How fast is your website, really?

Macklin Hartley

https://macklin.me

Spoiler Alert

 User-Centric Performance

  Capture Metrics from End User Devices

⬛  Continuous Monitoring

Not in this Talk

❌  Web Performance Improvement

Page Load event

const start = Date.now();

window.addEventListener('load', () => {
  console.log({ load: Date.now() - start });
});

Performance Inspector

Navigation Timing API

https://www.w3.org/TR/navigation-timing-2/

window.performance.timing

Adobe

Airbnb Inc

Akamai Technologies

Alibaba Group

Apple, Inc.

Baidu, Inc.

BBC

Cloudflare

Facebook

Fastly

Google LLC

Microsoft Corporation

Mozilla Foundation

Salesforce

Shopify

The New York Times

Wikimedia Foundation

W3C Web Performance Working Group

User Centric PerformancE

  • Perceived Load Speed
  • Load Responsiveness
  • Runtime Responsiveness
  • Visual stability
  • Smoothness

https://web.dev/user-centric-performance-metrics/

User-Centric PerformancE

https://web.dev/user-centric-performance-metrics/

TTFB

Time to First Byte

FCP

First Contentful Paint

LCP

Last Contentful Paint

FID

First Input Delay

INP

Interaction to Next Paint

CLS

Cumulative Layout Shift

https://developer.mozilla.org/en-US/docs/Glossary/time_to_first_byte

the time between the browser requesting a page and when it receives the first byte of information from the server.

TTFB

Time to First Byte

https://www.mindspun.com/blog/time-to-first-byte-ttfb/

TTFB

Time to First Byte

GOOD

POOR

NEEDS
IMPROVEMENT

800ms

1.8s

https://web.dev/ttfb/

TTFB

Time to First Byte

https://web.dev/fcp/

the time from when the page starts loading to when any part of the page's content is rendered on the screen...

FCP

First Contentful Paint

https://macklin.me/7-useful-tools-for-monitoring-web-performance-user-experience

FCP

First Contentful Paint

GOOD

POOR

NEEDS
IMPROVEMENT

1.8 sec

3.0 sec

https://web.dev/fcp/

FCP

First Contentful Paint

https://calibreapp.com/blog/largest-contentful-paint

...tracks how many seconds it takes for your page’s most data-intensive, above-the-fold element to load.

LCP

Largest Contentful Paint

https://macklin.me/7-useful-tools-for-monitoring-web-performance-user-experience

LCP

Largest Contentful Paint

GOOD

POOR

NEEDS
IMPROVEMENT

2.5 sec

4.0 sec

https://web.dev/lcp/

LCP

Largest Contentful Paint

https://web.dev/fid/

the time from when a user first interacts with a page[...] to the time when the browser is actually able to begin processing event handlers...

FID

First Input Delay

https://requestmetrics.com/web-performance/first-input-delay

FID

First Input Delay

GOOD

POOR

NEEDS
IMPROVEMENT

100ms

300ms

https://web.dev/fid/

FID

First Input Delay

https://web.dev/inp/

...represents a page's overall responsiveness by measuring all click, tap, and keyboard interactions made with a page.

INP

Interaction to Next Paint

https://web.dev/inp/

INP

Interaction to Next Paint

GOOD

POOR

NEEDS
IMPROVEMENT

200ms

500ms

https://web.dev/inp/

INP

Interaction to Next Paint

https://web.dev/cls/

[a calculated score] for every unexpected layout shift that occurs during the entire lifespan of a page...

CLS

Cumulative Layout Shift

https://macklin.me/7-useful-tools-for-monitoring-web-performance-user-experience

CLS

Cumulative Layout Shift

CLS

Cumulative Layout Shift

GOOD

POOR

NEEDS
IMPROVEMENT

0.1

0.25

https://web.dev/cls/

CLS

Cumulative Layout Shift

Core Web Vitals

https://developers.google.com/search/blog/2020/11/timing-for-page-experience

Interactivity
Interaction to Next Paint (INP)

Loading
Largest Contentful Paint (LCP)

Visual Stability
Cumulative Layout Shift (CLS)

