skip an element's rendering work, including layout & painting, until it is needed – and therefore makes the initial load much faster! Source: http://pa.ag/2Wxn399

Higher quality, (more) efficient data compression, smaller files as well as loading strategies for faster image delivery Image optimisation

pa.ag @peakaceag 34 What are you looking for in your Lighthouse report? Opportunities > Properly size images

pa.ag @peakaceag 35 60% of all web traffic is made up of images... The average website transfers between 900 and 970KB of images per URL! Source: https://pa.ag/2xwHOFN

pa.ag @peakaceag 36 Basic optimisation for all images: put ‘em on a diet! tinyPNG & tinyJPG for smart (lossy) compression & removal of metadata et al. API access, various plug-ins (WP , etc.) as well as direct integration into Photoshop. Source: http://tinypng.com | http://tinyjpg.com

imagemin is your friend: "lossy" or "lossless" compression for JPEG, PNG, GIF, SVG & WebP DIY image compression

pa.ag @peakaceag 38 WebP: Google’s alternative to JPEG, PNG, and GIF Lossy & lossless compression, transparency, metadata, colour profiles, animation, and much smaller files (30% vs. JPEG, 80% vs. PNG) – but only in Chrome, Opera & Android. Everything about WebP: http://pa.ag/1EpFWeN / & WebP support: http://pa.ag/2FZK4XS

pa.ag @peakaceag 39 You can use WebP with an on-the-fly replacement Swap PNG and JPEG images per re-write (i.e. using nginx/Apache configuration). Alternatively: the element allows you to manually specify multiple file types. VS.

pa.ag @peakaceag 40 bestbuy.com could save >50% in image size/traffic Better compression combined with modern, responsive images & formats

pa.ag @peakaceag 41 Say hello to AVIF – supported in Chrome 85 & Firefox 80 Developed by the Alliance for Open Media in collaboration with Google, Cisco, and Xiph.org to be an open-sourced and royalty-free image format: Source: https://pa.ag/3gK9Gdk AV1 (.avif) is basically a super-compressed image type. Netflix has already considered .avif superior to the JPEG, PNG, and even the newer WebP image formats for its image quality to compressed file size ratio. i

pa.ag @peakaceag 42 Also, ensure to generate responsive images! One image for all screen resolutions and devices is not enough. An image per pixel is too much; responsivebreakpoints.com can help! More: https://pa.ag/2NNBvVm & https://pa.ag/2C6t6aQ

pa.ag @peakaceag 43 What are you looking for in your Lighthouse report? Opportunities > Defer offscreen images

pa.ag @peakaceag 44 Off-load your images using the “lazySizes” loader The high-performance lazy loader for images, iFrames and more, detects any visibility changes (e.g. through user interaction, CSS or JS) without configuration. More on GitHub: https://pa.ag/2VOywil Especially important for mobile since you only want to load images that are actually visible! You can even lazy load responsive images (with automatic sizes calculation)

pa.ag @peakaceag 45 Things are getting much easier now: loading = lazy Performance benefits paired with SEO friendliness (and no JS) simultaneously Tip: This now also works for :

Remember: to prevent any CLS, it‘s of utmost importance to specify width and height attributes, e.g. for images, iframes, videos, etc. Always specify dimensions!

Pretty, varied, colourful and if not implemented correctly – incredibly slow! Custom web fonts

pa.ag @peakaceag 48 What are you looking for in your Lighthouse report? Diagnostics > Ensure text remains visible during webfont load

pa.ag @peakaceag 49 >70% of all websites use at least one non-standard font Result: 111KB of additional data and on average, 4-5 additional HTTP requests Source: https://pa.ag/2I7vAHC

pa.ag @peakaceag 50 The traditional method: using external CSS Easy to use, but with one big disadvantage: it’s render-blocking! CSS’s (font) call to Google causes the render to stop/block until the download has finished!

FOIT (flash of invisible text) or FOUT (flash of unstyled text) can cause irritating flickering Asynchronous?

pa.ag @peakaceag 52 Fighting the flash of unstyled text/content Make your fall-back font match the intended web font (letter spacing, heights, etc.), otherwise this will cause layout shifts: Give it a try: https://pa.ag/2qgE8EH

pa.ag @peakaceag 53 Fighting the flash of invisible text/content Stuff to play around with: various “font-display” strategies for CSS More: http://pa.ag/2eUwVob ‘font-display’ enables the text to be displayed while the font itself is still loading.”

pa.ag @peakaceag 54 Non-blocking CSS loading means massive gains Optional results in a 100ms blocking period, but no swap - even after it’s downloaded (only on “next page” view). This feels much faster! Go to your CSS file, look for @font-face and add ’font-display:optional’ – there hasn’t been a safer & easier gain in #webperf in a long time! Invisible Fallback Webfont 3s 0s 100ms 3s 100ms Block Swap Fallback Optional

pa.ag @peakaceag 55 Btw: if you‘re using Google Fonts, you’ll love this one: Adding “display=swap“ to the URL will achieve the same result! However, I’d rather not rely on (external) web fonts at all. Source: https://pa.ag/2BbLK03 https://fonts.googleapis.com/css?family=Lato https://fonts.googleapis.com/css?family=Lato&display=swap

Because latency and asset delivery matter, especially for international sites TTFB/CDNs, pre-* & more

pa.ag @peakaceag 57 What are you looking for in your Lighthouse report? Opportunities > Reduce server response times (TTFB)

pa.ag @peakaceag 58 Should I worry about my TTFB at all? Spoiler: YES! And what‘s an acceptable result to aim for? More: http://pa.ag/2lKCIRH & http://pa.ag/2mkJTMY Many possible causes of slow server responses, and therefore many possible ways to improve: ▪ Optimise the server's application logic to prepare pages faster. ▪ Optimise how your server queries databases (or migrate to faster database systems). ▪ Upgrade your server hardware to have more memory or CPU.

