Most website design functionality evaluation criteria you'll find online are opinion dressed as a checklist — “is it modern?”, “does it pop?” Those questions can't be failed, which makes them useless. This scorecard uses 42 checks that are either objectively true or objectively false, grouped into the six categories that actually predict whether a site earns money or quietly loses it.
Use it two ways: to audit the site you already have before you agree to a rebuild, and to grade the work of whoever builds the next one. Every check below is pass/fail, and most can be verified in under a minute with free tools. Score honestly — a site that scores 30/42 is not “a B-minus,” it's a site with twelve specific defects you can now name and price.
Why most evaluation criteria fail
The standard failure is grading aesthetics instead of function. Aesthetics matter — but they matter because of what they signal, not because of taste. The most-cited research here is Stanford's, and it's routinely misquoted: you'll see “75% of users judge a company's credibility on website design” repeated across hundreds of agency blogs with no primary source attached.
The actual study is worth knowing, because it's stronger than the misquote. B.J. Fogg's team at Stanford's Persuasive Technology Lab had 2,684 people evaluate the credibility of live websites and comment freely on what drove their judgment. “Design look” appeared in 46.1% of comments — more than any other factor, including information accuracy, company reputation, and customer service (Fogg et al., How Do Users Evaluate the Credibility of Web Sites?, DUX 2003).
So credibility is a real criterion. But it's one of six, and it's the only one where taste enters at all. The other five are measurable to the decimal point.
1. Credibility and first impression (7 checks)
These are the checks a stranger's brain runs in the first few seconds, before reading a word of your copy.
- A specific value proposition is visible without scrolling. Not a slogan. A sentence a stranger could repeat back correctly.
- Real photography of real people, places, or work. Stock photos of a diverse team high-fiving in a glass conference room are a negative credibility signal, not a neutral one.
- A physical address and a real phone number appear in the footer. Absence reads as evasion.
- Named human beings appear somewhere on the site — an about page with faces and titles, not “our team of experts.”
- Proof is specific and attributable. “Increased leads by 40% for Miller HVAC in Q2” beats five unattributed five-star quotes.
- Typography is consistent — two typefaces maximum, a deliberate scale, and no more than one weight change per hierarchy level.
- Nothing on the page is visibly broken at any common viewport: no overlapping text, no images stretched off-ratio, no horizontal scroll on mobile.
2. Performance: the criteria with a price tag (7 checks)
This is the category where evaluation stops being subjective entirely. Deloitte and Google analyzed 30 million user sessions across 37 brands over four weeks and isolated the effect of mobile site speed. A 0.1-second improvement in load time produced an 8.4% lift in retail conversions with a 9.2% higher average order value, and a 10.1% lift in travel conversions (Deloitte, Milliseconds Make Millions).
Google's Core Web Vitals give you the thresholds for free. Note that Interaction to Next Paint (INP) replaced First Input Delay as a Core Web Vital on 12 March 2024 — if an agency's audit still reports FID, the audit is stale (web.dev).
- Largest Contentful Paint (LCP) under 2.5s at the 75th percentile of real users, not in a lab test.
- Interaction to Next Paint (INP) under 200ms. If the report says FID, it's out of date.
- Cumulative Layout Shift (CLS) under 0.1. Nothing jumps as the page loads.
- Field data exists in Search Console's Core Web Vitals report — meaning the site has enough real traffic to be measured at all.
- Images are served in a modern format (WebP or AVIF), correctly sized, and lazy-loaded below the fold.
- Total page weight under ~2MB on the homepage and key landing pages.
- The site is usable on a throttled 4G connection — test it in Chrome DevTools' network throttle, not on your office fiber.
3. Conversion architecture (7 checks)
A beautiful site that doesn't route visitors toward a decision is an expensive brochure. This is the category most agencies skip in their own evaluation criteria, because it's the one that exposes whether the site was designed around your buyer or around the designer's portfolio.
- Every page has one primary call to action, and it's the same one a reasonable visitor to that page would want.
- The primary CTA is reachable without scrolling on mobile, and repeats at natural decision points on long pages.
