Step 1: Name the use case
Four different kinds of work get sold under the single word “design”, and the agencies that do each of them best are frequently not the same agencies. Before you compare anyone, decide which of the four you are buying. This site ranks agencies separately for each, because a strong result in one says little about the others.
The four use cases
Website design is the marketing site or corporate web presence: the pages that explain what you sell to a buying committee and turn interest into a conversation. The output is pages, templates and a content system your marketing team can operate without a developer. Success is measured in pipeline and in how quickly you can publish a new page. See the B2B website design agencies ranking.
Product design is the software your customers use after they have signed: SaaS applications, platforms and portals. The output is workflows, every screen state, and a component system your engineers build from. Success is measured in task completion, adoption and the volume of support requests. See the B2B product design agencies ranking.
Branding is identity, rebrand and messaging: positioning, naming, a visual system and the guidelines that let other people apply it. The output is a set of decisions plus the assets that express them. Success is measured in recognition, consistency and whether your sales team stops rewriting the story on every call. See the B2B branding agencies ranking.
AI product design is AI features, copilots and AI-native products. It is product design with an extra problem: the system is sometimes wrong, and the interface has to make that safe. The output covers confidence, review, correction, hand-off to a human and failure states, in addition to the usual flows. See the B2B AI product design agencies ranking.
What goes wrong when it is misidentified
Hiring a branding studio to fix a product produces a beautiful interface that people cannot finish a task in, because the studio optimised for impression rather than for the fourth click of a workflow. Hiring a product studio to build a marketing site produces something dense and logical that does not persuade anybody, because it treated a page as a screen and a visitor as a user who already wants to be there. Treating AI as a visual layer, a chat box bolted to an existing product, produces a feature that works in the demo and gets switched off in month two, because nobody designed what happens when the answer is wrong.
A subtler failure is the bundled brief: “brand, website and product, one agency”. A few agencies are genuinely strong across two of these. Very few are strong across three, and a bundle usually means you are buying the average of their capabilities and the maximum of their fees.
A ten-second test
Finish this sentence: “After this engagement, [who] can [do what] that they cannot do today.” If the verb is understand, recognise or believe, you are buying website or branding work. If it is complete, configure or approve, you are buying product design. If it is trust, correct or delegate, you are buying AI product design. If the sentence needs two different verbs, pick the one where failure costs you most and treat the other as a scoped add-on, not a second lead.
If you are an early-stage company with no product in market yet, the question changes shape, and the guide for B2B startups works through it by funding stage.
Step 2: Set a budget from real ranges
Set the number before you talk to anyone, and set it from evidence. Across the 18 agencies on this site, hourly rates run from $50 to $400, and stated minimum engagements from $10,000 to $300,000. Those two figures behave very differently, and the second one filters harder.
Start with the minimum, not the rate
A minimum engagement is the smallest piece of work the agency will accept. If your budget is below it, the conversation ends, whatever the hourly rate. All 18 agencies on this list state a minimum, and those minimums fall into three bands:
- Below $100,000: five agencies. UX Studio has the lowest floor at $10,000, Fuselab Creative, Blink UX and Praxent start at $25,000, and Clearleft at $60,000.
- $100,000 to $150,000: six agencies. AREA 17 starts at $100,000, Doberman and Artefact at $120,000, and Instrument, Fantasy and TEAGUE at $150,000.
- $200,000 to $300,000: seven agencies. Bounteous and Pentagram start at $200,000, Work & Co, Huge, IDEO and Rightpoint at $250,000, and AKQA at $300,000.
Thirteen of the 18 agencies set their minimum at $100,000 or above. A budget under $100,000 still reaches five agencies on this list, but most of the market is closed to it, and it’s better to know that before you write a brief.
What the rate tells you
The hourly range tells you what a change costs once work is under way, and it lets you convert a budget into capacity. Arithmetic is enough: $150,000 at $200 an hour is 750 hours, roughly nineteen person-weeks. $50,000 at $150 an hour is about 333 hours, or eight person-weeks. $15,000 at $50 an hour is 300 hours, a little under eight. Ask any agency to translate its proposal into weeks and people, and the number stops being abstract.
Across the four use cases the listed ranges look like this:
- Website design: $100 to $350 an hour, minimums $25,000 to $250,000.
