How to find what slows pages down
Find what slows pages down with repeatable mobile tests. Trace delays to the server, images, scripts, or styles, then verify the fixes.
Free app. Runs on your Mac with the Claude or Codex plan you already have.
What it means to find what slows pages down
To find what slows pages down, follow a page from the first server response to the moment someone can see and use its content. A page speed audit identifies the work that takes time, which pages share that work, and what change you can test. Check when a large file loads before deciding whether shrinking it would speed up the main content.
Start with three kinds of evidence. Field data captures experiences from real visits over a reporting window. Lab data comes from a controlled browser test where you can see which resources load and what the browser does with them. Transfer measurements describe response times, headers, and file sizes. Use field data to establish where users struggle, lab data to investigate why, and transfer measurements to check what the site sends.
What it does for the business
- Give developers a repair they can estimateA ticket can name the template, the asset or rule to change, and the measurement to repeat. The developer can use that evidence to estimate the work.
- Fix shared causes onceA repeated delay across articles, offer pages, and the pricing page may come from a shared component. The client can schedule one change and budget for checks on every affected template.
- Protect the pages people need to actMap findings to the pages that carry a signup, quote request, or purchase. Given repairs with similar evidence and effort, the client can put customer paths first.
- Make release reviews easierA baseline gives the team a reference when a new theme, tag, or media upload changes how pages load. A retest helps the client decide whether to keep or adjust the change.
How to build a page speed audit to find what slows pages down, step by step
Choose pages and capture a repeatable baseline
A performance result is useful when someone else can repeat the measurement.
What to do
- Pick pages that exercise different templates. Include the home page, a main offer page, pricing, the blog index, and a long article where those exist. Add other layouts and language versions that load different content or resources.
- Gather field data when available. For Core Web Vitals diagnosis, record the device category and whether each result describes an individual URL, a group of pages, or the whole site. Core Web Vitals measure loading, responsiveness, and visual stability.
- Test each URL in a mobile browser profile. Keep the device, network throttling, cache settings, and location consistent. Log the test time, tool and version, and build if known. Repeat the run and keep every result.
- Collect transfer measurements as well. Record time to first byte, the wait from the request to the start of the response, and the compressed and decoded sizes. Capture headers and sizes for the stylesheets, scripts, fonts, and images the page references.
Why
A banner and a long article put different demands on the browser. Testing both helps you distinguish template-specific delays from shared ones. Repeated runs show whether a slow response is worth tracing.
Follow the delay through requests and rendering
The cause of a delay is the work the browser is waiting for when the delay occurs.
What to do
- Read the network waterfall, a timeline of requests and responses. Check for redirects ahead of the document, slow or variable server responses, and missing compression. For versioned assets, inspect the cache policy and test repeat visits separately.
- Match Largest Contentful Paint, the render time of the largest eligible image or text block visible in the viewport, to the element that produced it. For an image, see when the browser found it, how long the download took, and how long the browser waited before drawing it.
- Inspect render-blocking resources, such as stylesheets the browser must process before drawing the page. Check whether a synchronous script stops the browser from reading the HTML, and whether the main image is buried in code the browser discovers later.
- Review the performance trace, a timeline of the browser's work. Find long tasks during a slow interaction and identify the scripts running then. Investigate each layout shift, a visible element moving unexpectedly, by checking the element that moved and the content that pushed it.
- Compare the raw HTML with the rendered page. Check for headings, links, or main content inserted by scripts. Inspect excessive inline data and review image dimensions, loading settings, font files, and off-screen scripts across templates.
Why
A late request points to how the image is discovered. A long transfer points to the file or its delivery. A wait after download sends you back to the browser trace to see what may be delaying the draw.
Test one bounded repair at a time
An improvement has a unit, a baseline, and a test you can repeat.
What to do
- Turn each supported cause into a specific change. If a banner is requested late because it is marked for lazy loading, change how that one banner loads and check when its request starts. Keep lower-page images on their existing loading rules for that test.
- When an asset is oversized for its display, create a local replacement at the needed dimensions and compare bytes on the same basis. Check that it still looks acceptable on the target devices. Record the byte difference and measure browser timing separately if you can.
- Use the same URLs, device profile, throttling, and cache settings for a browser retest. Compare all runs before and after, and include an unchanged page to help spot a shift in the test environment or a wider change in delivery.
- Rank supported repairs by measured cost and pages affected, then consider effort and release risk. Give each unmeasured hypothesis a test that could resolve it. If the evidence points to excess CSS, identify the suspected file and test the change across templates that load it.
Why
A smaller asset gives you a measurable byte saving. A faster paint depends on whether that asset held up rendering in the first place. Keeping each repair small makes it easier to attribute a change in the trace and to catch side effects, such as a menu losing styles on a page outside the first test.
