You can run a complete pre-launch QA pass on a small site in one afternoon. The trick is order: broken forms and links cost money the moment you go live, while a missing alt tag does not. This checklist runs in that order, and every item tells you what to click, where to look, and what proves a pass.
You will get a three-tier checklist (stop-ship, should-fix, backlog), six testing blocks covering conversion paths, links, SEO, speed, mobile, accessibility, security and analytics, plus a short 72-hour post-launch re-test. Everything is written for one non-engineer testing alone.
The path in 7 steps
The steps of this page, in order.
-
01 Know your blockers Sort every issue into stop-ship, should-fix or backlog before you start clicking.
-
02 Test the money path Walk every form, signup and checkout the way a real visitor would.
-
03 Check links and index Crawl the site for 404s, bad redirects and duplicate canonical tags.
-
04 Measure speed Run key templates through a field or lab performance tool and compare to published thresholds.
-
05 Try small screens Open your top pages on a real phone and a second browser.
-
06 Sweep access, security, tags Check alt text, contrast, keyboard use, HTTPS, exposed paths and analytics events.
-
07 Re-test for 72 hours After launch, confirm forms, redirects and tags still fire on the live domain.
01 / 12
What absolutely has to be tested before launch?
Test in this order: conversion paths, links and index tags, speed on key templates, mobile, accessibility basics, security hygiene, then analytics and consent.
The first four items on the list move revenue or rankings right away. The last three protect you from silent damage that compounds over time. Fix them in this order so the most expensive problems get caught first.
A solo tester at a small business can finish the stop-ship tier in two to three hours on a small site, and the full pass in one afternoon. What you can actually cover depends on your CMS, your hosting and whether you have staging. Note those three before you start.
-
Open staging and set a baseline
Load the site in a clean browser profile. Note the CMS, hosting and staging setup, then start a timer.
-
Test conversion paths
Submit every form, complete a checkout or signup, and confirm confirmation emails arrive and data lands where it should.
-
Check links, redirects and index tags
Click primary navigation and footer links, verify 301s point to the right pages, and confirm no stray noindex tags or wrong canonicals.
-
Measure speed on key templates
Run PageSpeed Insights or Lighthouse on home, a product or service page, and a blog post. Record LCP, INP and CLS against Google's published thresholds.
-
Verify mobile, accessibility and security basics
Resize to a phone viewport, tab through forms, check heading order and alt text, and confirm HTTPS, security headers and no exposed admin pages.
-
Confirm analytics, events and consent fire
Use browser dev tools or a tag inspector to watch pageviews, key events and consent signals trigger in real time before you sign off.
02 / 12
Is the site fast enough on the pages that matter?
Measure only the templates that carry traffic or revenue, homepage, top landing pages, a product or service page and the blog post template.
Start with field data: the Chrome UX Report (CrUX) shows how real visitors experienced your URLs over the last 28 days. If a template has no field data, run Lighthouse or PageSpeed Insights in lab mode and treat the result as a hypothesis, not a verdict.
Compare against Google's published Core Web Vitals thresholds, an LCP of 2.5 s or less, INP of 200 ms or less, CLS of 0.1 or less at the 75th percentile. A 95 performance score with a failing LCP is still a fail; the score is not the metric.
Worked example
From raw LCP delay to a passing LCP
- Largest contentful element
- Hero image (product page)
- Measured LCP (field, 75th percentile)
- 3.1 s
-
TTFB0.6 s
server response + connection
HTML arrives after 600 ms.
-
Resource load time1.8 s
download + decode of the hero image
The image is served at 2400 px wide for a 600 px slot.
-
Render delay0.7 s
main-thread work before paint
A third-party chat script blocks rendering.
-
LCP before fix3.1 s
0.6 + 1.8 + 0.7
Above the published 2.5 s threshold.
-
Hero image saving−0.8 s
resize to 600 px + convert to WebP/AVIF
Resource load drops from 1.8 s to 1.0 s.
-
LCP after fix2.3 s
0.6 + 1.0 + 0.7
Below 2.5 s without touching TTFB or the third-party script.
03 / 12
What should I have ready before I start clicking?
Gather five things first, or you will bounce between tabs and lose the thread.
Set up a staging URL that mirrors production content, a CMS login, and a test email inbox you control. Without all three, you cannot verify a form submission or check what a real confirmation message looks like.
Then write down your top five pages by traffic or revenue, and inventory every form, signup, and checkout on the site. Keep one spreadsheet or issue tracker open so each failure lands there with a screenshot and the exact URL.
Staging access
Confirm the staging URL loads the same content as production. Bookmark it so you never test on the live domain by accident.
CMS login
Log in once before the pass starts. If a password reset is pending, you will lose twenty minutes mid-checklist.
