A website can look ready and still fail at 320px.
A full-page screenshot is useful evidence, but it cannot prove that a button is reachable, a download works, or text is readable at a narrower width. Here's a practical launch check you can run yourself, based on mistakes found while testing Moruneva's own website.
We tested eight public Moruneva pages across 22 browser viewport combinations in an isolated Chromium session. That self-audit found a phone-width navigation overflow, a link that wrapped awkwardly and several text contrast issues. Fixes were subsequently checked with browser measurements and source tests. This is our own site, not a client's project, and a passing browser check is not a security certification.
1. Decide what must work before choosing screenshots.
Write down three user journeys before launch: find the right page, complete the main intended action, and return or recover if something goes wrong. A beautiful homepage is not a test plan. For a five-page site, choose the home page, pricing or services, an important content page, the action page and one supporting detail.
For each journey, record expected behavior in plain language: “At 320px the inquiry button stays fully visible and opens a correctly addressed draft email” is testable. “The site feels polished” needs a separate human review.
2. Check real page width at 320px, not only a 390px phone screenshot.
The WCAG 2.2 Reflow guidance uses an equivalent width of 320 CSS pixels as an important reference. Narrow layouts should not require two-dimensional scrolling for ordinary text and controls; some content with a necessary two-dimensional layout has exceptions.
Our own first run demonstrated why this matters. At a 320px viewport the document measured 344px and an inquiry button extended to x=328px. After correcting box sizing and responsive spacing, the same recorded case fit within 320px and the button's right edge measured x=304px. The change was supported by before/after measurements, not just a green build status.
| Width | What to inspect | Common failure |
|---|---|---|
| 320px | Navigation, CTA, long headings, tables | Content extends beyond viewport |
| 390px | Card stacking, touch targets, email actions | Actions wrap into misleading or tiny targets |
| 768px | Tablet columns and intermediate breakpoints | Desktop styling appears too early |
| 1280px | Content measure and paired cards | Two featured cards occupy one third each, leaving dead space |
Free check: in browser developer tools compare document.documentElement.scrollWidth with the active viewport width, and inspect the exact element that extends past the edge. For ordinary text, an unexpected wider document is a warning. A large image or genuine data table can require different handling.
3. Make screenshots and interaction evidence separate.
Use screenshots to capture visual state and transitions. Then actually test the links, navigation, keyboard focus, back button and promised outputs. A screenshot of “Download result” proves only that the control was rendered; inspect the downloaded JSON or CSV using fictional inputs to confirm the file exists and contains the expected fields.
Playwright can capture full-page and element screenshots for free; its web assertions can wait for a URL, heading, visibility or download-related state. If you already know how to use Playwright, you may not need to pay anyone to collect the images.
// Minimal read-only viewport sanity check (illustrative)
await page.setViewportSize({ width: 320, height: 800 });
await page.goto('https://example.com', { waitUntil: 'domcontentloaded' });
const width = await page.evaluate(() => document.documentElement.scrollWidth);
if (width > 320) throw new Error('Unexpected horizontal overflow');
await page.screenshot({ path: 'page-320-full.png', fullPage: true });
This example checks only one property; it does not test access permissions, user flow, production errors or whether the page is actually usable. Never use real credit cards, customer records or live account credentials as test data without the owner's explicit authorization.
4. Report failures so another person can reproduce them.
A useful launch report includes a short headline, severity, URL, viewport, observed outcome, expected outcome, reproduction steps and evidence filename. If there is no bug, the report should still say exactly what was inspected and what was not inspected. A page saying HTTP 200 is not proof that its content is correct.
Example bug note (our own site, already corrected): “P1 — inquiry action clipped on narrow viewport. URL: /work-with-us. Width: 320px. Actual: document measured 344px, CTA right edge x=328px. Expected: CTA fits the visible content width. Reproduce: open the page at 320px, scroll to the top navigation/action, inspect document width. Evidence: before/after browser measurements and images.”
5. Keep release gates separate.
A healthy pre-launch checklist has at least four independent answers: the source tests passed, the website was deployed, the live HTTPS content loaded, and real browser interactions matched the brief. Further questions—cross-browser compatibility, accessibility conformance, security policy behavior and customer conversion—need their own evidence. None is automatically proved by another one.
We learned this the hard way: early source-level regression checks passed even while visible responsive issues remained. We therefore added actual Chromium element-position tests at six widths to our build workflow. These are useful for preventing the exact problems they assert; they still do not replace a review of the published site.
Before you publish: a five-minute handoff checklist
- List the pages and the two or three user actions that must work.
- Open at 320, 390 and desktop width; check text, cards, nav and CTA geometry.
- Click important links, press Back, check keyboard focus and inspect real test downloads.
- Save browser/device width, actual result, expected result and an image for each issue.
- State limits: untested browsers, gated flows, account data and anything not in scope.
Need a review another person can act on?
Moruneva has an introductory, fixed-scope US$95 website launch check: up to five public URLs, phone and desktop widths, reproducible findings, screenshots and one ZIP handoff. It is an inquiry-based pilot, not instant checkout. You can also do the checks above yourself for free.
Sources and disclosure
- W3C WAI — Understanding Reflow (WCAG 2.2)
- Playwright — Screenshots
- Playwright — Test assertions
- Playwright — Visual comparisons and environment limitations
Moruneva field note, 11 November 2026. The numerical example is a documented self-audit of this website, not third-party client results. It does not establish that all devices, search engines, payment systems or security features have passed. No external website was scanned for this article; technical examples are illustrative.