Choosing a UX/UI Agency: Checking Research, Prototypes and Case Studies

By Edip Kilinc, founder of Webwavers. Published October 31, 2026. Last updated October 31, 2026. Reading time: about 16 minutes.

Short answer: A good UX/UI agency shows you, before the first screen, how it justifies its decisions: with user research, a clarified information architecture and a clickable prototype that has been tested. In the portfolio you recognize this by case studies that describe the starting point, method, findings and result, not only finished interfaces. An agency that shows only screens shows design, but no process.

The key points

• UX/UI covers research, structure, interaction and visual design. Check each phase individually.
• Ask for case studies with method and findings. Pretty screens do not prove user research.
• A prototype is only useful once it has been tested with real users before development.
• For complex B2B applications, experience with roles, permissions, data volumes and multilingual use counts.
• Ask about the handoff: design system, documentation and collaboration with development.

Table of contents

• What a UX/UI agency does
• The process: which results you get per phase
• User research: how to recognize real research
• Information architecture
• Prototypes: what clickable means
• Checking the portfolio and case studies
• Checklist for checking a UX portfolio
• 12 questions for the first call
• Warning signs
• Special cases: product interfaces, portals, industry
• Handoff to development
• Accessibility from the start
• Putting effort and cost in context
• Agency, freelancer or in-house
• When you do not need a UX/UI agency
• How we do it
• FAQ

What a UX/UI agency does

UX stands for user experience, UI for user interface. The two terms are often used together and yet mean different things. In the glossary you find the definitions of UX and UI. In short: UX asks whether people reach their goal. UI designs the interface through which they try.

A UX/UI agency usually covers four areas:

1. Research. Understanding who the users are, what they want and where they fail.
2. Structure. Arranging content and functions so that people can find them. This is called information architecture.
3. Interaction. Designing flows, states and error cases, usually in wireframes and prototypes.
4. Visual design. Layout, typography, colors, components, often as a design system.

The catch: not every agency that calls itself this does all four areas with the same weight. Some are design studios at heart that wear UX as a label. Others are research-strong product teams with less design ambition. Your job in the selection is to find out where the strength lies and whether it fits your project.

The process: which results you get per phase

The process is a good test, because every phase should have a tangible result. If an agency cannot name artifacts for you, they probably do not exist.

Phase	What happens	Tangible result	Proof in the portfolio
Discovery	Goals, stakeholders, existing-state analysis, analytics	Target picture, problem hypotheses, list of assumptions	Case study names the initial problem and goals
Research	Interviews, observation, surveys, existing data	Findings report, personas or jobs-to-be-done	Quotes, method, number of participants
Structure	Card sorting, sitemap, user flows	Information architecture, flow diagrams	Diagrams or sitemap in the case
Concept	Wireframes, interaction logic	Low-fidelity drafts, states and error cases	Wireframes next to the final design
Prototype	Clickable draft, usually in Figma	Testable prototype	Prototype link or video
Test	Usability test with real users	Test report, prioritized problems	Changes that trace back to tests
UI design	Visual design, components	Design system, final screens	Components and states
Handoff	Documentation, collaboration with development	Specifications, designs with states, review during implementation	Mention of the collaboration

If you know that your project does not need research (for example, because you already have data), the agency can shorten phases. It should then explain what it uses as a basis instead.

User research: how to recognize real research

User research is where agencies differ most. "We analyzed your target group" can mean googling or thirty interviews. Ask.

The Nielsen Norman Group arranges UX methods along three axes: what people say compared to what they do (attitudinal and behavioral), qualitative and quantitative data, and the context of use. The group lists 20 methods, including usability tests, field studies, interviews, card sorting, tree testing, analytics and A/B tests (NN/g, Christian Rohrer, 2022, last checked 2026). The takeaway for your selection: a good agency combines methods and explains why it chooses which.

How to recognize real research work:

• It names method, number of participants and time frame, not just "user research."
• It separates observation from opinion. What people say and what they do are rarely the same.
• It shows raw material: anonymized quotes, observation notes, analyses.
• It explains how findings fed into design decisions. "Because six of eight participants overlooked the filter, we moved it to the top" is proof. "We understood the users" is not.
• It has a plan for access to users. In B2B especially, this is the biggest hurdle: who organizes interview partners, and how many can realistically be reached?