- Product design: $50 to $280 an hour, minimums $10,000 to $250,000.
- Branding: $170 to $400 an hour, minimums $120,000 to $300,000.
- AI product design: $100 to $400 an hour, minimums $25,000 to $250,000.
Product design has the widest spread and the lowest floor, because it includes the studios priced for smaller companies. Website and AI work start at $25,000 but sit mostly above $100,000. Branding has the highest ceiling, driven by the most established identity practices.
Price is not a proxy for quality
A higher rate does not reliably buy a better result. Compare the rates in the table below on this page with the composites on the ranking pages: agencies at very different rates sit close together, and the difference between them tends to be in specific criteria rather than across the board. A larger agency buys you capacity, a multi-market launch date and delivery certainty. It does not necessarily buy better design on a single product.
What sits outside the quote
The design fee is one line of the real cost. Budget separately for engineering time to build what is designed, participant recruitment and incentives for research, content and copywriting, licences for fonts and imagery, and your own team’s hours in workshops and reviews. Ask each agency what it expects you to supply and what it assumes is included. The answers differ more than the rates do.
If your budget is under the bands above, or you are pre-revenue, go to the startup routing guide before you go further.
Step 3: Score against six criteria
Every agency on this site gets a score out of ten on each of six criteria. The composite is the weighted average of those six scores. It is the main input to each ranking, with ties broken alphabetically, but the order also reflects our editorial judgement of fit for each use case. Each ranking page shows the six scores alongside the composite, so you can see where an agency earned its place and where it is weaker.
The weights reflect how B2B work fails. A B2B website is read by a buying committee rather than one visitor. A B2B product is used every day by admins, approvers and analysts who did not choose it. And the client usually keeps the work for years after the agency has gone. For that reason, three criteria share the top weight: how the work performs, how it is structured, and how it holds up after handoff.
Live product quality: 20 percent
This criterion asks how the agency’s released work performs for the people who rely on it. For a website, that means the pages a buyer actually opens: pricing, product and contact. For a product, it means the logged-in application. For branding, it means the identity as it appears in sales decks, on the website and at events. For AI work, it means the feature in production, including what it does with vague or incomplete input. It is a top-weighted criterion because you can judge it yourself without trusting anyone’s account of a project.
Check it: Ask for access to a demo environment, or a recorded walkthrough of a live product, using an ordinary user role rather than an admin account. For a website, submit the contact form on a phone and run the homepage through a free speed test such as PageSpeed Insights.
UX structural depth: 20 percent
B2B work is structurally heavy. Products have roles, permissions, multi-step approvals and dense data tables. Websites have solution pages for several industries, several buyer roles and several product lines. This criterion asks whether that structure was designed deliberately or simply accumulated over time. It carries a top weight because structural problems are the most expensive to fix later.
Check it: Find a past client’s site with pages for more than one industry, and see whether those pages follow the same logic. In the pitch, ask how the agency would add a new user role to a product, or a new product line to a website, without reorganising everything that already exists.
Research evidence: 15 percent
B2B users are hard to reach. Procurement managers, clinicians and finance controllers are busy and specialised, and some work in regulated settings. Agencies often fill the gap with your sales team and internal stakeholders. This criterion rewards research with real external participants, and methods documented well enough for someone else to examine.
Check it: Ask how the agency recruited a hard-to-reach role on a past project, and how many of the participants were outside the client company. Ask which design decision the findings changed.
Accessibility, tested: 15 percent
In B2B buying, accessibility increasingly comes up in procurement. Enterprise and public-sector buyers ask for an accessibility conformance report, Section 508 applies to US federal purchasing, and the European Accessibility Act has applied to many digital products and services in the EU since June 2025. We score what has been tested on released work, not what a policy says.
Check it: Ask whether the agency has produced a conformance report or test results for a client, and which assistive technologies it tested with. Then tab through the demo request form on one of its client websites without using a mouse.
Post-handoff coherence: 20 percent
B2B clients keep their websites and products far longer than the agency stays involved. A marketing team will publish new pages every week, and an engineering team will add features long after the launch. This criterion measures what that team inherits: a governed component library, documented templates, and a site or product that stays consistent when other people extend it. We weight it as highly as live quality because, in B2B, the years after launch cost more than the launch itself.