Test inbox
Use an email address you own and can open. This is where you confirm form notifications and autoresponders actually arrive.
Page list
Name your top five pages by traffic or revenue. Those get tested first, and they set the bar for everything else.
Form inventory
List every form, signup, and checkout on the site. A form you forgot to count is a form you will not test.
Issue tracker
Open one spreadsheet or tracker before you start. Every failure goes in with a screenshot, the URL, and the browser you used.
04 / 12
How do I test forms and the conversion path?
Submit every form the way a real visitor would, then confirm the success page, the error state and the email that lands.
Start with a valid submission: fill every required field, use an email address you can open right now, and keep the message short. Then break it on purpose. Submit with an empty required field, a made-up address and a 2,000-character message. You want to see a clear error next to the field, not a generic banner or a silent failure.
Once the success message appears, check that it names the right form. Then open the inbox: the confirmation email should arrive in a few minutes, not in spam. If you take payment, run a live test with a small amount when your processor supports it, or at least verify the sandbox end-to-end and confirm the live key is switched on.
Required fields
Submit with one required field blank. You should get a specific message next to that field, and the form should not send anything.
Validation messages
Type an email like test@test and a phone number with letters. Each field should show its own plain-English error, not a generic red banner.
Success state
Complete a valid submission. The page you land on should name the form you filled in, so you know it is the right one.
Error state
Force a failure: use a duplicate email, or temporarily block the send endpoint. The user should see a message with a next step, and no data should silently disappear.
Email delivery
Open the inbox you used. The confirmation should arrive within a few minutes and sit in the main inbox, not in spam or promotions.
CRM or autoresponder record
Search the CRM or email tool for the exact address you submitted. The record should exist with the right tags and list, within a minute or two.
Going further
See common form failures
Special case
Why forms fail on launch day even when they worked in staging
Forms usually break on launch day because of environment switches that nobody flipped between staging and production.
The first cause is an SMTP placeholder. In staging, the sending address is often something like no-reply@localhost, and the mail server quietly drops it. In production, that placeholder is still there, so the visitor sees a success page but no email ever arrives.
The second cause is a different spam policy. Staging mail often bypasses filters, while production sends through a real domain. If SPF, DKIM or DMARC are not set up, the confirmation lands in spam or is rejected outright.
The third cause is missing reCAPTCHA keys. The test keys work on localhost, but the live keys are tied to your production domain. If they were never pasted into the live environment, the form either fails visibly or lets bots through.
- Sending addressConfirm the live form uses a real domain address, not localhost or a placeholder.
- Authentication recordsCheck that SPF, DKIM and DMARC are published for the sending domain.
- Spam testSend a real submission to a Gmail and an Outlook address you own, and see where it lands.
- API endpointsCompare the live keys and URLs against the ones used in staging, one by one.
05 / 12
How do I find broken links, bad redirects and wrong canonicals?
Crawl the site with one free tool, then spot-check the highest-value URLs by hand.
Run a free crawler against staging so it reports the whole link graph at once. Export the internal 404s and the redirect chains, then fix any chain longer than one hop: a single 301 from the old URL to the new one is what you want.
Then open your top five pages and read the canonical tag, the line of HTML that names the preferred URL for a page. It must point to that page on the live domain, not to staging, and www versus non-www plus http versus https must resolve the same way everywhere.
Crawl output, checked by hand
| URL tested | Expected redirect | Canonical target | Pass/Fail |
|---|---|---|---|
| /old-pricing | 301 to /pricing | /pricing | Pass |
| /blog/post-1 | None | /blog/post-1 | Pass |
| https://www.example.com/about | None | /about | Fail: canonical points to staging |
| /services/seo | 301 to /services/search | /services/seo | Fail: chain of two hops |
Hypothetical examples. Your pass/fail column is the one that matters: every canonical must match a live URL.
06 / 12
Does the site work on phones and other browsers?
Open your top pages on one real iPhone, one real Android and one non-default desktop browser, not just the responsive preview.
Start with your three highest-traffic pages plus one form page. On each, check that tap targets are at least 44 pixels, body text is readable without pinch-zoom, and nothing forces horizontal scroll.
Then walk one full conversion path on a real phone: open the menu, tap the form, fill it, submit, and confirm the email arrives. Watch that the sticky header never covers the submit button and cookie banner buttons stay tappable.
Mobile and cross-browser sanity matrix
| Page | iPhone Safari | Android Chrome | Desktop Chrome | Desktop Safari |
|---|---|---|---|---|
| Homepage | Pass | Pass | Pass | Pass |
| Top landing page | Pass | Pass | Pass | Pass |
| Form / contact page | Pass | Pass | Pass | Pass |
| Full conversion path | Pass | Pass | Pass | Pass |
Run this grid on real devices over Wi-Fi and one slower cellular connection. Mark any cell with a tap-target, zoom, horizontal scroll, header overlap or menu failure as a fail.