A note on the size of tests: in 2000, Jakob Nielsen argued that five test participants already uncover roughly 85% of usability problems, and that several small tests are better than one large one (Nielsen Norman Group, 2000). That applies to qualitative tests aimed at finding problems, not to figures you want to back up statistically. If an agency proposes "five users per round," that is a solid basis. If it claims to deliver representative statements with that, it is not.

You can find more on interviews in the glossary under user interview.

Information architecture

Information architecture (IA) decides whether people find what they are looking for. It covers navigation, labels, grouping of content and the hierarchy between pages or screens. Especially for websites with a lot of content, for portals and for software with many functions, it matters more than any visual detail.

Two methods that show real IA work: in card sorting, participants sort labeled cards into groups that make sense to them. This shows how users mentally organize content. Tree testing, by contrast, checks an existing structure by having participants solve tasks in a text-only tree (NN/g: Card Sorting, 2024). If an agency does not propose card sorting or tree testing before the redesign of a large website, ask how else it validates the structure.

What you can expect as a result: a sitemap or a screen flow, a rationale for labels and groups, and a description of what changes compared to before and why. This is relevant for websites too. More on this in the article Page structure of a company website.

Prototypes: what clickable means

A prototype is a draft you can try out before it is built. The glossary explains the term under prototype. The level of detail is what matters, because "prototype" can mean many things.

Level	Look	Suitable for	Warning
Low fidelity	Gray, boxes, placeholder text	Testing structure and flow early	Looks unfinished, users focus on layout instead of content
Mid fidelity	Real text, simple design	Testing content and interactions	Often the best compromise for tests
High fidelity	Almost final look, real interactions	Convincing stakeholders, final tests	Expensive, changes cost time

What you should ask for: a prototype that is actually clickable and maps the central user paths, not just an image gallery with hotspots. And even more important: it should be tested with real users before development. A prototype that only the project team has seen is a nice document, but no protection against risk.

Ask about edge cases too. What does the empty state look like, the error state, the loading state? In software and portals, these are the states where users get stuck most often. If the prototype only covers the happy path, only half of it was designed.

Checking the portfolio and case studies

The portfolio is the most important public proof. But it only tells you something if you read it correctly. Many portfolios are galleries: beautiful screens, a project name, a logo. From that you learn that the agency can design. You do not learn how it arrives at decisions.

A reliable case study has these building blocks:

1. Starting point. What was the problem, for whom, with what goal?
2. Role of the agency. What exactly did it do, and what did the client's team do?
3. Method. What research, what tests, with how many participants?
4. Findings. What emerged, what was surprising?
5. Decisions. Which changes resulted, with before and after?
6. Result. What changed, by which metric and over what period?
7. Limits. What did not work? A case study without a single "we did that differently" is mostly advertising.

Note: not every project may publicly state figures. Many B2B clients forbid it. In that case the agency should be able to show in conversation what it would measure and what it measured in similar projects. Figures that appear in a case but name no metric and no time period are decoration.

Check relevance. A strong portfolio for consumer apps says little about a dealer portal. Look for experience with your context: complex B2B applications, multilingual platforms, customer portals, product presentation for products that need explanation. Whoever shows you a similar project can also explain where it got critical.

A note on awards: they say something about design quality and perception, but little about research and impact. More on this in the guide Design awards when choosing an agency.

Checklist for checking a UX portfolio

Go through two to three case studies per agency and tick off:

1. The initial problem and goal come first, not the result.
2. The agency's role is clearly delineated.
3. Methods are named (interviews, observation, analytics, card sorting, usability test).
4. The number of participants and the target group of the research are stated.
5. Raw material is visible: quotes, notes, analyses, diagrams.
6. Information architecture is documented (sitemap, flow, rationale).
7. Wireframes or intermediate stages are shown, not just the end result.
8. There is a clickable prototype or a video of it.
9. At least one test with users is described, including the changes derived from it.
10. States are visible (empty, error, loading), not just the ideal path.
11. Results name a metric and time period, or explain why no figures are given.
12. There are references to accessibility and to the handoff to development.
13. A project in your context (industry, complexity, user roles) is included.
14. The agency mentions what it would do differently in hindsight.

As a rough rule of thumb (our own assessment, not a study): an agency that meets nine or more of these works with a recognizable process, and one that meets four or fewer mainly shows looks.

12 questions for the first call

The call is the moment where portfolio and reality meet. These questions separate preparation from routine.