Mobile Friendly

HTTPS

No Intrusive Interstitials

Search signals

for web page

experience

Core
Web
Vitals

Atomicity

TTFB ≈ 290ms

FCP ≈ 500ms

LCP ≈ 1.625 sec

https://csswizardry.com/2022/08/measure-what-you-impact-not-what-you-influence/

https://web.dev/vodafone/

Vodafone Case Study

Goals

✅  User-Centric Performance

⬛  Capture Metrics from End User Devices

⬛  Continuous Monitoring

Ad-hoc Testing

Synthetic Monitoring

Real User Monitoring (RUM)

Capture Methods

Ad-hoc Testing

Lighthouse for
Google Chrome

Ad-hoc Testing

web.dev/measure

Ad-hoc Testing

👍  Free, no setup required

👎  Once-off, manual tests only

👎  Unlikely to represent real world traffic

Synthetic Monitoring

Calibre

https://calibreapp.com/features/pull-request-reviews

Calibre

👍  CI/CD + scheduled automation

👍  Monitor change over time

👎  Not free, although competitively priced

👎  Still not representative of real world traffic

Global Mobile Market Share

https://engineering.linecorp.com/en/blog/the-baseline-for-web-development-in-2022

Global Browser Market Share

https://engineering.linecorp.com/en/blog/the-baseline-for-web-development-in-2022

Low End vs High End Performance

Low End Device
Motorola Moto E13

10x

High End Device
Apple iPhone 15 Pro

Geekbench 5 Single-core

Geekbench 5 Multi-core

The Performance INequality Gap

# of devices

performance

50th percentile
Apple iPhone 11 (2019)
Samsung Galaxy S20 Ultra (2021)

The Performance INequality Gap

# of devices

performance

75th percentile
Motorola Moto E13 (2023)
Google Pixel 4a (2020)
Apple iPhone 6 (2015)
Samsung Galaxy S6 (2015)

The Performance INequality Gap

# of devices

performance

25% of people receive
a poorer experience

4g is widely available

https://engineering.linecorp.com/en/blog/the-baseline-for-web-development-in-2022

86.8%

Network Budgets

https://www.speedtest.net/global-index/united-states

3G

4G

 ~1 second

 ~2.3 seconds

4 second budget

5G

 less than 1 second

Median Mobile Network Speed

46.26 Mb/s
Reno, NV

181.74 Mb/s
Glendale, AZ

390%

https://www.speedtest.net/global-index

Median Mobile Network Latency

https://www.speedtest.net/global-index

101 ms
Anchorage, AK

39 ms
Plano, TX

35%

Performance is Situational

Alex Russel - Progressive Performance (Chrome Dev Summit 2016)
https://www.youtube.com/watch?v=4bZvq3nodf4

Real User Monitoring (RUM)

Real User Monitoring (RUM)

Capturing performance metrics directly from devices in the real world to observe differences in functionality, reliability, and responsiveness.