- Forms ask for the minimum viable information. Every additional field is a tax on completion rate.
- Phone numbers are tap-to-call links (
tel:) on mobile — a shockingly common miss on service-business sites. - There's a defined next step for visitors who aren't ready to buy — a guide, a tool, a self-assessment. Not everyone converts on visit one.
- Form submissions are tracked as conversion events in analytics, and someone can tell you last month's number without opening a spreadsheet.
- Thank-you pages exist and are distinct URLs — otherwise you cannot attribute a lead to a campaign.
4. Accessibility and legal exposure (7 checks)
This category has moved from “nice to have” to “quantifiable liability.” UsableNet tracked 4,928 web accessibility lawsuits filed in 2025, with e-commerce accounting for roughly 69% of them — and notably, 1,427 of those suits targeted companies that had already faced a prior claim (UsableNet 2025 Year-End Report).
- All meaningful images have descriptive alt text, and decorative images have empty alt attributes.
- Text contrast meets WCAG 2.1 AA — 4.5:1 for body copy, 3:1 for large text. Light-gray-on-white is the single most common failure.
- The entire site is keyboard navigable, with a visible focus indicator. Tab through it yourself.
- Form inputs have programmatically associated labels, not just placeholder text.
- Heading structure is semantic and sequential — one H1, no skipped levels, headings used for structure rather than font size.
- Video content has captions, and nothing critical is conveyed by color alone.
- An automated scan returns zero critical violations (axe DevTools or Lighthouse) — necessary but not sufficient; automated tools catch roughly a third of real issues.
5. Technical SEO and discoverability (7 checks)
You can fail every one of these while having a site that looks immaculate. That's precisely why they belong in a functionality evaluation rather than a design review.
- Every important page has a unique title tag and meta description written for a human, not stuffed with keywords.
- An XML sitemap exists, is current, and is submitted in Google Search Console.
- Canonical tags are correct and there's exactly one indexable version of the site (https, one of www/non-www).
- Structured data validates — Organization at minimum, plus LocalBusiness, Article, or FAQPage where applicable.
- Internal linking is deliberate: important pages are reachable in three clicks and receive links from related content.
- No orphaned or broken pages — crawl it and check.
- New and updated pages get submitted for indexing automatically, not by hand when someone remembers.
6. Ownership and maintainability (7 checks)
This is the category that separates a website you own from a website you rent, and it's almost never on anyone else's evaluation criteria — because most agencies would rather you didn't ask.
- You own the domain registration, in an account in your company's name that you can log into right now.
- You own the hosting account, or can move hosts without permission from anyone.
- You have the source code — in a repository you control, not on a vendor's laptop.
- Content can be edited by a non-developer without filing a ticket and waiting three days.
- Automated backups run on a schedule, and someone has actually tested a restore.
- The platform and plugins are on supported versions, with a documented update cadence.
- Analytics and Search Console are in accounts you own, not the agency's. This one bites people at the exact moment they change agencies.
Scoring: what your number actually means
A rebuild at this score usually destroys value rather than creating it — you'd be resetting accumulated SEO equity and known-good conversion paths to chase a visual refresh. Fix the specific failed checks, invest the rebuild budget in content and the systems behind the site, and re-score in six months.
Most sites land here. The failures usually cluster in two categories — commonly performance and conversion architecture — which means a focused engagement fixes the majority of the gap for a fraction of a rebuild's cost. Get the failed checks priced individually before anyone quotes you a full redesign.
Below 24, the failures are usually structural: the platform, the template, or the ownership arrangement is the root cause, and patching individual checks won't hold. Rebuild — but write these 42 checks into the contract as acceptance criteria so the next site can't score this low on delivery day.
What a real before/after comparison looks like
When you search for website design before/after comparison examples, you'll mostly find screenshot pairs — old site on the left, new site on the right, no numbers anywhere. That's a portfolio, not evidence. A defensible before/after states the same measurements on both sides:
Ask any agency you're considering for one case study in the right-hand format. The request itself is a useful filter: firms that measure their work can produce it in a day, and firms that don't will send you more screenshots.
