Skip to content
Tutorials Center
My learning

Topics

Lesson 37 of 37 ยท Projects and Practice

Practice: A Performance Audit Checklist

Put it all together with a step-by-step audit checklist for any website, practice scenarios to diagnose, a self-scoring tool and ideas for your next steps.

You now know the individual techniques. Real sites do not announce which one they need, so this final lesson gives you a method: a checklist you can run on any page, followed by practice scenarios where you decide the diagnosis.

The audit in order

Always work from the user's experience down to the details, and from the biggest wins to the smallest.

flowchart TB
  A["1. Real-user data<br/>which metric is bad, for whom?"] --> B["2. Lab reproduction<br/>throttled, cold cache"]
  B --> C["3. Network<br/>waterfall, size, chains"]
  C --> D["4. Rendering<br/>critical path, shifts, LCP"]
  D --> E["5. JavaScript<br/>long tasks, bundle, third parties"]
  E --> F["6. Fix one, re-measure"]

Start with the metric that is failing, then follow it down to a cause.

1. Real-user data

  • Which Core Web Vital fails at p75 (LCP, INP, CLS)?
  • Is it only mobile? One country? One template?

2. Reproduce

  • Incognito, cache disabled, CPU 4x slowdown, Slow 4G.
  • Run at least five times and record the median.

3. Network

  • Is the server response (TTFB) fast? Is the HTML cacheable and compressed?
  • Are there long request chains or late-discovered resources?
  • Are static files cached with long lifetimes and fingerprinted names?
  • Are images the right format and size, with width and height?
  • Are fonts limited to what is needed, with font-display set?

4. Rendering

  • What is the LCP element, and when is it discovered and loaded?
  • Is anything blocking first paint (CSS, synchronous scripts)?
  • What shifts? Are sizes reserved for images, ads and embeds?

5. JavaScript

  • How much JavaScript ships, and how much of it runs on this page?
  • Which long tasks exist? Which function is behind them?
  • Which third parties cost the most main-thread time?
  • Are listeners, timers and caches cleaned up?

6. Fix and verify

  • Fix the biggest cause, re-measure with the same setup, then repeat.
  • Add a budget or a CI check so it cannot regress.

Practice scenarios

Read each symptom and decide the most likely cause and first action before opening the answer. Answers are in the quiz explanations below and in this self-scoring tool, which you can run to check your thinking.

  1. Slow LCP on mobile only; the hero image is 2.4 MB.
  2. Page paints quickly, but clicking "Add to cart" takes 700 ms to respond.
  3. The article text jumps down when the page loads.
  4. The first byte takes 1.8 s even on fast Wi-Fi.
  5. A single-page app slows down after an hour of use.
  6. Lighthouse is good but real users complain.
JS Scenario trainer

Press Run (or Ctrl + Enter) to see the console output here.

No output.

A scoring tool for your own audit

Run your own site through the checklist and let this small script total the result. Change the true/false values to match your site.

JS Score your audit

Press Run (or Ctrl + Enter) to see the console output here.

No output.

Habits to keep

  • Measure first. Never optimize from a hunch.
  • Biggest win first. One 2 MB image beats ten micro-optimizations.
  • One change at a time. Otherwise you cannot tell what worked.
  • Protect your gains. Budgets, CI and alerts keep a fast site fast.
  • Think about the slowest users. Test on a mid-range phone and a slow network.

Where to go next

Check your understanding

Quick check ยท 5 questions

of 5 answered You scored / 5 ยท perfect!

  1. 1. In the audit, what should you check first?

    Starting from the failing user-facing metric points you to the right layer.
  2. 2. A page paints fast but clicking a button takes 700 ms to respond. What is the likely cause?

    Slow responses to input are an interaction problem (INP), usually caused by long main-thread tasks.
  3. 3. Text jumps down after the page loads. Which fix is most direct?

    Layout shifts come from late content without reserved space.
  4. 4. Lighthouse is green but real users are slow. What does this suggest?

    Field data reflects real conditions; lab tests are only a model.
  5. 5. Which habit best protects performance gains?

    Continuous checks catch regressions before users do.

Summary

  • Audit from user-facing metrics down to causes: data, reproduce, network, rendering, JavaScript, fix.
  • Match symptoms to causes: big image, long task, layout shift, slow server, leak, lab/field gap.
  • Measure, change one thing, re-measure and guard the result with budgets and monitoring.

Next: Web Performance course overview