Check it: Speak to the person who maintains the work today at a past client, such as an engineering lead or a marketing operations manager. Ask how much of the original design is still in use and what had to be rebuilt.
Independent signals: 10 percent
This criterion covers awards, analyst placements and verified client reviews. It carries the lowest weight because the most-awarded work tends to be consumer campaigns, and much B2B work sits under client confidentiality and never reaches an awards jury. A low score here is common for strong specialists in enterprise, government and healthcare.
Check it: Prefer reviews from named B2B companies over anonymous ones, and analyst reports in your own segment over general design awards.
Reweighting for your case
The weights suit a typical B2B buyer, and your situation may call for different ones. If you sell to government or regulated industries, treat accessibility as a pass-or-fail condition rather than one factor among six. If no one in-house will maintain the work, post-handoff coherence matters even more. If you are commissioning an AI feature, weigh research evidence and live product quality most heavily, since that is where failures show first. For a first marketing site at an early-stage company, live quality and the minimum engagement usually matter more than anything else.
To apply your own weights, copy the six criterion scores from the relevant ranking page into a spreadsheet, multiply each score by your own weight, and compare the totals. When your result and ours disagree, trust yours.
Step 4: Shortlist and write the brief
A shortlist of three to five agencies is the right size. Fewer and you have no comparison, more and you cannot evaluate the pitches with the attention they need.
Building the shortlist
Apply the filters in an order that removes the most agencies first:
- Use case. Start from the ranking that matches Step 1.
- Minimum engagement against your budget. Remove every agency whose minimum you cannot meet.
- Criteria that matter to you. If accessibility is a gate, drop agencies that are weak on it, however high the composite.
- Sector and client fit. An agency that has worked in your domain skips the education a generalist needs, and you should weigh that against a slightly higher composite elsewhere.
Then read the record of each remaining agency and keep the three to five whose strengths match the criteria you weighted highest, not the five with the highest totals.
The brief
Write it once and send the same document to everyone. Two to four pages is enough. Include:
- The problem, in the language of your business rather than the language of design. What is failing today, and what does it cost?
- Who uses the result, by role. An admin, an end user, an approver and a support agent are four different users, and a brief that says “the portal” without naming them will be estimated for one.
- The measure of success: the number that will be different if this works.
- Constraints: the technology stack, the content system, compliance obligations, and any component library that already exists.
- What has already been tried or decided, so a dead end is not paid for twice.
- Your budget range and your date.
- How the decision will be made and by whom.
State the budget. Buyers hold it back to avoid anchoring the price, and what usually happens instead is that agencies price to the top of what they guess, or decline without telling you why. A stated range gets you proposals scoped to it, and you can compare them.
Step 5: Ask the right questions in pitches
A pitch is a performance, and the useful information sits in the gaps of it. Ask the same questions of every agency, write down the answers, and score them afterwards, not during the meeting.
Who will actually do the work
Ask for names, roles and the percentage of each person’s time on your project. A senior designer at 20 percent and a mid-level designer at 100 percent is a very different engagement from the reverse, and both are described as a senior team. Ask who will be in the room after the pitch team leaves, and ask to meet them.
How they handle what they do not know
Ask what they would want to learn in the first two weeks, and what would change their proposal. A good agency has specific unknowns and a way of resolving them. A weak one has a template. Ask, too, what they would leave out to fit your budget, because scoping is the clearest sign of judgement in a pitch.
Evidence you can examine
Ask for live products, a research example and an accessibility method, using the checks from Step 3. Ask for one project that did not go as planned, and how it was recovered. An agency with no such story either has not done much or is not telling you the truth about it.
Handoff and afterwards
Ask what your engineers will receive, in what format, and on what day. Ask how many designers stay involved during the build, and what the agency does when a developer says a design cannot be built. Ask what happens six months later, when you want to add a screen: can your team do it alone, using what was delivered?
Commercial terms
Ask whether the price is fixed or time-and-materials, what triggers a change order, and what the rate is for extra work. Ask how notice and termination work. Ask which parts of the work are subcontracted, and to whom.
Scoring the pitches
Score each agency from zero to five on the same five areas above, add a line for the strongest and weakest thing they said, and compare the sheets a day later. Pitches favour the most charismatic person on the team, and a sheet corrects for that. If two agencies are close, ask both for a paid, fixed-scope discovery phase, which is the fairest test available: you will learn how they work before you commit the larger budget.
