Table of contents

Measuring Page Performance Using Playwright - Best Practices

The need for fast and responsive applications has never been greater because of the move from desktop to mobile. Still, web applications have been increasing in complexity and size. It is clear why the topic of webpage performance is more popular today than it ever was.

This article gives a practical introduction to the whys and hows of web performance without getting lost in the depth of this massive topic.

The time it takes for a service to become usable influences a user’s perception. Helpful features, great design and other prominent characteristics become irrelevant when an online service is slow, and users navigate away.

You can build the best web application in the world, but be mindful that each user has limited time to invest in your service to solve their problems. Exceed that amount, and you risk losing them to a different, more performant solution. Especially for new users, a fast experience is essential because they haven’t been given proof of the quality of your service yet.

There is a brighter side to the topic: if low performance can sink an online platform, high performance can very well help it rise to the top. Speed and responsiveness can be a service differentiator, prompting users to choose it over the competition. Therefore an investment in this area will almost always pay off. Some notorious real-world examples from known businesses include:

1. Pinterest decreased user wait time, and increased both traffic and conversions.
2. Zalando applied small load time improvements and found a direct correlation with increased revenue per session.
3. The BBC discovered that every extra page load second led to 10% of users leaving the page.

Given the importance of page performance, it is no coincidence that browsers expose a ton of insights into performance metrics. Knowing how your application scores against these across time will provide the feedback you need to keep it performant for your users.

Several approaches can be combined to gather the best insights:

1. Real user monitoring to understand what performance actual end-users of your service are experiencing.
2. Synthetic monitoring to proactively gather intel on service performance and find issues before users stumble into them.
3. Performance testing to avoid releasing performance regression to production in the first place.
4. Regular audits to get an overview of your page’s performance and suggestions on how to improve it, e.g. with tools such as Google Lighthouse.

With Google pushing for a faster web, the Web Vitals metrics should be on your radar. Metrics such as Time to First Byte (TTFB), Total Blocking Time (TBT) or First Contentful Paint (FCP) are good user experience indicators and worth monitoring.

Google recommends focusing on the three most important ones – Largest Contentful Paint (LCP), First Input Delay (FID) and Cumulative Layout Shift (CLS). These three metrics are considered the Core Web Vitals and give a good idea of a page’s loading behavior, interactivity, and visual stability.

As much as we should be striving to build performant applications, we should commit to monitoring and testing performance to enable continuous feedback and rapid intervention in case of degradation. Playwright provides a great toolkit to power synthetic monitoring and performance testing.

1. Access to the Web Performance APIs.
2. Whenever testing against Chromium, access to the Chrome DevTools Protocol for traffic inspection, network emulation and more.
3. Easy interoperability with performance libraries from the Node.js ecosystem.

Modern browsers support many APIs to gather web performance metrics and web vitals.

The Navigation Timing and the Resource Timing performance APIs are W3C specifications. The MDN docs define the scope of both:

Navigation timings are metrics measuring a browser’s document navigation events. Resource timings are detailed network timing measurements regarding the loading of an application’s resources. Both provide the same read-only properties, but navigation timing measures the main document’s timings whereas the resource timing provides the times for all the assets or resources called in by that main document and the resources’ requested resources.

The Navigation Timing API allows us to retrieve timestamps of key events in the page load timeline. A Navigation Timing entry includes metrics such as the navigation response time, the used protocol and document load time.

const { chromium } = require('playwright')

;(async () => {
  const browser = await chromium.launch()
  const page = await browser.newPage()
  await page.goto('https://danube-web.shop/')

  const navigationTimingJson = await page.evaluate(() =>
    JSON.stringify(performance.getEntriesByType('navigation'))
  )
  const navigationTiming = JSON.parse(navigationTimingJson)

  console.log(navigationTiming)

  await browser.close()
})()

