Name one approver before the review, judge the design against the agreed goal, merge comments into one list and record why. A design approval process. To get a design signed off when five people hold five opinions, settle the process before anyone sees the work.
To get a design signed off when five people hold five opinions, settle the process before anyone sees the work. Name the one person who approves and the people who advise. Agree what the design is judged against. Collect each reviewer's comments in writing, merge them into one list, and resolve the conflicts before that list reaches the designer. Then write down what was decided and why, in a paragraph.
In our view, a stalled design approval is a process problem before it is a taste problem: an argument nobody set up to end. When no one has said who decides, or against what, every opinion counts the same, and the loudest or most senior one can win, late. This page is for the person in the middle, a product manager or the sponsor who has to take the work upstairs, and it covers the whole design approval process from setup to sign-off.
Why design reviews stall
A design review gathers people who each know something the others do not. Sales knows what prospects ask for, support knows where users get stuck, engineering knows what is expensive to build. The point of the review is to pool that knowledge into a better decision.
Groups are poor at this unaided. In a 1985 study by Garold Stasser and William Titus, small groups chose between candidates after each member had read only part of the facts. The authors' model of what happened is that discussion favours information members already share and information that supports their existing preferences. The groups chose the candidate a plurality had favoured at the start, not the strongest one, and discussion tended to perpetuate members' distorted pictures rather than correct them.
A live design review has the same shape, with a senior voice added. Two practical lessons follow, and they shape everything below:
- Collect opinions separately, in writing, before any meeting. The meeting is then for resolving differences, not for discovering them while the most senior person speaks first.
- Give the discussion something to argue against other than preference. That is the agreed goal, and it has to exist before the review does.
Name who decides and who advises
The single most useful step is to write down, before the first review, who approves the design and who only advises. Several decision-rights frameworks do this. Atlassian's DACI play names four roles: a Driver who corrals stakeholders and gets the decision made by an agreed date; one Approver, "the one person (yes: one!) who makes the decision"; Contributors with subject knowledge, who advise but do not vote; and the Informed, who are told the outcome. RAPID divides the work differently (Recommend, Agree, Perform, Input, Decide) but makes the same point: ideally one person decides.
For a design review, that maps like this:
- Driver: the product manager. They set the review up, chase the comments and consolidate them.
- Approver: one person, whoever owns the outcome the design serves. For example, the head of product or the sponsor. Never a committee.
- Contributors: engineering, sales, support, marketing, a customer-facing specialist. Their input is wanted and will be weighed; it is not a vote.
- Informed: teams whose work changes because of the design, told once the decision is made.
Two refinements matter in practice. First, some reviewers hold a genuine veto on narrow grounds: legal, accessibility, security, brand. Name them, and name what their veto covers, so a brand reviewer can block a misused logo but not a navigation structure. RAPID's Agree role covers this kind of mandatory requirement and is assigned sparingly; when Agree and Recommend cannot reconcile, the Decide role still makes the final call. Second, if the senior sponsor is not the Approver, they need to say so, in writing, to the people involved. An Approver whose boss can overturn them silently is not the Approver.
Review against the goal, not against taste
A reviewer asked "what do you think?" will answer with what they like. A reviewer asked "does this help a first-time admin finish setup without calling support?" will answer about the product. Before the review, write the goal the design is meant to serve, the constraints it has to respect and the questions you want answered at this stage.
Good goals are specific enough to judge against: the user, the task, the measure that would tell you it worked, and the business limits a designer has to work inside. If the design team does not have those limits, the review will surface them late; our piece on what a designer needs to understand about your business covers what to hand over first.
Stage matters as much as goal. Feedback on button colour in a review meant to choose between two flows is noise. The Government Digital Service's design community advises opening a crit by specifying what you want feedback on and giving context, which keeps the conversation on track. Say plainly which decisions are open in this round and which are already closed.
Show options with trade-offs, not one design to like
One design invites a verdict: like or dislike. Two or three options, each with its trade-offs written beside it, invite a judgement: which cost are we willing to pay? That moves the review from taste to the goal.
Keep the options genuinely different, not three shades of the same idea, and say for each one what it is good at, what it gives up and which part of the goal it serves best. Add a recommendation. A set of options without a recommendation hands the reviewers the designer's job; options with one give them something to test. It is how we work with clients: we bring options, trade-offs and a clear recommendation, and the client keeps the product decisions.
A pre-review setup checklist
Before the first review, the Driver should be able to tick every line:
- The Approver is named, and has agreed to be the Approver.