1. What does your approach look like if we start from zero? What results do I get per phase?
2. How do you handle user research if we ourselves have no access to end users?
3. How many interviews or tests do you typically plan, and what do you base that on?
4. How do you validate the information architecture before you design it visually?
5. How do you test the prototype, and what happens with the results?
6. What data (analytics, support tickets, heatmaps) do you evaluate before you develop new ideas?
7. Show me a project where research disproved your assumption. What did you change?
8. How do you work with development, internal or external?
9. What do you deliver at handoff: design system, specification, components, review during implementation?
10. How do you handle accessibility, and where in the process do you check it?
11. Who from your team works on my project, and who is my fixed point of contact?
12. What happens after launch: measurement, iteration, support?

Good agencies answer concretely and with examples. Evasive answers like "We do that individually" are acceptable if they are followed by an example.

Warning signs

Design before research. If the offer has final screens after four weeks and users are never mentioned, the work is design, not research.

Prototype without a test. It helps alignment in the team, but it does not reduce risk.

Personas that are just wishful thinking. Personas without a data basis are inventions. Ask where they come from.

One-size-fits-all phases. The offer looks the same for every project, regardless of complexity and user group.

Numbers without context. "Plus 120 percent conversion" without a metric, a time period and a baseline value.

No answer to "What went wrong?" Every project has trade-offs. An agency that cannot name them has not reflected.

Handoff without involvement. If design and development run strictly separately and nobody accompanies implementation, details get lost.

Special cases: product interfaces, portals, industry

UX requirements differ greatly depending on the context. Three cases where the selection needs particular attention.

Product interfaces in IT and software. This is about workflows, data volumes, roles and permissions, states and errors. Good agencies can explain how they reduce complexity without hiding functions that power users need. You can find more context on the page IT, software and SaaS. Marketing sites for software have their own patterns, for example the hero that explains software.

Customer portals and dealer platforms. What matters is task orientation, clear status displays, search and filters, several roles per account and often multilingual use. The guide Customer portal redesign with UX goes into this in more detail.

Industry and engineering. The focus here is often on presenting products that need explanation: variants, technical data, certificates, configurators. The goal is that a buyer or technician reaches the right variant without a phone call. More on this on the page Industry and engineering and in the guide Datasheets and certificates on industrial websites.

In all three cases, what counts is: does the agency have experience with your kind of complexity? A consumer onboarding is something completely different from a permissions model with seven roles.

Handoff to development

Many UX/UI projects lose quality in the handoff. The design is clean, the implementation deviates, and nobody notices early enough. So ask about the handoff concept before signing the contract.

What should be part of it:

• Design system. Reusable components, colors, typography, spacing, states. This reduces deviations and speeds up later changes. The guide Design systems: Figma to Webflow handoff describes the path in detail.
• Specification. Behavior at different screen sizes, states, animations, input validation.
• Documentation of decisions. Why does something look the way it does? This saves follow-up questions in development.
• Support during implementation. Design review at intermediate stages, not just acceptance at the end.

When design and development come from different houses, the interface is the biggest source of errors. Why many companies therefore award both together is covered in the guide Design and development from one team. There are also good reasons to separate them, for example when the development team is in the company.

Accessibility from the start

Accessibility belongs in the concept, not in the final check. Contrast, focus order, labels, target sizes and understandable error messages are design decisions. Depending on the offering, the Accessibility Strengthening Act (Barrierefreiheitsstärkungsgesetz, BFSG) has applied in Germany since June 28, 2025. You can find an overview in the article BFSG and your website and in the glossary under accessibility.

Ask the agency where in the process it checks accessibility, which standards it works with and whether it plans tests with users with disabilities. An agency that sells accessibility only as a check at the end becomes expensive when errors turn up.

Putting effort and cost in context

For UX services alone, such as research, information architecture and prototypes, there is no reliable, independent market range. The field is too uneven: a two-week audit and a three-month product design project cannot be captured in one number. Whoever gives you a flat figure without a scope is guessing.

What you can put in context are hourly rates. The website cost report by adfera lists €80 to €120 per hour for agencies in web work and €60 to €90 for freelancers (adfera, 2026). Tricks.de lists an average of about €120 for advertising agencies (Tricks.de, 2026). Specialized research and product design teams can charge more. What matters is what is in the hours.

How to compare quotes:

• Ask for a breakdown by phase, not just a total.
• Check whether research participants, incentives and tools are included.
• Ask how many feedback rounds and test rounds are included.
• Clarify whether the handoff to development and design reviews are included.
• Ask about options: a compact start with an audit and prototype that can be expanded later is often more sensible than one large overall package.