07 / 12
What accessibility basics should I check?
Four checks catch most real-world failures, and every one is doable without a screen reader.
Keyboard testing is the fastest signal: press Tab from the address bar and watch where focus goes. Every link, button and field needs a visible ring, in an order that matches the reading flow. A focus indicator hidden behind a sticky header fails, not just an invisible one.
Then check text alternatives: every image needs alt text that carries its meaning, and decorative images need an empty alt. Buttons and links must make sense read alone, and body text must hit WCAG 2.2 AA contrast, the W3C's 5 October 2023 standard.
Keyboard focus is visible
Tab through the homepage and landing page. Every interactive element must show a clear focus ring. If you cannot see where you are, keyboard users cannot either.
Focus order is logical
Tab from top to bottom. Focus should move in the same order you would read the page. A skip link should be the first thing a keyboard user reaches.
Images have alt text
Check each image in the inspector. Informative images need a short description of what they convey; decorative images need an empty alt attribute, not a missing one.
Link and button text works alone
Read every link and button out of context. "Learn more" fails; "Download the 2026 benchmark" passes. This also helps screen reader users who navigate by links list.
Contrast meets WCAG 2.2 AA
Test each body-text and font pairing with a contrast checker. Normal text needs 4.5:1; large text (18pt or 14pt bold) needs 3:1. WCAG 2.2 was published by the W3C on 5 October 2023.
Form fields have labels
Every input needs a visible, programmatically associated label. Placeholder text is not a label. Check that error messages name the field and say how to fix it.
Going further
See what changed in WCAG 2.2
Concept
Where WCAG 2.2 differs from 2.1 for typical marketing sites
WCAG 2.2 adds three checks that affect marketing sites directly: focus not obscured, dragging movements, and target size.
WCAG 2.2 kept everything from 2.1 and added nine success criteria. For a typical marketing site, three matter most. Focus Not Obscured (Minimum) requires that when an element receives keyboard focus, it is not entirely hidden by sticky headers, cookie banners or chat widgets. Dragging Movements says any action done by dragging must also work with a single pointer without dragging. Target Size (Minimum) sets a 24 by 24 CSS pixel floor for interactive targets, or enough spacing to avoid accidental taps.
These map to simple checks. Tab to every link near the top of the page and confirm the focus ring is visible below any sticky bar. If you have a slider or map, test it with click or tap only, no drag. On mobile, tap the smallest buttons and links, like footer social icons and pagination dots, and see whether your thumb hits the right one.
The rest of WCAG 2.2 is mostly relevant to complex apps or authentication flows. For a brochure or landing page, the 2.1 checks plus these three cover the practical surface area. If you are unsure, the W3C's own "Understanding" pages give plain-language examples for each criterion.
- Focus not obscuredWhile tabbing, confirm no focused element is fully hidden behind a sticky header, cookie bar or floating chat button. Scroll position must adjust if needed.
- Dragging alternativesAny drag interaction (carousel, map, range slider) must have a click or tap-only equivalent. Test with a mouse and with touch.
- Target size minimumInteractive targets should be at least 24 by 24 CSS pixels, or have enough spacing. Check footer links, icon buttons and carousel dots on a phone.
- No new 2.1 regressionsWCAG 2.2 does not remove 2.1 criteria. Keep your existing checks for contrast, labels, alt text and keyboard order.
08 / 12
What security and privacy items should I verify?
Six hygiene checks catch the mistakes that put a small site in the news, and none of them require a security background.
Load every key page and watch the console. If a secure page pulls an image or script over insecure HTTP, that is mixed content, and the padlock may disappear or a warning may show. Browse the HTTP version of each URL too; it should redirect to the HTTPS one.
Try opening /admin, /wp-admin and /staging in a private window. They should not be reachable. On the CMS login, confirm 2FA is on (a second code from an app or text) and no default usernames remain.
HTTPS on every page
Load each key URL and confirm the address starts with https:// and the padlock is intact. If a page loads over HTTP, it fails.
No mixed content
Open the browser console and reload. Any line naming a http:// resource on an https:// page is mixed content and must be fixed.
HTTP redirects to HTTPS
Type the http:// version of a page. It should send you to the https:// version automatically. If it loads insecurely, it fails.
Admin and staging paths are not public
In a private window, try /admin, /wp-admin and /staging. You should get a login wall, a 404, or a block, never a live page.
No default credentials, 2FA on
Check the CMS user list for names like 'admin' or 'editor'. Turn on two-factor authentication for every account that can publish.
Privacy policy and consent in place
Confirm the privacy policy is dated and current. Before any non-essential cookie is set, a consent banner should ask the visitor to accept or reject it.
