How to Review a Website: 7 Dimensions + 30-Point Checklist

A website review checks a site against fixed dimensions, before launch or after a redesign, to confirm it is usable, understandable, credible and converting — and turns every issue into a record that can be assigned and re-verified. It evaluates whether the site meets its goals, not whether every feature was implemented to spec. The latter is testing; the former is review.

You need both, but they use different checklists, involve different people and produce different outputs.

This article gives you a review process you can reuse as-is: separate review from testing first, walk seven dimensions, record findings against a 30-point checklist, and turn every issue into a record you can re-verify.

How review, testing, acceptance and usability testing differ

The four terms get used interchangeably, but they answer different questions:

Website review QA testing Acceptance (UAT) Usability testing
Answers Can this site meet its goals Does functionality work as specified Does the business accept the delivery Can real users complete the task
Who Product / design / marketing / client QA / developers Business owner / client Researcher + real users
Output Issue list + priority + evidence Defect tickets + pass/fail Acceptance decision Task completion rate + observation notes
When Before launch, before a redesign, on a schedule Every release Before handover Before or after design or launch

The one-line split: testing proves "it isn't broken", review judges "it is useful". An issue found in review is not necessarily a bug — it might be wrong information architecture, unclear copy, or a pricing page missing a restriction.

Scope first: what a review has to answer

The most common failure in a review is not carelessness. It is having no scope: open the homepage, scroll down, note whatever you see, and end up with 80 conflicting opinions.

Answer three questions first:

  • Who is it for? New visitors, existing customers, or internal stakeholders. The three care about different things.
  • Which pages? Don't spread across the whole site; pick samples by template.
  • What decision follows? Can we launch, do we need a redesign, which three things do we fix first. A review with no decision turns into opinion collection.
  • How to pick sample pages

    • The homepage and one core landing page
    • Every step of the core conversion path (sign-up, checkout, contact)
    • One page per template (content, listing, detail)
    • Exception states: 404, empty results, form errors, no permission
    • At least one pass through the main path on mobile

    10–30 pages are enough for a small or mid-sized site. The more pages there are, the more important it is to group them in advance, or something will be missed.

    The seven dimensions

    Dimension What to look at Most common problems
    Goals and information architecture Can the homepage explain what this is and who it's for; is navigation organized around user tasks Navigation named after company departments; homepage sells the vision, not the product
    Content Are headings, body copy, pricing and FAQ complete, accurate and current Placeholder copy; pricing page missing a billing unit; invented case studies
    Visual consistency Are components, spacing, colors and images consistent across pages Three styles for the same button; stretched images
    Usability and interaction Forms, error messages, keyboard operation, mobile tap targets Errors that only say "invalid input"; going back loses data
    Functionality and links Do core flows work end to end; are there 404s; do forms give feedback Broken external links; submission succeeds with no confirmation
    Performance and mobile First-screen load, image weight, horizontal scrolling, offline messaging A large hero image blocks the first screen; tables overflow on mobile
    Trust and compliance Privacy policy, terms, contact details, real case studies Refund rules cannot be found; no physical address

    Of the seven, content and trust are the ones technical teams skip most, because unlike functionality they have no clear pass/fail. But that is usually where users decide whether to stay.

    The seven-step process

    Step What you do Output
    1 Set scope Answer "who it's for, which pages, what decision" A one-page review scope
    2 Set the environment Devices, browsers, network, account roles Environment list
    3 Walk the dimensions Work through all seven on every page; don't skip any Raw issue list
    4 Keep evidence URL, screenshots, reproduction steps, console At least one piece of evidence per issue
    5 Classify Blocker / important / suggestion Issue list with priorities
    6 Assign Every issue has an owner and a due date Tasks or tickets
    7 Re-verify Walk the original path again Closure record

    Spell out the classification in step 5, or "important" becomes "whoever shouts loudest":

    • Blocker: a core flow is broken, or an external promise is wrong. Cannot launch without fixing.
    • Important: it works, but clearly hurts understanding, trust or conversion. Fix before launch where possible.
    • Suggestion: preferences, optimizations, future considerations. Does not block launch.

    The 30-point checklist

    Grouped by the seven dimensions, with one criterion each. Copy it straight into a spreadsheet.