General notes on reading quotes are given in the guide Comparing web agency quotes.

Agency, freelancer or in-house

Not every UX project belongs in an agency. The decision depends on scope, duration and existing knowledge.

Option	Strength	Limit	Fits
Agency	Several disciplines, variety of methods, capacity	Less context about your product, project end	New direction, relaunch, clearly defined projects
Freelancer	Flexible, direct line, often cheaper per hour	One person, cover and capacity limited	Focused tasks, supplementing a team
In-house	Deep product knowledge, permanent availability	Building the team takes time, methods know-how has to be there	Products under continuous development
Hybrid model	Strategy external, maintenance in-house	Needs clear roles	Growing teams

You can find a detailed comparison in the guide UX: agency, freelancer or in-house.

When you do not need a UX/UI agency

There are situations in which an agency is not the best choice:

• Small adjustments. For a manageable improvement to an existing page, an analysis of the data, a few tests and an experienced designer are often enough.
• Missing data basis. If you do not yet know whether the product is needed, market validation helps more than interface design.
• Strong in-house team. If you already have research and design in the company, support at specific points makes more sense than a complete project.
• Standard requirements. A marketing website with clear goals often gets by with a good structure and an experienced web designer.

Tools help too: heatmaps and A/B tests give inexpensive hints about where users get stuck. They do not replace qualitative research, but they show where research is worthwhile.

How we do it

Webwavers is a premium agency for branding, web design and web development. UX/UI design is one of our services, which you can find on the page UX/UI design. You can ask us the questions above just as you would ask any other agency. If you are looking for design and implementation together, we work with Webflow as our focus and advise neutrally on the choice of system. How we set up projects is described on the page Process: every client gets a dedicated project lead and their own Slack channel, and every phase ends with a feedback round and approval. On the page UX/UI design, the five phases of start, structure, design, system and handoff are described. We use user tests when the project needs them, and often existing data and a clickable prototype are enough.

FAQ

How can I tell whether a UX agency does real user research? By the portfolio and the conversation. A reliable case study names the method, number of participants, findings and the decisions derived from them. In the conversation, the agency shows raw material and can name a project where research disproved its assumption.

What is the difference between a wireframe and a prototype? A wireframe shows structure and arrangement, usually gray and static. A prototype is clickable and maps flows, so that users can try it out. For tests you need a prototype, for early alignment wireframes are enough. Both terms are explained in the glossary.

How many users do I have to test? For qualitative tests aimed at finding problems, the Nielsen Norman Group argues that about five participants uncover roughly 85% of the problems, and that several small tests are better than one large one. For statistically reliable statements, you need significantly more participants and different methods.

Should the UX agency also develop? Not necessarily. Design and development from one team reduces friction at handoff. A separation makes sense if a development team is in the company. What matters is that the handoff is planned: design system, documentation and review during implementation.

What does UX/UI design cost at an agency? For UX services alone there is no reliable, uniform market range, because scope and depth vary greatly. Hourly rates give some orientation: according to adfera, €80 to €120 at agencies and €60 to €90 at freelancers in web work. Ask for a breakdown by phase.

How do I check whether an agency can handle complex B2B applications? Have them show you a project of similar complexity: several roles, a lot of data, permissions, multilingual use. Ask how findings from interviews fed into the structure and how the agency handled edge cases such as empty states and errors.

Do I need a UX audit before a relaunch? Often yes. An audit based on analytics, heatmaps and user observation shows which problems a redesign has to solve and which it does not. This prevents a new design from merely repackaging old weaknesses. Ask the agency whether it offers an audit as a separate, smaller step.

Conclusion

You recognize a good UX/UI agency by process and proof: research with a named method, validated information architecture, tested prototypes, documented handoff. In the portfolio, case studies with a starting point, findings and limits count for more than screens. Ask the twelve questions, use the checklist and have examples from your context shown to you. That way you compare agencies by what they do, not by what they promise.

Support with the selection: If you are planning a UX audit or a UX/UI project, you can reach out to us via the contact page.

Sources

1. Nielsen Norman Group: When to Use Which User-Experience Research Methods (Rohrer, 2022)
2. Nielsen Norman Group: Why You Only Need to Test with 5 Users (Nielsen, 2000)
3. Nielsen Norman Group: Card Sorting, Definition (Tankala and Sherwin, 2024)
4. adfera: Website-Kosten-Report 2026 (website cost report 2026)
5. Tricks.de: Website-Agentur Kosten und Preise (2026) (website agency costs and prices)