How to give a website critique worth listening to

Start with a visit, rather than a verdict.

Decide which page and task you are reviewing. “Can someone on a phone understand the garden-room offer and find a way to enquire?” is a useful question. “Is this a bad website?” leaves rather too much room for opinion.

Record the viewport, capture date and part of the page you actually inspected. In our demonstration, the browser viewport is 390 × 844 pixels. That gives the observation a reproducible setting; it does not stand in for every phone, visitor or page on the site.

• Check the main offer, navigation and intended next action.
• Keep a frame showing the issue in context, with the page around it.
• Note what you have not tested, such as the rest of a contact form.
• Repeat a surprising measurement before making it the centre of your pitch.

Our website audit checklist gives you a wider set of checkpoints. Choose the relevant ones; a short personal review need not become an exhaustive compliance report.

Make a smaller claim. Make it stronger.

Separate the observation, your interpretation and your proposed change. They do different jobs. A gallery extending beyond the captured screen is visible evidence. Saying it might make browsing less comfortable is an interpretation. Recommending cards that fit the available width is a proposal.

“In this mobile visit, the project cards run beyond the right edge. I’d make each card fit the screen, so the projects are easier to explore.”

That is a recommendation someone can inspect and discuss. You have not measured lost enquiries, so there is no reason to announce that the business is losing customers. You have not run a controlled change, so there is no basis for promising a conversion increase.

The same care applies to search and accessibility. A missing description tag does not establish that a site cannot rank. An image with empty alternative text may be decorative. One browser timing is not a field measurement of the whole audience.

Keep the proof beside the spoken finding.

A prospect should be able to move from “you said this” to “I can see what you mean”. In our evidence timeline, each reviewed finding links to an original captured frame and the moment it is spoken in the video.

Those are two clocks. Our retained gallery frame comes from 27 seconds into the capture, while the presenter discusses it 7.54 seconds into the finished video. Guessing one from the other would give the appearance of precision without the proof.

Review the actual narration, then set the speech timings. Keep the approved script and video together. If either changes, review the evidence again. AI can help draft the explanation; the agency still needs to check that every claim is supported.

See this in our Mere & Timber walkthrough, including what its retained recording cannot prove.

Write an introduction someone would actually say.

Start with the business and the specific visit. Choose one or two useful observations. Explain each in ordinary language, then offer a practical next step. Reading the draft aloud is a good way to catch a sentence that belongs in a report rather than a conversation.

1. “I had a look at your garden-room website on a mobile.”
2. Show the finding and explain the proposed change.
3. “If useful, we can talk through what changing that would involve.”

These are illustrative lines, not a promise that a particular script will win work. Personalisation is the relevance of the observation, not the number of times you repeat somebody’s business name.

Give the ending room to land.

Say what you would like the viewer to do next, and show how to do it. Use your agency’s contact details and a closing message that fits the offer. Preview the whole ending, including the transition into the contact card.

We found this in our own demonstration: cutting away almost as soon as the last word finished made the video feel unfinished. Our composition now leaves at least two seconds for the presenter to settle, then uses the closing card and a final fade. A written CTA alone did not fix an abrupt edit.

The email and audit page should make the same invitation. A request to reply, a contact link and a booking option can support one next step; several competing offers make the choice harder to understand.

Measure whether the conversation moved on.

Delivery, an open event, playback, a booking and won work are distinct events. Keep their denominators visible. An open signal does not prove someone read the email, and a video play does not make a qualified opportunity.

Record replies and commercial outcomes when they actually happen. Respect the recipient’s playback choice and suppression status. Use the evidence to improve the next introduction, without turning a flattering chart into a claim it cannot support.

A useful review makes its limits easy to understand. This method helps structure a personal website conversation; it is not a full security assessment, accessibility certification or ranking guarantee.