Check the evidence behind the highest-ranked findings
The stronger the recommendation, the easier its evidence should be to inspect.
What to do
- Ask a reviewer to open the evidence for the top findings and verify the affected URLs and assets. Have them recompute byte savings on the same basis and tie the cause to the original waterfall, rendered page, or trace.
- Check that the runs being compared share the conditions recorded in the baseline. If a faster run also used a warm cache or a different build, repeat the comparison with matching settings before reporting the timing difference.
- Read each claim for the scope its evidence supports. Report a group-level field result at the group level. Describe a reproduced browser delay with its test settings. Without field data, request it before assigning a Core Web Vitals verdict.
Why
An arithmetic or attribution error can send a developer to the wrong resource. A reviewer can also catch a sitewide claim inferred from one page, before the client shares the report with a hosting provider or developer.
Give the client a short list of changes to make
Every finding should tell the reader where the delay occurs and how to check the repair.
What to do
- Open the report with the largest supported cause and the evidence behind it. List the affected pages or templates, observed cost, proposed change, measured saving where available, and confidence in the finding.
- Link each repair to the relevant source location, tag, header, or build setting. Mark likely locations as hypotheses when you only have the public site. Give the owner a test and a clear result to look for after the change.
- Record what you inspected and found in good order, along with test coverage and evidence gaps. Give the client the most useful follow-up measurement, such as supplying the mobile field report or repeating the browser test after a template update.
Why
A short report helps an owner decide what to fund and gives the developer the detail needed to begin. The evidence also gives the SEO team a baseline for checking visitor measurements after release.
Common mistakes
- Testing the home page aloneA shared theme can perform differently on a long article, a product page, and a page with an embedded tool.
- Keeping the fastest runA favorable run can hide a delay that returns when the server or connection varies.
- Blaming the biggest fileCheck when the browser loads the file and what page work depends on it before assigning it a cost.
- Giving a field verdict from a lab runUse visitor measurements for a Core Web Vitals verdict and controlled browser runs to diagnose behavior under known conditions.
- Reporting bytes as time savedReport a smaller transfer in bytes until a comparable test shows a change in timing.
What a good result looks like
- Sampled URLs cover the important templates and language versions.
- Each performance figure has its timing, evidence source, and test conditions.
- Every supported cause points to an asset, an element, or a browser task.
- Each proposed change names its scope and the comparison that would verify it.
- The ranked list identifies measured gains, confidence, and hypotheses to test.
And then what? Using the result
- File the first repair as a ticket with the affected URLs, trace or response evidence, and retest steps.
- Give the owner of each shared template the full list of page types to check before release.
- Record the release date and rerun the same tests on changed pages and an unchanged page.
- Review field data after the change has had time to reach visitors and the reporting window reflects the release.
- Use unresolved hypotheses to choose the next capture if the delay persists.
When to run it
- After a template or hosting changeCompare new captures with the previous baseline under matching conditions. Check redirects and response times along with changes to the page's resources.
- When mobile visitors describe a delayUse the reported page and action to choose what to reproduce. A slow first view and a delayed filter change call for different browser measurements.
- When a client asks where to spend development timeRank repairs by the size and reach of the measured issue. Attach a test to each ticket so the client can determine whether the work helped.
Questions people ask
How do you find what slows pages down?
Test representative pages under repeatable conditions and trace each delay to the request, page element, or task responsible. Compare page types to find shared causes, then test a focused repair against the baseline.
What is the difference between field data and lab data?
Field data measures real visits over time, while lab data records what happens in a controlled browser test. Use the field view to see where visitors are affected and the lab trace to investigate causes you can reproduce.
How do you tell if the server is slowing a page down?
Check whether the document repeatedly has a slow time to first byte under comparable conditions. Use redirects, connection timing, and server-side measurements where available to locate the wait. Then trace the load after the first byte.
How do you find what delays Largest Contentful Paint?
Identify the element reported for Largest Contentful Paint and trace the time before it appears. For an image, compare the request start, download time, and wait before drawing. For text, check whether styles, fonts, or script work hold up its display.
How do you test whether a page speed fix worked?
Repeat the original measurements with the same URLs, device, network, and cache settings. Compare the full ranges and include an unchanged page. State the result in the unit you measured.
Can you audit a page without source code?
Yes, public pages and supplied reports can support findings about requests, rendering, and asset sizes. Mark the likely source location as an assumption when you cannot inspect the project, and give the developer a way to verify it.
More jobs in SEO agencies
Find what slows pages down, on your own Mac.Free app. Runs on your Mac with the Claude or Codex plan you already have. Your files stay on your computer.
Download for Mac