Run in Checkly

[{
  name: 'https://danube-web.shop/',
  entryType: 'navigation',
  startTime: 0,
  duration: 1243.7999999998137,
  initiatorType: 'navigation',
  nextHopProtocol: 'http/1.1',
  workerStart: 0,
  redirectStart: 0,
  redirectEnd: 0,
  fetchStart: 0.10000000009313226,
  domainLookupStart: 1.2000000001862645,
  domainLookupEnd: 11.100000000093132,
  connectStart: 11.100000000093132,
  connectEnd: 336.8000000002794,
  secureConnectionStart: 102.89999999990687,
  requestStart: 336.89999999990687,
  responseStart: 432.39999999990687,
  responseEnd: 433.70000000018626,
  transferSize: 971,
  encodedBodySize: 671,
  decodedBodySize: 671,
  serverTiming: [],
  workerTiming: [],
  unloadEventStart: 0,
  unloadEventEnd: 0,
  domInteractive: 1128.8999999999069,
  domContentLoadedEventStart: 1128.8999999999069,
  domContentLoadedEventEnd: 1130.8999999999069,
  domComplete: 1235.3999999999069,
  loadEventStart: 1235.3999999999069,
  loadEventEnd: 1235.3999999999069,
  type: 'navigate',
  redirectCount: 0
}]

The Resource Timing API allows us to zoom in on single resources and get accurate information about how quickly they loaded. For example, we could specifically look at our website’s logo:

const { chromium } = require('playwright')

;(async () => {
  const browser = await chromium.launch()
  const page = await browser.newPage()
  await page.goto('https://danube-web.shop/')

  const resourceTimingJson = await page.evaluate(() =>
    JSON.stringify(window.performance.getEntriesByType('resource'))
  )

  const resourceTiming = JSON.parse(resourceTimingJson)
  const logoResourceTiming = resourceTiming.find((element) =>
    element.name.includes('.svg')
  )

  console.log(logoResourceTiming)

  await browser.close()
})()

Run in Checkly

{
  name: 'https://danube-web.shop/static/logo-horizontal.svg',
  entryType: 'resource',
  startTime: 1149.1000000000931,
  duration: 96.89999999990687,
  initiatorType: 'img',
  nextHopProtocol: 'http/1.1',
  workerStart: 0,
  redirectStart: 0,
  redirectEnd: 0,
  fetchStart: 1149.1000000000931,
  domainLookupStart: 1149.1000000000931,
  domainLookupEnd: 1149.1000000000931,
  connectStart: 1149.1000000000931,
  connectEnd: 1149.1000000000931,
  secureConnectionStart: 1149.1000000000931,
  requestStart: 1149.6000000000931,
  responseStart: 1244.3000000002794,
  responseEnd: 1246,
  transferSize: 21049,
  encodedBodySize: 20749,
  decodedBodySize: 20749,
  serverTiming: [],
  workerTiming: []
}

The Paint Timing API provides information on the first paint and the first contentful paint. Access the entries via performance.getEntriesByType('paint') or performance.getEntriesByName('first-contentful-paint').

const { chromium } = require('playwright')

;(async () => {
  const browser = await chromium.launch()
  const page = await browser.newPage()
  await page.goto('https://danube-web.shop/')

  const paintTimingJson = await page.evaluate(() =>
    JSON.stringify(window.performance.getEntriesByType('paint'))
  )
  const paintTiming = JSON.parse(paintTimingJson)

  console.log(paintTiming)

  await browser.close()
})()

Run in Checkly

[
  { name: 'first-paint', entryType: 'paint', startTime: 1149.5, duration: 0 },
  { name: 'first-contentful-paint', entryType: 'paint', startTime: 1149.5, duration: 0 }
]

The Largest Contentful Paint API provides information on all large paints. Use this API to evaluate the Core Web Vital Largest Contentful Paint (LCP).

const { chromium } = require('playwright')

;(async () => {
  const browser = await chromium.launch()
  const page = await browser.newPage()
  await page.goto('https://danube-web.shop/')

  const largestContentfulPaint = await page.evaluate(() => {
    return new Promise((resolve) => {
      new PerformanceObserver((l) => {
        const entries = l.getEntries()
        // the last entry is the largest contentful paint
        const largestPaintEntry = entries.at(-1)
        resolve(largestPaintEntry.startTime)
      }).observe({
        type: 'largest-contentful-paint',
        buffered: true
      })
    })
  })

  console.log(parseFloat(largestContentfulPaint)) // 1139.39

  await browser.close()
})()

