Should I Hire a Web Designer? Six Defects a Comp Cannot Contain

Hire a web designer when the decisions are unmade — when nobody can say what the site is for, who it is talking to, or which of three directions is right. Do not hire one to produce pictures of a site you have already decided on. That is the split, and it explains why a designer can be excellent and the resulting site still ship broken.

A comp is a picture. What ships is a system, and the gap between the two is where the defects live. This post measures that gap on a real site — this one — using the defects our own audits found on pages that looked completely finished.

The three jobs people mean by "web designer"

The question is unanswerable as asked because it bundles three purchases that have different prices, different deliverables, and different failure modes.

Most people asking "should I hire a web designer" have already made the direction decisions — they know they sell project management software to agencies — and are really shopping for composition. That is the job a template does, because composition is the part that generalises. Direction does not generalise, and implementation is a different skill than either.

The honest version of the advice: buy direction from a human when you don't have it, buy composition from a template when you do, and never assume either one gave you implementation.

What a comp cannot contain

Here is the concrete part. This site scores Lighthouse accessibility 100 on all eight measured pages, on desktop and mobile, with desktop performance and SEO also at 100. It is built from 91 components across five layers, on 38 colour tokens and 480 explicit dark-mode variants. By any reasonable reading it is a finished, professionally-executed design.

Audits of it have found, so far:

1. /blog had no <h1> at all. The index title rendered as an h2 because the shared heading component defaulted to one. Visually identical; the page's subject was simply never stated.
2. /docs skipped h1 → h3. The first section heading jumped a level. Again invisible — the classes made it look right.
3. BlogTag failed AA contrast by 0.01. text-primary (#465fff) on bg-primary-25 (#f2f7ff) at 12px measures 4.49:1. The threshold is 4.5:1.
4. The FAQ accordion had two of five WAI-ARIA requirements. On /pricing — one of the pages scoring 100.
5. The dashboard tabs had three of six.
6. All five form status messages were live regions that announced nothing, because each was mounted in the same render as its first message. That one is fixed in the post published alongside this one.

Not one of those six is visible in a comp. Three of them (the accordion, the tabs, the live regions) do not exist in a comp at all, because a comp has no time dimension — it cannot express "when this changes, this gets announced." The other three are semantic rather than visual: a comp shows a heading's size, not its level.

A designer is not being criticised here. A comp is doing exactly what a comp is for. The point is that "we hired a designer" answers none of these questions, and a project plan that treats the comp as the finish line has budgeted for none of the work that catches them.

The 0.01 problem

BlogTag deserves its own paragraph, because it is the cleanest example of why this is a systems question rather than a taste question.

The colour was #465fff on #f2f7ff. Both are from the project's own token scale — a designer picked them, they are on-brand, and they look fine. The ratio is 4.49:1 against a 4.5:1 requirement. Nobody eyeballs a contrast ratio to two decimal places. The only ways to catch it are to run a checker on every pair you ship, or to build the scale so that the failing combination cannot be expressed.

That is a design-system decision — and it is the kind of thing you are buying when you buy a template that already has one, or that you are commissioning when you hire someone to build one. It is not what an hourly composition engagement produces.

So when should you hire one?

Genuinely, in these cases:

The row that costs people the most is the second-to-last. "Accessible and fast" is frequently written into a design brief and then delivered as a picture of a site that would be accessible and fast if someone built it that way. Ask instead: who is running the audit, on which pages, against which standard, and what happens when it fails?

What to ask, whichever way you go

The same four questions work on a freelancer, an agency, and a template:

1. Show me the states. Not the hero — the error state, the empty state, the loading state, the 40-character-name state. Most of a real interface is states, and most comps contain one of them.
2. What is the type and colour scale, and what enforces it? A palette is a picture; a token file is a constraint. Ask what stops someone using the 4.49:1 pair.
3. Which pages have you measured, and at what widths? "Responsive" is a claim. We only found three of the defects above because a follow-up pass measured every main page and mobile, which had never been measured at all — the first eight-page audit had missed pages added later.
4. What happens on the second change? Cheap to build and expensive to change is the standard failure mode. Ask how a new section gets added six months from now, and by whom.

A template answers 1–4 by being inspectable before you buy: you can open the live preview, resize it, tab through it, and read the token file. That is a genuine advantage over a proposal, and it is worth saying plainly that it is also the limit of what a template answers — it cannot tell you whether the direction is right for your business.

Common mistakes

Frequently asked questions

Is a template just a worse version of hiring a designer? It is a different purchase. A template is the composition job, already done, inspectable before payment, at a fixed price. It cannot do the direction job — it does not know your market. If your positioning is settled, that missing half is not a loss. If it isn't, the template will faithfully render an unclear message.

What does a web designer actually cost? Ranges are meaningless without a scope, which is why the published ones span two orders of magnitude. The more useful framing is to price the artifacts you will receive rather than the pages — we work that through with real counts in website development cost, and the recurring half in what maintenance actually costs.

Can I hire a designer to customise a template? Yes, and it is often the best value in this whole post: the composition is settled, so the engagement is direction plus a brand pass, which is a fraction of the scope. The thing to protect is upgradeability — a rebrand that rewrites every component blocks you from ever merging the template's updates. The edit order that keeps both possible is here.

How do I check accessibility myself before paying anyone? Three minutes, no tools. Tab through the page and confirm you can always see where focus is. Zoom to 200% and check nothing is cut off. Then submit a form with a deliberate error and see whether the message is announced — that last one is the check that caught the sixth defect on this list, and it is the one nothing automated will do for you.

Templates for the composition half

ASoc Brief is a product-designer résumé and portfolio site — the template a designer would use for their own site, which is a reasonable proof that the composition holds up. ASoc Byte is an engineering-studio site built around outcomes and social proof. ASoc Canvas is a no-code page-builder marketing site, with the editor preview and pricing structure that category expects.

Browse the full sets: Next.js landing page templates, Tailwind landing page templates. If you're weighing the build itself rather than the design, template versus building from scratch is the same question one layer down.