Going further
See the header checklist
Method
Reading the five security headers that matter most
The five headers that quietly do the most work, and where to see them in devtools.
Security headers are instructions your server sends with each page. They cost nothing and close common gaps. You can see them in the browser: open devtools, go to the Network tab, reload the page, click the document request and look under Response Headers.
Strict-Transport-Security (HSTS) tells the browser to always use HTTPS for your domain. Content-Security-Policy (CSP) lists which sources of scripts and images are allowed, which limits injected code. X-Content-Type-Options: nosniff stops the browser from guessing file types. Referrer-Policy controls how much of your URL is passed to other sites. Permissions-Policy turns off browser features like camera or microphone on pages that do not need them.
- Strict-Transport-SecurityPresent with a max-age value, for example max-age=31536000.
- Content-Security-PolicyPresent and not set to allow everything.
- X-Content-Type-OptionsSet to nosniff.
- Referrer-Policy and Permissions-PolicyPresent with a value you understand, such as strict-origin-when-cross-origin.
09 / 12
Do analytics, events and consent actually fire?
Open the browser's network tab or your tag debugger, then walk the conversion path and watch each event land.
Tag firing is the moment an analytics or marketing script actually sends data to its platform. Use your tag debugger or the network tab and confirm each event fires once, with the right page path.
Check that the consent banner blocks non-essential tags before acceptance, allows them after, and remembers the choice. Then verify the email or CRM handoff gets the same identifying data as analytics.
Pageview
Load each key page and confirm exactly one pageview hit in the network tab or tag debugger, with the correct URL path.
Key events
Walk the conversion path (button click, scroll depth, form start) and verify each event fires once with the expected parameters.
Conversion pixel
Submit a test form and confirm the conversion pixel fires once, with no duplicate hits from redirects or thank-you pages.
No duplicates
After a single action, check the network log for repeated hits from the same event; duplicates usually come from a tag firing on both page load and click.
Consent gating
Before accepting cookies, confirm non-essential tags (analytics, ads) are blocked. After accepting, they fire. Reload and check the choice is remembered.
CRM handoff
Submit a test lead and verify the email or CRM receives the same email address, name, and source parameters that analytics records, so attribution stays intact.
10 / 12
What should I re-test in the first 72 hours?
The launch is not the finish line. Some checks that passed on staging can fail on the live domain, so re-run them within 72 hours.
Staging lies in small ways. Mail relays, spam filters and API endpoints usually point somewhere else on the live domain, so the same form that passed last week can silently drop submissions. Send one real test through every form on production and confirm the message arrives in the inbox you actually monitor.
Redirects and canonicals also resolve differently once DNS switches. Watch your analytics for new 404s and compare conversion volume with the prior week. Then confirm the cookie banner, privacy policy and consent log all match what is live, and log every error you find.
Form delivery (hour 1)
Submit each production form with a real address and confirm arrival, reply-to and spam placement.
Redirect map (hour 1)
Click every high-traffic legacy URL and verify it lands on the intended live page, not a soft 404.
404 log (hour 24)
Review server and analytics 404 reports for URLs that were live on staging but missing in production.
Analytics baseline (hour 24)
Compare sessions and conversions against the previous week to catch a broken tag or a silent drop.
Consent log (hour 48)
Load the banner in a clean browser and confirm the recorded choice matches the live privacy policy.
Error log review (hour 72)
Read the application and CDN error logs once more, then move anything unresolved into your issue tracker.
11 / 12
Which of these issues must block the launch?
Only the stop-ship tier blocks the launch. Everything else can ship today with a named owner and a date.
Stop-ship covers anything that loses money or leaks access: a broken form or checkout, an exposed admin or staging path, missing HTTPS, a broken canonical on a top page, a dead consent banner. Fix these before the site goes live.
Should-fix within a week includes missing alt text on key images, one broken internal link on a low-traffic page, or a heavy image on a secondary page. Backlog is cosmetic issues, one-off warnings, and improvement ideas, record each one, name an owner, set a deadline.
Launch today with a short fix list
- Clears the stop-ship tier only
- Releases value immediately
- Keeps the fix list short and trackable
- Should-fix items wait a week
- Backlog can stall without a named owner
For teams whose stop-ship tier is already clear and who can name an owner per remaining item.
Delay 24 hours to clear should-fix items
- Alt text, broken links and heavy images get fixed
- One short sprint, no project drift
- Slips a dated campaign or announcement
- Tempts scope creep into the backlog
For launches tied to a dated campaign where a one-day slip costs little.
Delay a week to clear the backlog
- Nothing unplanned ships
- Time to re-test after each fix
- Rarely worth it for cosmetic-only items
- A week of drift invites new scope
For sites where every item is genuinely user-facing and the go-live date is flexible.
12 / 12










