Short answer: Evaluate a design agency on four things, in this order — who is actually doing the work, how they scope, how they handle disagreement, and what happens when the project goes wrong. Portfolio quality is table stakes and tells you less than people think, because every agency shows its three best projects. The single most useful step is a small paid trial: a two-to-three week UX audit or one discrete flow, with real money and a real deliverable. You will learn more from that than from six calls. The six disqualifying red flags are: refusing to name the working team, no screen inventory in the proposal, "unlimited revisions," no questions about your users, no documented handoff process, and claiming they have never had a project go badly.
Below: the eighteen questions, what strong and weak answers sound like for each, and the scoring framework we give clients who are evaluating us alongside competitors.
Why portfolios are the weakest signal
Every agency on your shortlist has a beautiful portfolio. That is what a portfolio is for. What it does not tell you:
- Whether the people who made that work still work there
- How long it took, and whether it was on budget
- How much of the direction came from the client
- Whether it shipped, and whether anyone used it
- What the agency did when the client changed their mind in week six
You should still look. Look for range of problem type rather than range of aesthetic — a studio that has designed a permissions model, a data-dense table, and an onboarding flow has solved harder problems than one showing ten beautiful landing pages. And look for states, not screens. Any case study that shows only happy-path screens is showing you a portfolio piece, not a product.
Then move on to the questions, which are where the actual information is.
The 18 questions
On the team
1. Who specifically will work on this project? Names, roles, and what percentage of their time?
Strong answer: Named people, their portfolios, their seniority, and a rough allocation. "Priya leads, roughly 60% of the design hours; Arjun on UI, 30%; I review weekly."
Weak answer: "Our senior design team." This almost always means the person on the sales call is not the person doing the work, and you will meet a different team at kickoff.
2. Will the people on this call be on the project?
Strong: A clear yes or a clear, explained no. "I'm the founder, I'll review weekly but won't be in the file daily — Priya will be your day-to-day."
Weak: Vagueness. This is the most common bait-and-switch in the industry.
3. What is your designer-to-project-manager ratio, and what percentage of my billed hours will be non-design?
Strong: A real number. 15–25% non-design is healthy.
Weak: Not knowing. An agency that cannot answer this does not track it, which tells you something about how they run projects.
4. What happens if the designer on my project leaves mid-engagement?
Strong: A described process — shared files, documented decisions, an overlap period. It has happened to them and they know what they did.
Weak: "That won't happen."
On scoping
5. How many screens do you think this needs, and how did you get to that number?
Strong: A screen inventory, even rough, with states counted separately. Pushback on your estimate — because your estimate is almost certainly low. This is the single most predictive question on the list; see UI/UX design agency pricing for why screen and state count dominates project cost.
Weak: Accepting your number without comment. They are either not thinking about it or planning to change-order you later.
6. What is in scope and what is explicitly out?
Strong: A written exclusions list. Good agencies write down what they are not doing.
Weak: Only an inclusions list.
7. What would you cut if the budget were 30% lower?
Strong: Immediate, specific, and it reveals their priorities. A good answer usually cuts breadth, not depth.
Weak: "We'd need to discuss." They should already know what the optional parts are.
8. What do you need from us, and how often?
Strong: Specific cadence and specific asks — a weekly 45-minute decision call, access to users, analytics access, a named decision-maker.
Weak: "Just feedback on the designs." Projects fail on the client side more often than the agency side, and an agency that has not thought about your obligations has not learned that yet.
On process and craft
9. Walk me through a project where the first direction was wrong. What did you do?
Strong: A real story with a real cost. Look for whether they noticed early or late.
Weak: Not having one. Everyone has one.
10. How do you decide between two directions you both like?
Strong: A decision mechanism — testing, a stated principle, a metric, or an explicit "the client decides and we advise." Any of those beats aesthetics.
Weak: "We present options and let the client choose." That is abdication dressed as service. You are paying for judgement.
11. What research do you do, and what happens if we say we cannot afford it?
Strong: A tiered answer, and honesty about the risk of skipping it.
Weak: Either "research is mandatory" (inflexible) or "we don't really do research" (they are decorating).
12. Show me a design system you have built and explain a decision inside it.
Strong: They can explain why a component has the variants it has, and what they deliberately left out. Systems thinking shows up here or nowhere. More on this in design systems for teams under 20 people.
Weak: A screenshot of a colour palette and a type scale.
13. How do you handle empty, loading, error and permission states?
Strong: As a standard part of the work, with a stated approach.
Weak: Surprise at the question. This is the clearest technical competence signal in the whole list and it takes one question.
14. What does handoff include, exactly?
Strong: Annotated specs, documented tokens, a developer walkthrough, and a support window. See the 2026 state of design handoff.
Weak: "We share the Figma file."
On commercials and risk
15. What are your revision terms, and what counts as a round?
Strong: A defined number per milestone, a clear definition of a round, and a stated rate for extra rounds.