Chrome User Experience (CRUX) Report

https://developers.google.com/web/tools/chrome-user-experience-report

👍  Free, publicly available data

👎  Frustrating setup process

👎  Delayed, monthly aggregation

👎  Google Chrome devices only

Chrome User Experience (CRUX) Report

Core Web Vitals Checker

https://calibreapp.com/tools/core-web-vitals-checker

https://calibreapp.com/tools/core-web-vitals-checker

👍  On-demand access

👍  Zero configuration

👎  Same upsides and downsides as CrUX Report

Core Web Vitals Checker

DDD MELBOURNE Sponsor Leaderboard

Sponsor TTFB FCP LCP ↑ FID INP CLS
🥇 Liberty 🟢 542ms 🟢 1.18s 🟢 1.53s 🟢 14ms 🟢 180ms 🟡 0.14
🥈 Umbraco 🟢 450ms 🟢 948ms 🟢 1.58s 🟢 5ms 🟢 111ms 🟢 0.02
🥉 Github 🟡 955ms 🟢 1.37s 🟢 1.69s 🟢 6ms 🟢 98ms 🟢 0.06
Luminary 🟢 789ms 🟢 1.29s 🟢 1.84s 🟢 8ms 🟢 82ms 🟢 0.03
Webjet 🟢 440ms 🟢 1.26s 🟢 1.86s 🟢 23ms 🟡 459ms 🟢 0.08
Qoria 🟢 579ms 🟢 1.48s 🟢 1.88s 🟢 5ms 🟢 44ms 🟢 0.00
Kodez 🟡 1.15s 🟢 1.69s 🟢 1.96s 🟢 49ms 🟢 0.01
DevExpress 🟢 686ms 🟢 1.62s 🟢 2.15s 🟢 3ms 🟢 47ms 🟢 0.00
Octopus Deploy 🟡 1.11s 🟡 2.43s 🟡 2.5s 🟢 5ms 🟢 62ms 🟢 0.03
SSW 🟡 1.4s 🟡 2.32s 🟡 2.77s 🟢 10ms 🟢 78ms 🟢 0.02
DDD Melbourne 🔴 2.35s 🟡 2.51s 🟡 2.94s 🟢 8ms 🟢 79ms 🟢 0.01
Kaleida 🟡 1.56s 🔴 3.35s 🟡 3.53s 🟢 0.00
Netwealth 🔴 2.63s 🔴 3.81s 🔴 4.15s 🟢 2ms 🟢 55ms 🟢 0.01
Insight 🟡 1.6s 🔴 5.18s 🔴 5.71s 🟢 8ms 🟢 128ms 🟡 0.21
JustDigitalPeople
​The Sizzle
Sponsor TTFB FCP LCP ↑ FID INP CLS
🥇 Liberty 🟢 542ms 🟢 1.18s 🟢 1.53s 🟢 14ms 🟢 180ms 🟡 0.14
🥈 Umbraco 🟢 450ms 🟢 948ms 🟢 1.58s 🟢 5ms 🟢 111ms 🟢 0.02
🥉 Github 🟡 955ms 🟢 1.37s 🟢 1.69s 🟢 6ms 🟢 98ms 🟢 0.06
Luminary 🟢 789ms 🟢 1.29s 🟢 1.84s 🟢 8ms 🟢 82ms 🟢 0.03
Webjet 🟢 440ms 🟢 1.26s 🟢 1.86s 🟢 23ms 🟡 459ms 🟢 0.08
Qoria 🟢 579ms 🟢 1.48s 🟢 1.88s 🟢 5ms 🟢 44ms 🟢 0.00
Kodez 🟡 1.15s 🟢 1.69s 🟢 1.96s 🟢 49ms 🟢 0.01
DevExpress 🟢 686ms 🟢 1.62s 🟢 2.15s 🟢 3ms 🟢 47ms 🟢 0.00
Octopus Deploy 🟡 1.11s 🟡 2.43s 🟡 2.5s 🟢 5ms 🟢 62ms 🟢 0.03
SSW 🟡 1.4s 🟡 2.32s 🟡 2.77s 🟢 10ms 🟢 78ms 🟢 0.02
DDD Melbourne 🔴 2.35s 🟡 2.51s 🟡 2.94s 🟢 8ms 🟢 79ms 🟢 0.01
Kaleida 🟡 1.56s 🔴 3.35s 🟡 3.53s 🟢 0.00
Netwealth 🔴 2.63s 🔴 3.81s 🔴 4.15s 🟢 2ms 🟢 55ms 🟢 0.01
Insight 🟡 1.6s 🔴 5.18s 🔴 5.71s 🟢 8ms 🟢 128ms 🟡 0.21
JustDigitalPeople
​The Sizzle
Sponsor TTFB FCP LCP ↑ FID INP CLS
🥇 Liberty 🟢 542ms 🟢 1.18s 🟢 1.53s 🟢 14ms 🟢 180ms 🟡 0.14
🥈 Umbraco 🟢 450ms 🟢 948ms 🟢 1.58s 🟢 5ms 🟢 111ms 🟢 0.02
🥉 Github 🟡 955ms 🟢 1.37s 🟢 1.69s 🟢 6ms 🟢 98ms 🟢 0.06
Luminary 🟢 789ms 🟢 1.29s 🟢 1.84s 🟢 8ms 🟢 82ms 🟢 0.03
Webjet 🟢 440ms 🟢 1.26s 🟢 1.86s 🟢 23ms 🟡 459ms 🟢 0.08
Qoria 🟢 579ms 🟢 1.48s 🟢 1.88s 🟢 5ms 🟢 44ms 🟢 0.00
Kodez 🟡 1.15s 🟢 1.69s 🟢 1.96s 🟢 49ms 🟢 0.01
DevExpress 🟢 686ms 🟢 1.62s 🟢 2.15s 🟢 3ms 🟢 47ms 🟢 0.00
Octopus Deploy 🟡 1.11s 🟡 2.43s 🟡 2.5s 🟢 5ms 🟢 62ms 🟢 0.03
SSW 🟡 1.4s 🟡 2.32s 🟡 2.77s 🟢 10ms 🟢 78ms 🟢 0.02
DDD Melbourne 🔴 2.35s 🟡 2.51s 🟡 2.94s 🟢 8ms 🟢 79ms 🟢 0.01
Kaleida 🟡 1.56s 🔴 3.35s 🟡 3.53s 🟢 0.00
Netwealth 🔴 2.63s 🔴 3.81s 🔴 4.15s 🟢 2ms 🟢 55ms 🟢 0.01
Insight 🟡 1.6s 🔴 5.18s 🔴 5.71s 🟢 8ms 🟢 128ms 🟡 0.21
JustDigitalPeople
​The Sizzle

