Most websites are designed from a guess. The first meeting should end with decisions.

---

QUICK ANSWER

A first meeting with a web agency is not a chat about taste. Every question should settle one decision the designer would otherwise guess, and you should leave with a plan you keep: pages by phase, a calendar worked back from launch, and what you owe by when. Below: the questions, and what each decides.

---

The owner of a four shop bakery meets a web agency. Asked what look she likes, she says modern, clean, warm, professional. Three weeks later, a design arrives.

It is modern, clean and professional. It is also wrong. The tone reads like a bank, and the one thing her customers do on the site, order a cake for Saturday, sits three clicks deep. The designer had four adjectives and filled in the rest.

A brief written in adjectives is a guess

Nobody asks for a website that is old, cluttered and confusing, so modern, clean and professional carry almost no information. A law firm and a surf school can both mean "modern" and want opposite things. The designer's own taste fills the gap, and the client first meets it in a finished design.

Guesses are paid for in revision rounds

A wrong guess comes back as feedback: make it warmer, it does not feel like us. Each round costs days, and trust. By the third, the client suspects the agency is not listening. Usually the client knew exactly what they wanted. They were never asked in a way that drew it out.

Elevates' own contract caps revisions at two rounds per page. A cap protects both sides, but it only works if the first design is close. The fix is upstream: better questions before anything is designed.

Every question should change something the designer does

If an answer would change nothing the designer does next, cut the question. "What is the one thing a visitor must do on your site?" passes: it sets the main button on every page.

Elevates built its discovery session around exactly this rule. Every question in it is paired with the decision it unlocks, and a question that unlocks nothing does not survive. The figure shows a few of them. None needs technical knowledge to answer, and each one removes a guess.

Some of the most valuable questions sound almost too simple:

• Who is the one person who approves? Not the committee, the person. Projects stall when approval is shared by everyone and owned by no one.
• Which languages, and which one is the default? This decides how the site is built underneath. Adding a language later costs more than planning for it now.
• Who will update the site, and how often? A site updated daily by a shop manager needs a different system from one updated twice a year by a founder.
• What already exists? A logo, photographs, texts, guidelines. What is missing is where the timeline risk is.

Ask about the business in words. Ask about the look with pictures

Here is the part most first meetings get backwards. They ask about the look in words ("what style do you like?") and about the business in pictures (a mood board of other companies' sites).

Clients cannot describe a look in words. Almost nobody can. But every client can point. So the look should be settled by choosing: between pairs of layouts, between palettes, between typefaces, each rendered with the client's own name on it. The business, meanwhile, is settled in words, because only the client knows it: who buys, what they buy, what success looks like in six months.

Elevates' session follows that split. Business questions are asked. Aesthetic questions are answered by picking. The next article in this series explains why pointing works so much better than describing, and what the research on first impressions says about it.

Rank the features before anyone prices them

Every website idea arrives with a wish list: online booking, a blog, a members' area, a shop, a newsletter, three languages. If the list is priced as one block, it is either too expensive or quietly trimmed by the agency, and the client finds out what was cut at launch.

A better way is to put every feature on a card and sort them together into three columns: must have for launch, later, not needed. The client does the sorting, because only the client knows which feature earns money on day one.

That board then becomes the sitemap and the phase plan. What is in the first column is the first version of the site. What is in the second is the next phase, with its own price. Nothing is hidden, and nothing is cut without the client having cut it.

Price, by the way, does not belong in that meeting. A budget said aloud in the first conversation anchors everything that follows, and scope gets squeezed to fit a number before anyone knows what the number should buy. The scope comes first, from the board. The price follows in writing.

Start from the launch date and work backwards

"When would you like it live?" is usually answered with "as soon as possible", which is not a date.

The better question is what drives the date: a season, a trade fair, an opening, a funding round, a new product. Once there is a real date and a reason for it, the whole plan can be worked backwards from it: when design starts, when the first design is shown, when it must be approved, when the texts and photographs must be delivered, when the site is checked before launch.

Worked backwards, the calendar reveals the risk on the first day instead of the last week. A long list of must have features against a launch six weeks away is visible immediately. So is a content deadline that falls in the client's busiest month. Both can be solved in the meeting. Neither can be solved in the week before launch.

The meeting should end with a plan you keep

Most first meetings end with "we will send you a proposal". The client leaves with nothing but a feeling.

A discovery session should end differently: with a route the client can see before the call ends and keep afterwards.

That route holds the sitemap, with each page marked for the first version or a later phase. It holds the phases, weighted in weeks. It holds the calendar worked back from the launch date. And it holds the part clients are rarely shown: a list of what the agency needs from them, each line with an owner and a date. Photographs by this date. Texts by that one. Access to the domain by the end of the week.

That list is where most delays really come from, and writing it down with names is what stops them. It also makes something visible that is usually hidden: a website is built by two teams, not one.

What Elevates built to run this

The session described here is not theory. Elevates built a tool for it, called Atlas, and ran its first live client session in September 2026.

It runs for about thirty minutes, on a shared screen or a single tablet, in the client's own language. The business questions are asked. The look is settled by rounds of choosing between pairs. Answers save as they are given. And the last screen is not a summary: it is the route, with the client's own site sketched on desktop and phone, the pages by phase, the calendar back from launch, and the list of what is owed by whom.

Nothing in that route is invented. Every line traces back to something the client said or chose, or it does not appear. That is the practical meaning of listening: not nodding in a meeting, but being able to show, line by line, where each decision came from.

What to ask of any agency before the first meeting

Whoever you hire, you can test whether they will listen before you sign anything.

• Ask what the first meeting will decide. A good answer names decisions: the main action, the pages, the phases, the launch date. A weak answer says "we will get to know you".
• Ask how they will settle the look. If the answer is "tell us what you like", expect revisions. If it is "you will choose between options", expect fewer.
• Ask whether you will leave with anything. A plan, a sitemap, a list of what they need from you. If you leave with nothing, the proposal will be their guess, written neatly.
• Ask who approves on their side, and who on yours. Two named people save weeks.
• Ask what they need from you, and when. An agency that cannot answer this yet has not thought about your timeline.
• Ask how many revision rounds are included, in writing. Then ask what they do to make sure the first design is close.

These questions cost nothing. They work on any agency, including Elevates. An agency worth hiring will be glad you asked them.

Where good questions stop

Good questions remove guesses. They do not replace judgement.

A client can choose between two layouts; deciding what the third one should be is design. A client can rank features; knowing that a booking system on one platform will cost twice as much to change later is engineering. A client can name a launch date; knowing that a date six weeks away and a list of twelve features cannot both hold is experience.

So the session is only half of it. The other half is what the agency does with the answers: whether it turns them into a design the client recognises as theirs on the first showing, and a build that holds up when the business changes.

What a guess costs you, and what Elevates does instead

For a business commissioning a site, the cost of a guessed brief is rarely the first invoice. It is the weeks of revision rounds, the launch that slips past the season it was meant for, and a site that looks finished but does not do the one thing customers came for. Then it is the second project, a year later, to fix what the first one assumed.

There are two ways this goes. Many businesses already know exactly what they need, and a short, structured conversation with any capable team gets it right the first time. If that is you, use the questions above and hold your agency to them. Others have never had their business drawn out of them in a way a designer can use, and that is where the session earns its place. Elevates runs it before design starts, so the first design you see is built from what you said and chose, and the plan you keep says who does what, by when.

The question to ask this week

If you are about to brief an agency, ask them one thing: what will I have in my hands when our first meeting ends?

If the answer is a plan with dates and names, they intend to listen. If the answer is a proposal next week, they intend to guess, and you will pay for the guess in revisions.