    # Dimension Check Criterion
    1 Goals and information architecture Homepage 5-second test A first-time visitor can say what this is and who it's for
    2 Goals and information architecture Main navigation items No more than 7, named in user language, not internal jargon
    3 Goals and information architecture Next step on every page Every page has an obvious next action
    4 Goals and information architecture Length of key paths Sign-up / purchase / contact in no more than 3 clicks
    5 Content Headings Headings name the user's problem, not the feature
    6 Content Three things on a feature page What it is, who it's for, how to start — all three present
    7 Content Pricing and limits Billing unit, limits and refund rules are explicit
    8 Content Placeholders and stale content No lorem ipsum, sample copy or outdated years
    9 Content Real information About and contact pages have a real entity, address or contact method
    10 Visual consistency Consistent components Buttons, forms and cards look the same on every page
    11 Visual consistency Contrast and type size Body text has enough contrast; mobile body text is at least 16px
    12 Visual consistency Image quality Not stretched, blurred or distorted
    13 Visual consistency State design Empty, loading and error states are all designed
    14 Usability and interaction Form labels Labels sit outside the input; errors explain how to fix them
    15 Usability and interaction Keyboard operation The main flow can be completed by keyboard alone, with a visible focus
    16 Usability and interaction Mobile tap targets Primary buttons are at least 44×44px
    17 Usability and interaction Back and close Back, close and refresh do not discard entered content
    18 Usability and interaction Empty results Search / filter with no results explains why and offers a next step
    19 Functionality and links Core paths Sign-up, login, checkout and contact each complete at least once
    20 Functionality and links Dead links Sample core pages and external links; no 404s
    21 Functionality and links Form feedback Clear confirmation on both success and failure
    22 Functionality and links State persistence Login state, filters and pagination survive a refresh
    23 Performance and mobile First-screen time Main content visible within 3 seconds on 4G
    24 Performance and mobile Image weight No uncompressed hero images; images lazy-load
    25 Performance and mobile Horizontal scrolling No horizontal scrollbar on mobile
    26 Performance and mobile Weak-network messaging An understandable message when offline or timed out
    27 Trust and compliance Policy reachable Privacy policy and terms reachable from every page
    28 Trust and compliance Data explanation Forms say what is collected, why, and how to delete it
    29 Trust and compliance Social proof Case studies, reviews and client logos are real and checkable
    30 Trust and compliance Commercial rules Pricing, refunds and scope of service are findable

    A checklist is not better for being longer. The value of 30 points is that every review uses the same ruler, so two results can be compared.

    How to capture evidence

    A review comment most often dies on one sentence: "the button on the second screen is off." Three months later, nobody knows which page, which screen, which browser.

    A verifiable issue carries at least five things:

  • Page URL (keep the parameters if there are any)
  • Device and viewport (desktop 1440 / iPhone 14 / tablet)
  • Browser and OS
  • Reproduction steps (where you came from, what you clicked)
  • Screenshot or recording with annotation (point at the specific element; don't just capture the full screen)
  • Then add evidence by issue type:

    Issue type What to include Why
    Visual / layout Screenshot + viewport + browser Layout problems often appear at only one breakpoint
    Content / copy URL + screenshot + expected copy Avoids a vague "change this somewhere"
    Interaction / flow Reproduction steps + recording + account role Different roles see different states
    Functionality / errors URL + console error + failed request + timestamp Separates front-end, API and data causes
    Performance URL + network conditions + first-screen time "Slow" needs a comparable baseline
    Mobile Device, OS, browser + screenshot The same page differs a lot between device models

    A screenshot shows position but not environment. When BugCapturer's browser extension takes a screenshot or recording, it brings the page URL, viewport, browser and OS, console errors and failed requests along with it, and produces a share link that opens without installing anything. Review comments no longer need "the button on the second screen", and developers no longer have to ask "which browser were you on". See the browser extension and share links.

    What a review report looks like

    Don't hand over 80 scattered comments. One page is enough:

    text
    Scope: pre-redesign review of example.com; desktop 1440 + iPhone 14; Chrome / Safari
    Verdict: do not ship as is; 2 blockers
    Blocker: coupon does not apply on checkout (evidence #12)
    Blocker: mobile navigation covers content (evidence #18)
    Important: homepage headline is too abstract (evidence #03)
    Suggestion: case studies page lacks industry filters (evidence #27)
    Next: @Zhang on checkout, @Li on mobile; re-verify on the original paths on Friday
    

    Four parts, none optional: scope tells people what the verdict applies to; verdict spares the decision-maker from reading 80 items; classification decides what gets fixed first; owners and a re-verification date decide whether it actually gets fixed.

    Every issue in the report should link back to its evidence. Links expire and image hosts go down, so evidence needs to be kept with the record long term, not pasted into some chat window.

    Five common failures

  • Only capturing the homepage. The homepage is usually the most polished page; the problems live in template pages and exception states.
  • Only looking on your own laptop. Chrome at 1440 wide does not represent mobile, Safari or a slow connection.
  • Mistaking personal preference for a defect. "I don't like this blue" is not a finding; "the primary button lacks contrast against the background" is.
  • Issues with no owner. A list without owners is just opinions.
  • No re-verification. A review without re-verification hands the same issue to the next round.
  • Frequently asked questions

    How long does a review take? For 10–30 pages, 2–4 hours to walk through, plus 1–2 hours to organize and classify. The more pages there are, the more you should group and sample in advance rather than trying to see everything at once.

    Review before launch or after? Review the design files or staging before launch, focusing on goals, content and trust; then run a check within a week of launch, focusing on empty states, error states and performance under real data.

    What if the client won't cooperate, or there are no real users? Give them a task instead of asking how they feel. Ask the client to "sign up as a new user and complete one purchase" while you note where they get stuck; that is far more useful than "what do you think?"

    Do I need a tool? Under 20 issues, a spreadsheet plus screenshots is enough. Above 20, or when several people collaborate or you deliver across companies, use a tool. The value of a tool is that evidence travels with the record, not that you get another board.

    What is the difference between review and acceptance? Review is finding problems; acceptance is confirming whether the delivery is accepted. Review can happen at any stage; acceptance happens only at delivery points. Review outputs an issue list; acceptance outputs accept or reject.

    After the review comes fixing and verifying. For how to fix an issue until you can prove it is really gone, see how to fix bugs; for how to write a report a developer can start from, see the bug report format. If you need to judge whether functionality works as specified, use the website QA checklist and the web application testing checklist. To choose review and feedback tools, see website feedback tools compared and website markup tools.