pa.ag @peakaceag 59 What are you looking for in your Lighthouse report? Opportunities > Reduce the impact of third-party code

pa.ag @peakaceag 60 Especially for global businesses, CDNs can be a great help Use CDNPerf.com to find the one that suits you best, depending on where you are and which regions/countries you‘re predominantly serving. This will positively impact TTFB! Give it a try: https://www.cdnperf.com/ VS

pa.ag @peakaceag 61 CDN recommendation? Check out Cloudflare! Cloudflare Workers and their approach to serverless JavaScript is super powerful:

pa.ag @peakaceag 62 What are you looking for in your Lighthouse report? Opportunities > Preload key requests

pa.ag @peakaceag 63 Breakdown of requests for Netdoktor.de (waterfall view) When you‘re using a CDN, or getting resources from other, external (sub-) domains, make sure to pre-* respectively: DNS lookup for the asset server (static.netdoktor.de) takes ~300 ms

pa.ag @peakaceag 64 DNS pre-fetching in : 81ms = 75% time saved Very useful for other hosts' resources that you want to use at a later stage.

pa.ag @peakaceag 65 Taking it one step further: pre-connecting HTTPS Don't just pre-resolve DNS names – also allow for TLS-handshake.

pa.ag @peakaceag 66 Or: rel=preload late-discovered hero images faster You can use to optimize Core Web Vitals; specifically, how soon the primary imagery visible in the viewport loads, which positively impacts LCP: Source: https://pa.ag/31DGPmz Preload can substantially improve LCP, especially if you need critical images (like hero images) to be prioritized over the loading of other images on a page. While browsers will try their best to prioritize the loading of images in the visible viewport, can offer a significant boost in priority.

pa.ag @peakaceag 67 This has also just recently landed in Lighthouse: To optimise Largest Contentful Paint, you should preload your critical images. Lighthouse 6.5 has suggested how and where to do this Source: https://pa.ag/3mBkqOi

Create awareness (and limitations) beyond just one team; and make them all commit to hard numbers! Performance budgets

pa.ag @peakaceag 69 Google introduced performance budgets in Lighthouse LightWallet for budgeting: (e.g. page weight, request and timing-based budgets) Source: http://pa.ag/2WxNjQK "budgets": [ { "cpuThrottling": 1, "connectionType": "wifi", "pageWeight": [ { "resourceType": "script", "budget": 300 }, { "resourceType": "stylesheet", "budget": 50 } ], "timings": [ { "metric": "timeToInteractive", "budget": 5000, "tolerance": 1000 }, { "metric": "firstMeaningfulPaint", "budget": 2000, "tolerance": 500 } ] } ]

pa.ag @peakaceag 70 Don‘t know where to start? Use the performance budget calculator to synchronise your spending with your goals: More: http://bit.ly/perf-budget-calculator

Test on real phones with real network settings! Embrace performance budgets and learn to live within them

Besides “regular Lighthouse”, here are some tool recommendations to make your life a bit easier Tool recommendations

pa.ag @peakaceag 73 The labels “Poor”, “Needs improvement”, and “Good” are applied to a URL on specific device type. If a URL is below the threshold of data for a given metric, that metric is omitted from the report. In case you didn’t notice: things are much better in GSC! Search Console now contains a “Core Web Vitals“ report for desktop and mobile. This is current, real-world data based on the Chrome UX report: More: https://pa.ag/3eKHpEe i

pa.ag @peakaceag 74 Chrome Web Vitals extension Preview: comparing local metrics to field data from CrUX Get the extension: http://pa.ag/3mG1OMW | Preview: https://pa.ag/2KmQrfQ

pa.ag @peakaceag 75 Another Chrome extension: Core SERP Vitals Show real world Core Web Vitals data from the CrUX report in Google search results: Source: http://pa.ag/34w4fvv

pa.ag @peakaceag 76 Run multiple Lighthouse reports at once, for free! Visit batchspeed.com and throw in the URLs you want to measure with Lighthouse More: batchspeed.com

pa.ag @peakaceag 77

pa.ag @peakaceag 78 Lighthouse Parade: let‘s scale things up! A Node.js command line tool that crawls a domain and compiles a report with Lighthouse performance data for every page: Give it a try: http://pa.ag/2WAAiWu With a single command, the tool will crawl an entire site, run a Lighthouse report for each page, and then output a spreadsheet with the aggregated data. […] Each row in the spreadsheet is a page on the site, and each individual performance metric is a column. This makes it very easy to perform high-level analysis because you can sort the rows by whichever metric you are analysing.

pa.ag @peakaceag 79 CrUX report comparison at a glance Easy comparison of desktop vs phone results – or yourself vs competition, etc. More: https://crux-compare.netlify.app/ | Source: http://pa.ag/3nE02NB Q: Is there a difference between desktop and mobile ranking? A: At this time, using page experience as a signal for ranking will apply only to mobile search.

pa.ag @peakaceag 80 Very handy: Layout Shift GIF Generator Easily identify problematic layout shifts in the viewport on mobile and desktop. Available as a simple command line tool, or as an online tool. Source: https://pa.ag/3iIJdOU

pa.ag @peakaceag 81 SpeedCurve: all you need in #webperf monitoring By far the most comprehensive toolset on the market: More: https://speedcurve.com/

… but rather an ongoing process. #webperf is not a project!

Any questions? hi@pa.ag www.pa.ag twitter.com/peakaceag facebook.com/peakaceag Take your career to the next level: jobs.pa.ag Bastian Grimm bg@pa.ag