What You Should Be Asking Your Design Agency (But Aren’t)

Choosing a design agency is mostly done on feel. You look at the work, you meet the people, you compare the numbers at the bottom of the proposal, and you go with whoever seems most likely to get it. That is not unreasonable — but none of it tells you much about how the finished site will perform.

These are the questions we think clients should be asking. We include ourselves in that. Every one of them commits the agency to something specific, which is the only reason they are worth asking.

Why the portfolio tells you less than you think

Everyone starts with the work, and the work is always screenshots. The trouble is that a screenshot can only show you taste, and taste is not the variable that determines whether a website earns anything.

There is a further wrinkle. Tuch and colleagues, in the International Journal of Human-Computer Studies in 2012, tested 119 real websites and found that low visual complexity and high prototypicality — pages that look like what people already expect that kind of page to look like — produced the most positive first impressions, measurable at 17 milliseconds. Conventional beats inventive on first impressions. So a portfolio full of striking, unusual work is not straightforwardly evidence of a better outcome, and a portfolio of calm, familiar-looking sites is not evidence of a duller one.

Which is why the useful questions are all about what happens after the picture is approved.

The seven questions

1. Who will use this before it launches, and how many of them?

The single most useful thing anyone can do to a website before launch is watch a stranger try to use it. It is also the thing most likely to be missing from a proposal.

Jakob Nielsen of Nielsen Norman Group argued in March 2000 that a first study with five participants finds 85% of the usability problems in a design, and that you need at least 15 users to find all of them. Five is not a big ask. If the answer to this question is “our team will review it thoroughly”, that is not a test — everyone on the team already knows where everything is.

2. What will the first screen say, and who decided?

Ask to see the actual words that will appear at the top of the home page, and ask what evidence sits behind them. If the answer is that copy comes later, or that you will supply it, you have just found out who is really responsible for the most important part of the site.

It matters more than the layout around it. Writing in September 2011 on an analysis of more than 205,000 pages, Nielsen Norman Group concluded that users often leave within 10-20 seconds, and that to hold attention beyond that you must communicate your value proposition within 10 seconds. Whoever writes that sentence is doing the highest-leverage work on the project.

3. What performance target are you designing to?

This is the question that separates a design that will be fast from one that will have to be made fast later, expensively, against a layout everyone has already fallen in love with.

There is a published pass mark, so there is no excuse for vagueness. Google’s Core Web Vitals guidance asks for a Largest Contentful Paint within 2.5 seconds, an Interaction to Next Paint of 200 milliseconds or less, and a Cumulative Layout Shift of 0.1 or less, all at the 75th percentile of real page loads. And the value of hitting it is documented: in Milliseconds Make Millions, commissioned by Google and run by 55 and Deloitte across 37 brands, a 0.1 second mobile speed improvement moved 3.2% more retail shoppers from listing to product page and 9.1% more from product page to basket, with retail spend up 9.2%. Ask for a number in the contract.

4. What accessibility standard will this meet, and how will you show me?

Usually nobody has asked, so nobody has answered, and the site inherits whatever the theme happened to do about contrast, form labels and keyboard navigation.

It is close to universal. The WebAIM Million survey for 2026 found detected WCAG 2 failures on 95.9% of the top one million home pages, averaging 56.1 errors per page — and automated tools only find part of the problem. Name WCAG 2.2 AA, ask how it will be tested, and ask who fixes what the test finds.

5. What am I actually buying, and can I change it without you?

“Custom” covers an enormous range, from a genuinely bespoke build to a commercial theme with your colours in it. Neither is wrong. Not knowing which you bought is.

Templated platforms are now the default rather than the exception — W3Techs put WordPress alone on 40.7% of all websites and 58.9% of the CMS market when it surveyed in September 2026. So ask directly: which parts are third-party, what licences renew, what happens to them if we part company, and which pages can our own team edit without raising a ticket.

6. What happens after launch, and what does that cost?

Launch is the point at which most proposals stop and most websites start. Ask what the first three months look like, what is included, what is billed, who is responsible if something breaks at eight on a Sunday evening, and — crucially — what would trigger a change to the site rather than an addition to it.

7. What will you show me that is not a picture?

The best question of the seven, because it exposes the whole problem in one move. Designs are approved as images. Nearly everything that determines whether a site works cannot be seen in an image.

An agency that can produce those three things before launch is running a different kind of project from one that produces a set of layouts and a launch date.

What a bad answer sounds like

None of these are dishonest. They are simply answers that commit to nothing, and they are worth recognising:

• "We follow a proven process" — with no named stage, deliverable or date
• "We're very data-driven" — with no metric named and no baseline taken
• "That's a phase-two consideration" — usually said about speed or accessibility
• "We'll optimise it after launch" — the cost of which was not in the quote
• "Our clients love the designs" — an answer to a question about outcomes
• "We'll need your copy by then" — the proposition quietly becoming your problem

A good answer names a thing, a person and a date. It does not have to be a yes. “We do not do accessibility testing in-house, we bring in a specialist, here is what that adds” is a perfectly good answer — it is checkable, and now you can decide.

The question to ask yourself first

Before any of it: what is this website for, and how will we know if it worked?

If the honest answer is that the current site looks dated, say so plainly — that is a legitimate reason to spend money and it leads to a very different brief from “we need more enquiries from architects in the South East”. Most disappointing website projects are not badly executed. They are well-executed answers to a question nobody wrote down.