How should an enterprise write an RFP for a product design agency?

Illustrative situation: a product director at a large insurer sends a 40-page questionnaire to six design agencies for a redesign of the claims portal. Six responses come back in six shapes, among them a 90-page document, a set of unpaid mockups of the new dashboard, a quote for discovery only, and a team page led by a partner who, it turns out later, will attend two meetings. The scoring panel cannot compare any of it, so it picks the best presentation. The scene is invented, and every problem in it traces back to the request, not the agencies.

Humbleteam, a digital product design agency, publishes this guide. It describes no Humbleteam process or pricing.

This guide uses US federal and UK public procurement rules as reference points because both publish detailed rules for fair competition. They bind public buyers, not private companies. It is general information, not legal or procurement advice: if public-sector rules apply to you, follow your procurement team.

Is an RFP the right tool at all?

Not always. Three instruments do different jobs.

A request for information (RFI) gathers market information when you do not yet intend to award. The US Federal Acquisition Regulation says RFI responses are not offers and have no required format. The 18F De-risking Guide, published by the US General Services Administration, recommends something sharper: release a draft of the request itself during market research, because it is more likely to result in qualified proposals than an RFI.

A request for proposal (RFP) fits when you can describe the problem, the constraints and a budget envelope, and you want several proposals you can compare.

A direct paid discovery fits when you cannot yet describe the problem. The UK government’s Service Manual says a discovery typically takes around 4 to 8 weeks and starts by turning a predefined solution into a problem to solve. If “we need a new claims dashboard” is all you have, buy that work first and run the RFP on what it finds.

What goes in the RFP?

Write the problem, not the answer. The Service Manual’s own example: the problem is not “build an interactive map” but how people find their nearest contact center. A redesign brief works the same way.

Two items deserve their own lines. The first is a budget range. The UK charity body NCVO advises buyers to state the budget and the number of agencies invited, so bidders know where they stand. 18F’s contract format gives every bidder the same not-to-exceed ceiling, whatever each one proposes. The second is how you will score. Under the UK Procurement Act 2023, a public buyer with more than one criterion must state their relative importance, and US federal rules require factors and their relative importance to be stated clearly in the solicitation. A private company is not bound by either. It still gets better responses when agencies know what wins.

What should agencies submit, and what should you never ask for?

Ask for evidence you can check, in six parts.

• The named team and each person’s allocation. 18F describes a key personnel clause for this and warns against naming too many roles, because bidders then keep those people functionally benched until the contract is awarded. Name the lead designer, the research lead and the delivery lead.
• One or two relevant cases with the agency’s exact scope. “We redesigned the app” and “we ran research and designed the onboarding flow; the client’s team built it” are different claims.
• Their approach to discovery and research. In a 2003 article Jakob Nielsen argued that, before a project starts, reviewing an agency’s test methodology and project plans is the only way to assess its professionalism, since you cannot know whether the assigned designers will match its portfolio.
• Their plan for your design system and engineering handoff: how they will extend your components, document decisions and work with your engineers.
• Assumptions and exclusions: what the price leaves out and what they need from you.
• Price by phase, with the roles and days behind it.

Do not ask for free concepts of your product. AIGA, the US professional association for design, warns that speculative work risks compromised quality, because it skips the research and testing a real project needs, and creates intellectual property risks for both sides. If you want to see how a team thinks, NCVO suggests paying two finalists for a day of their time to run a workshop with you.

How do you keep responses comparable?

Fix the shape. Give a template with the same headings for every bidder, and a page limit. 18F recommends keeping proposals short, with a hard page limit, and lists bypassing page limits with tiny fonts as a red flag.

Keep one Q&A log. Under federal rules, specific information given to one potential bidder must be made available to all to avoid an unfair advantage. The UK’s current Digital Outcomes buyer guide tells buyers to post all questions and answers where every supplier can see them. Answer questions without naming who asked.

Then freeze the rules. The same guide says buyers cannot change criteria or weights once requirements are published.

How should the scoring rubric work?

Write it before responses arrive, give each criterion a weight, and describe what a 1, a 3 and a 5 look like, so two evaluators reading the same proposal reach the same number. Keep pass or fail gates, such as accepting your contract terms or completing your vendor security review, outside the weighted score.

Score independently, then calibrate. The UK’s Digital Marketplace scoring guidance, withdrawn at the end of 2025 but still a clear method, told each evaluator to score alone, share nothing until every supplier was scored, and then discuss and agree each score instead of averaging. It also warned against letting price influence quality scores. One practical way to follow that: score price last.

18F goes further and skips numbers: each evaluator lists pros and cons, and the team discusses them, which it says produces a detailed and defensible justification. The evidence column in the rubric below keeps that discipline. A score without a cited line from the proposal is not finished.

Price can also be scored by ratio to the cheapest quote, as the UK framework guidance did. UK guidance under the Procurement Act asks buyers to consider how relative scoring mechanisms can hurt value for money, so anchored descriptions are a reasonable alternative.

An illustrative scoring example

Two fictional responses to the claims portal RFP. Weighted score equals weight times score divided by 5.

North had the stronger presentation. In independent scoring, one evaluator gave its team a 4 because the partner who presented was impressive. At calibration, the panel checked the evidence column: the partner would spend 10 percent of their time on the project, and the day-to-day lead had no relevant case. The score dropped to 2. Harbor was more expensive but named a full-time lead with a matching case and listed what its price excluded. The panel still made the call, but by then it was arguing over lines in the proposals rather than over who had presented well.

Who should present at the shortlist stage?

The people who will do the work. 18F’s guidance has key personnel take part in verbal interviews so the buyer hears from the people doing the work rather than the bid team. The interview confirms the written proposal; it does not let the bidder revise it. NCVO describes the weak version: fancy slides, standard questions, and an hour or less of interaction before a decision.

Ask the named lead to walk through the case they submitted, explain what they would do in the first fortnight, and say where they disagree with your brief. Score the interview against the same rubric rows.

What keeps the process fair and confidential?

• One point of contact during the RFP, so no bidder gets a private briefing.
• Never pass one agency’s ideas to another. Federal rules protect a bidder’s confidential business strategy even inside a shared Q&A.
• State in the RFP how you will store and return proposal materials, and that unselected work stays with its authors.
• Ask evaluators to declare relationships with any bidder before reading.
• Keep an audit trail. The Digital Outcomes guide asks buyers to record why each score was given and why the criteria and weights were chosen.
• Tell unsuccessful agencies their scores. UK public buyers must send an assessment summary; for a private buyer it is optional, and a plain courtesy.

18F says a federal process ideally takes 4 to 12 weeks from release to award. A private company has fewer formal steps, and the rubric and the Q&A log are the two worth keeping.

What can you produce now?

Four documents, before any agency sees the brief: the RFP built from the outline above, the weighted rubric with your own weights and anchors, an empty Q&A log, and a calibration plan that names the evaluators and the meeting where they agree scores. If you still need names to invite, start with this overview of enterprise product design agencies, then judge every one of them with the same rubric.