A Mobile Performance Check for Industrial Websites

A fast home page does not establish that a buyer can use the product page, read a specification and send an enquiry from a phone. Performance needs to be checked along that route, including the awkward parts: a wide table, an attachment, a validation error and the return to a previous step.

This is an inspection guide for a digital lead who needs a defensible fix list. It does not estimate lost revenue from a speed score. That would require evidence about the audience, acquisition and sales process that a laboratory test cannot provide.

Choose a task, device and starting condition

Pick a relevant task such as finding an application limit and sending the question to sales. Start on the page buyers would actually enter, not automatically on the home page. Include a product or application page, its technical document and the enquiry route.

Record the device, browser, viewport, connection conditions and whether the browser cache is warm. Test repeat visits as well as a first load. Use emulation to find layout faults, then confirm important issues on a physical phone where possible.

Do not quietly remove the animation before measuring a site whose design depends on it. Inspect whether the experience remains usable while media loads, whether reduced motion is respected and whether returning to an earlier scroll position causes a failure.

Make a reproducible observation

Prioritise by the task that fails

An enquiry button covered by chat is a task failure even if all performance scores look good. A slow decorative image is a different problem if the buyer can already read and act. Describe the consequence before assigning priority.

For example, a technical table may scroll within its own container correctly while the rest of the page stays fixed. That is different from a page wider than the screen. Check touch and keyboard access, and keep the table's label visible enough to explain what is being compared.

When a fault is intermittent, retain the conditions under which it occurs. “Sometimes slow” gives an engineer little to reproduce. “Returning to the top after opening the menu leaves scrolling disabled” gives them a testable sequence. These are illustrative observations, not findings about a named customer site.

Use field data and lab tests for different jobs

What changed

From commit fb4d6f8 to 49bb07c: reserve mobile heading space, stabilise line breaks and align the hero from the top. Several other release changes were present, so this is not an isolated experiment.

Observed layout shift

CLS changed from 0.126 to 0.009. These are lab observations, not field Core Web Vitals or a claimed conversion improvement.

What did not materially improve

LCP was 3.76 seconds before and 3.75 seconds after. Correcting visual stability did not solve initial loading time.

Reproduction

Chromium through Playwright; Lighthouse mobile preset with simulated throttling. No physical iPhone measurement is implied. Compare the raw run conditions before reusing the numbers.

Read the sanitised measurement record for the exact metrics and environment. The subsequent full production crawl encountered connection timeouts and did not finish. That limitation belongs in the record: a successful lab run is not an end-to-end availability guarantee.

Field measurements describe observed experience across eligible visits. Lab tests help reproduce a condition and investigate causes. A single run cannot represent every device, network or page state. Compare like conditions and inspect variation rather than reporting the best score.

Google's page-experience guidance provides search context; WCAG 2.2 provides accessibility criteria. Neither establishes a specific conversion uplift for your business. Faster rendering also does not automatically fix crawl restrictions or indexing problems.

Retest the neighbouring behaviour

After changing an overlay, test navigation, chat and the form together. After optimising media, check the crop, first frame, loading state and motion preferences. A technically smaller file is not an improvement if it removes information the buyer needs.

Keep the before-and-after record with the change. Use the acquisition-versus-enquiry guide for the budget decision and the application-page guide for content structure. This inspection should produce reproducible defects and verified corrections, not a new name for every source of commercial friction.