Goals

✅  User-Centric Performance

✅  Capture Metrics from End User Devices

⬛  Continuous Monitoring

User Centric Performance Capture Metrics from Real Users Continuous Monitoring
window.onload()
Google Lighthouse
Calibre
CrUX Report
Web Vitals Checker
???

Continuous Monitoring

Request Metrics

https://requestmetrics.com

👍  Near real-time continuous monitoring

👍  Detailed reporting with zero sampling

👎  Can be expensive ($200/mo for 1 million sessions)

Request Metrics

Could we
Build it ourselves?

Prometheus + Grafana

Grafana

Prometheus

GET /metrics

POST /analytics

Analytics Service

web-vitals

https://github.com/GoogleChrome/web-vitals

import webVitals from 'web-vitals';

webVitals.getTTFB(sendToAnalytics);
webVitals.getFCP(sendToAnalytics);
webVitals.getLCP(sendToAnalytics);
webVitals.getFID(sendToAnalytics);
webVitals.getINP(sendToAnalytics);
webVitals.getCLS(sendToAnalytics);

export const sendToAnalytics = 
  data => navigator.sendBeacon('/analytics', JSON.stringify(data));

FrontEnd

import { Summary } from 'prom-client';

const ttfb = new Summary({ name: `ttfb`, labelNames: [`path`] });
const fcp =  new Summary({ name: `fcp`, labelNames: [`path`] });
const lcp =  new Summary({ name: `lcp`, labelNames: [`path`] });
const fid =  new Summary({ name: `fid`, labelNames: [`path`] });
const inp =  new Summary({ name: `inp`, labelNames: [`path`] });
const cls =  new Summary({ name: `cls`, labelNames: [`path`] });

BackEnd

express()
  .post(`/analytics`, (req, res) => {
    const { path, body: { metric, value } } = req;
    switch (metric) {
      case `TTFB`: ttfb.labels(path).observe(value); break;
      case `FCP`:  fcp.labels(path).observe(value);  break;
      case `LCP`:  lcp.labels(path).observe(value);  break;
      case `FID`:  fid.labels(path).observe(value);  break;
      case `INP`:  inp.labels(path).observe(value);  break;
      case `CLS`:  cls.labels(path).observe(value);  break;
    }
    res.status(200).send({});
  })
  .listen(8000);

Backend

Data Analytics

Amazon S3

Snowflake

POST /analytics

Analytics Service

👍  Near real-time continuous monitoring

👍  Free open source tooling

👎  Ongoing maintenance cost

👎  Almost certainly a bad idea

Build it yourself

Goals

✅  User-Centric Performance

✅  Capture Metrics from End User Devices

✅  Continuous Monitoring

Resources

Metrics

Tools

How Fast is your Website Really? (Melbourne)

By Macklin Hartley

How Fast is your Website Really? (Melbourne)

I presented this talk at DDD Perth 2022. Watch the talk here; https://macklin.me/talk-how-fast-is-your-website-really

  • 48