Weak: "Unlimited revisions." This is either untrue or priced with a large buffer you are funding. It also removes the pressure that makes feedback specific.
16. Who owns the files, and what happens to them if we part ways?
Strong: You own the output on final payment, files live in a shared organisation, and you get everything on exit.
Weak: Anything ambiguous. Get this in writing.
17. What is your process when we are unhappy with the work?
Strong: A described escalation — a conversation, a re-brief, a defined remedy.
Weak: Reassurance instead of a process.
18. Tell me about a client you were a bad fit for.
Strong: A clear description of who they should not work with. An agency that knows its limits has limits.
Weak: "We work with everyone." Nobody works with everyone well.
The six red flags worth walking away over
1. They will not name the working team. The most reliable predictor of disappointment. The work will be done by people you have not assessed.
2. The proposal has phases but no screen inventory. "Phase 2: UI Design — $34,000" is a number, not a scope. You cannot verify it, compare it, or hold them to it.
3. "Unlimited revisions." Either a lie or a buffer. Both cost you.
4. They asked nothing about your users. If an entire sales process passes without a question about who uses the thing and what they are trying to do, they are selling screens.
5. No documented handoff process. Design that engineering cannot build is an expensive mood board. This is where most of the real waste in this industry lives.
6. They claim no project has ever gone badly. Everyone has had a bad project. An agency that will not tell you about one is either inexperienced or not being straight with you, and you will find out which during yours.
A seventh, softer flag: the price is dramatically below everyone else's. Sometimes this is a genuine cost-base difference — an India- or LatAm-based studio really is cheaper, and that is legitimate. But a quote at a third of the next lowest, from a studio in the same market, usually means they have scoped a different project than the one you described. Ask for the screen inventory and the gap will explain itself.
The paid trial: the step that actually works
Six discovery calls tell you how an agency sells. Two weeks of paid work tells you how it works.
What to buy: a UX audit of your existing product, or the design of one discrete flow. Both are small, both are useful regardless of who you hire, and both require the agency to demonstrate judgement rather than describe it.
What it costs: $2,500–$7,000 for an audit from an India-based studio, $12,000–$35,000 from a US one.
What to watch for, in order of importance:
- Did they ask good questions before starting? The quality of the questions predicts the quality of the work more reliably than anything else.
- Did they disagree with you? A partner who agrees with everything is not adding judgement.
- Did they hit the date? A two-week engagement that slips is a twelve-week engagement that slips by six.
- Was the deliverable usable without them in the room? Good documentation is a habit, not an effort.
- What did they do when you pushed back? Defend with reasoning, or fold immediately? Both are bad in different directions. You want reasoning, then adjustment.
Run the same trial with two agencies if the main engagement is large enough to justify it. The comparison is worth far more than the cost.
A scoring framework
If you are comparing three or more agencies, score each on these, 1–5:
| Criterion | Weight |
|---|---|
| Named, verified working team | 20% |
| Scoping rigour (screen inventory, exclusions) | 20% |
| Evidence of systems thinking | 15% |
| Handoff and documentation | 15% |
| Demonstrated judgement (disagreed with you) | 15% |
| Relevant domain experience | 10% |
| Portfolio aesthetics | 5% |
Yes, portfolio is 5%. That is deliberate, and it is the part clients push back on most. Aesthetic quality is largely a filter you have already applied by getting to a shortlist — everyone left can make things look good. The weights above are what separates the shortlist.
Domain experience gets 10% rather than more because it is overrated except in genuinely regulated or high-complexity spaces. For fintech or healthcare, raise it to 20% — the constraints are real and unfamiliarity is expensive. For a standard B2B SaaS product, a strong generalist team beats a weak specialist one comfortably.
Questions clients ask us about this process
How many agencies should I talk to? Three to five for a serious engagement. Beyond five you are comparing sales processes rather than capability, and the marginal call teaches you nothing.
Should I send the same brief to everyone? Yes, and include your budget range. Withholding budget does not get you a better price; it gets you proposals scoped for the wrong project, which wastes everyone's time. The useful information is what each agency does with the same budget.
Is a bigger agency safer? More redundancy, more process, more overhead, and you are more likely to get junior staff. A 10–40 person studio usually gives the best ratio of seniority to price. Under five people, you are effectively hiring a freelancer with extra steps — which may be exactly right. See in-house vs agency vs freelancer.
How much should references matter? Less than you think, because you get the three happiest clients. Ask instead for a reference from a project that was difficult. Whether they can produce one is more informative than what that reference says.
What if I cannot afford a paid trial? Then ask for a scoped critique of one screen of your existing product, free, in a call. Not a full audit — thirty minutes of them reacting live. You will still learn whether they think structurally or visually.
Related reading
- UI/UX design agency pricing in 2026
- In-house designer vs agency vs freelancer
- UX audit benchmarks from 50 real product audits
- Design systems for teams under 20 people
Published by Devoq Design. This checklist is the one we hand to prospective clients evaluating us against other studios, including the questions we find hardest to answer well.
Last updated: 14 September 2026