Turning the scorecard into a hiring brief
The practical value of scoring your site is that it converts a vague feeling — “our website isn't working” — into a specific, priceable list. That changes the buying conversation completely. Instead of accepting a package quote against a one-paragraph brief, you hand over 12 named defects and ask what it costs to fix each one.
Send the failed checks, not a redesign request
“We need a new website” invites a template quote. “We fail LCP at 4.1s, have no tap-to-call on mobile, and don't own our analytics account” invites a scope. The second gets you a better price and a better partner, because it filters for firms that can actually read the list.
Write the criteria into acceptance terms
Specify the Core Web Vitals thresholds, the WCAG 2.1 AA standard, and the ownership checks as conditions of final payment. Reputable firms will agree without blinking. The ones that push back are telling you something valuable for free.
Decide whether the website is really the bottleneck
Sometimes a site scores 38/42 and the business still isn't growing — because the leak is in what happens after the form is submitted. If leads sit in an inbox for two days, or your team re-types every enquiry into three systems, a redesign fixes nothing. That's a workflow problem wearing a website costume.
That last one is where most of our own client work starts. We build custom websites and software, but the honest first question is always whether the site is the constraint or just the most visible symptom — the same question we walk through in is your business ready to develop custom software and how to define which processes to automate.
The honest verdict
Forty-two pass/fail checks will tell you more about your website in an afternoon than a year of “I think it needs a refresh.” Score it, count the failures, and group them: if they cluster in performance and conversion, you need remediation. If they cluster in ownership and platform, you need a rebuild. If they cluster in what happens after a lead arrives, you need neither — you need the workflow behind the site rebuilt instead.
Once you know which of those three you're in, the buying decisions get straightforward: what website design packages actually contain at each price tier, and how to choose a partner — whether that's a B2B or SaaS website design agency, a healthcare website design company that understands HIPAA, a manufacturing website design company that publishes specs engineers can actually use, a firm that understands law firm website design cost and bar advertising compliance, or one that knows what actually books jobs for home services trades.
If you'd rather not run the scorecard alone, that's what we do — we design and build the site, the software behind it, and the automation that connects them, and we'll tell you plainly if you don't need all three. Send us your failed checks and we'll price them individually, or start with the free automation audit to find out whether the website is the real constraint.
Frequently Asked Questions
What are the most important website design functionality evaluation criteria?
Six categories, in this order of business impact: performance (Core Web Vitals — LCP under 2.5s, INP under 200ms, CLS under 0.1), conversion architecture (one clear CTA per page, minimal forms, tracked events), accessibility (WCAG 2.1 AA), credibility signals, technical SEO, and ownership of the domain, code, and analytics. Performance is first because it's the one with measured revenue effect: Deloitte found a 0.1-second mobile speed improvement lifted retail conversions 8.4%.
How do I know if my website design is outdated?
Ignore visual trends and test function. A site is genuinely outdated if it fails Core Web Vitals on mobile, isn't keyboard navigable, has no structured data, uses placeholder text instead of form labels, or can't be edited without a developer. A visually dated site that passes all 42 checks is costing you far less than a fashionable one that fails twelve of them — and template-based sites often fail on performance specifically, because they load framework code for features you never use.
Is it better to redesign my website or fix the existing one?
Score it out of 42 first. Above 36, optimize — a rebuild resets accumulated SEO equity and proven conversion paths for a visual refresh you don't need. Between 24 and 35, remediate the specific failures; the fixes usually cluster in two categories and cost a fraction of a rebuild. Below 24, the problems are typically structural (platform, template, or ownership) and patching won't hold, so rebuild — but write the evaluation criteria into the contract as acceptance conditions.
Do automated tools like Lighthouse cover accessibility evaluation?
Only partially. Automated scanners reliably catch contrast failures, missing alt attributes, and unlabelled form inputs, but they identify roughly a third of real WCAG issues — they can't judge whether alt text is meaningful, whether focus order is logical, or whether an interaction works with a screen reader. Run the automated scan as a floor, then tab through the site with a keyboard yourself. With 4,928 accessibility lawsuits filed in 2025, the gap between “passes Lighthouse” and “actually accessible” is where the risk lives.