Step 6: Contract and handoff terms that protect you
Most disputes with design agencies trace back to something that could have been written in the contract and was not. These are the terms worth negotiating, in roughly the order they matter. Have your own counsel review the final wording, because this is a checklist of topics and not legal advice.
Scope, deliverables and acceptance
Define scope as outcomes plus a named list of deliverables. “Design of the onboarding flow” is not a deliverable. “Designs for every state of the onboarding flow across three roles, in the component library, with annotations” is. Attach acceptance criteria you can test: which screen sizes, which browsers, what level of accessibility conformance and how it will be tested, including the assistive technologies to be used.
Phasing and exit
Split the engagement into a paid discovery phase and a separately committed delivery phase, with the right to stop after discovery. It costs a small amount more in total and removes most of the risk. Ask for a notice period of 30 days, not 90, and check that any retainer states whether unused days roll forward.
People
Name the key personnel and require your approval before any of them is replaced. Ask that subcontractors be named in the contract and bound by the same confidentiality and ownership terms as the agency.
Ownership and files
The work should belong to you on payment, including source files, design tokens, illustrations, and any custom fonts or licences bought for the project. Where the agency keeps ownership of pre-existing tools or a framework it brings, the contract should give you a licence that survives the end of the engagement. Ask for files in the formats your team uses, and ask for them at every milestone rather than only at the end.
The handoff package
Write down what handoff means. A reasonable list for product and website work:
- A component library with definitions for every state, including empty, loading and error.
- Design tokens for colour, type, spacing and motion, in a format your code can consume.
- Accessibility annotations: focus order, labels, roles and the reasoning behind them.
- A written record of decisions, including what was deliberately left out and why.
- A walkthrough session with your engineers, recorded.
For branding, substitute usage guidelines, master files in every required format and templates your team can edit. For AI work, add the interaction rules for uncertainty: what the interface shows at each confidence level, how a user corrects an output, and when it hands over to a person.
Support after delivery
Ask for a defined window, commonly 30 to 60 days, in which the agency answers questions and corrects work that does not match the agreed designs, at no charge. After that window, a set hourly rate for follow-on questions, held at the contract rate for a year, saves a renegotiation at the moment you have the least leverage.
Money
Pay against milestones tied to deliverables, not to calendar dates alone, and avoid a large payment up front. State the process for change orders in writing: who can request one, how it is priced, and that no work under it starts without your written approval.
Step 7: Red flags
None of these is proof that an agency is wrong for you. Each one is a reason to ask a further question before you sign.
- Only case studies, no live product. If nothing they built can be opened and used, you are judging the photography.
- The team in the pitch is not the team on the work. Ask for names and time allocations in writing.
- A price with no assumptions. A number that does not say how many user types, screens or research sessions it covers cannot be compared with another number.
- Pricing per screen. A screen is not a unit of work. The fifth variant of a table costs a fraction of the first, and per-screen pricing rewards producing more of them.
- Research promised but not described. “User research” without a sample size, a recruitment method or a named output is a workshop with stakeholders.
- Accessibility described as a standards number. Ask which tools and people test it, and what was found on the last project.
- Handoff described as a Figma file. A file is not a delivery. Ask what states, tokens and documentation come with it.
- Refusal to name a reference. A client who used the work and came back is the best evidence available, and an agency with none should explain why.
- Pressure on timing. A discount that expires on Friday is a sales technique, not a cost saving.
- No questions from them. An agency that proposes before asking what you have already tried has decided what it will sell you.
- A bundle across all four use cases. See Step 1: one team is rarely strong across all of them.
Using the table on this page
The table below lists every agency on the site with its hourly range and minimum engagement, taken from each agency’s record. Use it as the first filter of Step 4, after you have chosen a use case, and not as a place to pick a winner. Rates are ranges because an agency’s rate depends on the seniority of the people staffed and the length of the work, so treat the range as an opening figure to test in the pitch.
Once you have your budget and your use case, go to the ranking that matches and read the records there. Start with website, product, branding or AI product design, and return to the home page if you want to see all four together. For a note on how this site operates, see the about page.