Run in Checkly

The Layout Instability API provides information on all layout shifts. Use this API to evaluate the Core Web Vital Cumulative Layout Shift (CLS).

const { chromium } = require('playwright')

;(async () => {
  const browser = await chromium.launch()
  const page = await browser.newPage()
  await page.goto('https://danube-web.shop/')

  const cummulativeLayoutShift = await page.evaluate(() => {
    return new Promise((resolve) => {
      let CLS = 0

      new PerformanceObserver((l) => {
        const entries = l.getEntries()

        entries.forEach(entry => {
          if (!entry.hadRecentInput) {
            CLS += entry.value
          }
        })

        resolve(CLS)
      }).observe({
        type: 'layout-shift',
        buffered: true
      })
    })
  })

  console.log(parseFloat(cummulativeLayoutShift)) // 0.0001672498

  await browser.close()
})()

Run in Checkly

The Long Task API provides information about all JavaScript executions taking 50 milliseconds or more. Use this API to evaluate the Web Vital and lab metric Total Blocking Time (TBT).

const { chromium } = require('playwright')

;(async () => {
  const browser = await chromium.launch()
  const page = await browser.newPage()
  await page.goto('https://danube-web.shop/')

  const totalBlockingTime = await page.evaluate(() => {
    return new Promise((resolve) => {
      let totalBlockingTime = 0
      new PerformanceObserver(function (list) {
        const perfEntries = list.getEntries()
        for (const perfEntry of perfEntries) {
          totalBlockingTime += perfEntry.duration - 50
        }
        resolve(totalBlockingTime)
      }).observe({ type: 'longtask', buffered: true })

      // Resolve promise if there haven't been long tasks
      setTimeout(() => resolve(totalBlockingTime), 5000)
    })
  })

  console.log(parseFloat(totalBlockingTime)) // 0

  await browser.close()
})()

Run in Checkly

If the browser performance APIs are not enough, the Chrome DevTools Protocol offers many great performance tools for us to leverage with Playwright.

One important example is network throttling, through which we can simulate the experience of users accessing our page with different network conditions.

const { chromium } = require('playwright')

;(async () => {
  const browser = await chromium.launch()
  const page = await browser.newPage()

  const client = await page.context().newCDPSession(page)
  await client.send('Network.enable')
  await client.send('Network.emulateNetworkConditions', {
    offline: false,
    downloadThroughput: (4 * 1024 * 1024) / 8,
    uploadThroughput: (3 * 1024 * 1024) / 8,
    latency: 20
  })

  await page.goto('https://danube-web.shop/')

  // your flow and assertions

  await browser.close()
})()

Run in Checkly

The DevTools Protocol is quite extensive. We recommend exploring the documentation and getting a comprehensive overview of its capabilities.

Lighthouse can easily be used programmatically with Playwright to gather values and scores for different metrics, like Time To Interactive (TTI):

const chromeLauncher = require('chrome-launcher')
const { chromium } = require('playwright')
const lighthouse = require('lighthouse')
const request = require('request')
const util = require('util')

;(async () => {
  const chrome = await chromeLauncher.launch()

  const resp = await util.promisify(request)(
    `http://localhost:${chrome.port}/json/version`
  )
  const { webSocketDebuggerUrl } = JSON.parse(resp.body)
  const browser = await chromium.connect({ wsEndpoint: webSocketDebuggerUrl })

  const { lhr } = await lighthouse(
    'https://danube-web.shop/',
    { port: chrome.port },
    null
  )

  console.log('Report complete for', lhr.finalUrl)

  console.log(`
    Time To Interactive - Score: ${lhr.audits.interactive.score},
    Value: ${lhr.audits.interactive.numericValue} ${lhr.audits.interactive.numericUnit}
  `)

  await browser.close()
  await chrome.kill()
})()

All above examples can be run as follows:

$ node measure-performance.js

1. The comprehensive MDN Web Performance documentation
2. Automatically finding accessibility issues with Playwright
3. web.dev’s performance section
4. Getting started with Chrome DevTools Protocol by Andrey Lushnikov
5. Get Started with Google Lighthouse