- Contributors are named, with the reason each one is there.
- Any narrow veto (legal, accessibility, security, brand) is named, with what it covers.
- The senior sponsor has either taken the Approver role or delegated it in writing.
- The goal is written: user, task, measure, constraints.
- This round's questions are written, and so are the decisions already closed.
- The options are ready, with trade-offs and a recommendation.
- The version under review is named, so every comment refers to the same thing.
- There is one place for comments, and a date by which they are due.
- Everyone knows what happens to a comment that arrives after that date.
Consolidate the feedback into one list
The Driver's real work starts when the comments come in. Each reviewer writes theirs separately, against the goal and the round's questions. The Driver then merges them into a single list, and nothing reaches the designer until conflicts on that list are resolved. A designer handed five people's raw notes has been handed the decision, without the authority to make it.
A simple table does the job. The rows below are an illustration, not a real project:
| Comment | From | Goal or question it bears on | Decision | Owner |
|---|---|---|---|---|
| The plan comparison hides the annual price | Sales | Goal: self-serve upgrade | Accept: show both prices | Designer |
| Move settings into the main navigation | Support | Question 2: navigation | Reject: tested worse in round one | Driver |
| The tone feels too casual | Marketing | None stated | Ask: which user or brand rule? | Driver |
| Fields fail contrast checks | Accessibility | Constraint: WCAG AA | Accept: required | Designer |
| Rework onboarding as well | Sponsor | Out of scope this round | Park: next round | Approver |
The rules for filling the Decision column are short:
- Tie every comment to a goal, a question or a constraint. If a comment ties to none, ask the reviewer which one it serves. Sometimes it reveals a goal nobody wrote down; then add it, openly.
- Evidence beats preference. Research, analytics, a support log or an engineering cost outweighs "I prefer".
- Conflicts go to the Approver, framed as a choice. "Sales wants both prices shown, marketing wants one; showing both serves the upgrade goal; recommend both." The Approver decides once, and the list records it.
- Parked is a real decision. Out-of-scope comments get a home and a round, not silence.
Send the consolidated list back to every reviewer, so each one sees what happened to their comment and why.
Bring the senior sponsor in early
The late veto, where a senior person sees the design for the first time near the end and reopens everything, is an expensive failure, because it arrives when change costs most. It need not come from bad faith: the sponsor may simply never have been asked the questions that matter to them while the answers could still change.
Three habits guard against it. Show the sponsor the goal and the options at the first round, not the polished result at the last. Ask them directly what would make them reject the work, and add the answer to the goal or the constraints. And agree, in advance, that a decision already recorded is reopened only for new information, not for a change of mind. If the sponsor cannot attend reviews, they delegate approval in writing and receive the decision record, so their silence is a choice rather than a gap.
Handling "I just don't like it"
The Government Digital Service's design community is blunt that feedback like "I don't like it", based on personal opinion, is not useful. That is right as a rule for what counts as feedback. It is less useful as a response to a senior stakeholder who has just said it.
Treat the dislike as a symptom and look for the cause. Ask which user it would fail, which goal it works against, or what they expected to see instead. There may be something real underneath: a customer they know well, a brand rule nobody wrote down, a competitor screen they have in mind. Once named, it can be judged like any other comment.
If nothing is underneath, say so kindly and refer back to the goal and the Approver. If the person saying it is the Approver, it is their call, and they are entitled to make it. Record it honestly: a decision taken on preference, against the recommendation, with the evidence noted. That record protects the team, and can prompt a second look on its own.
Write a short decision record
Decisions that are not written down get remade. The software world has a good model for this: Michael Nygard's architecture decision records, short documents that state the context, the decision, its status and its consequences, including the negative ones. A design decision needs the same thing, and a paragraph will do.
A template, to fill in after each decision:
Decision [number], [date]: for [screen or flow], we chose [option] over [the alternatives], because [the goal it serves and the evidence]. [Approver] approved it; [Contributors] were consulted, and [any dissent, by whom and why] was recorded. We accept [the known cost or risk]. We will revisit it only if [the condition, such as a measure falling below a stated level]. Status: [proposed, accepted or superseded by decision number].
Keep the records together, in order, where the team already works. When someone asks in six months why the navigation looks the way it does, the answer is a link, not a meeting.
What sign-off should mean
Sign-off is the Approver accepting a named version of the design against the agreed goal, with the open items listed. It is worth saying what it covers and what it does not:
- It covers a version. Name the file, prototype or build, so later changes are visibly later.
- It covers a scope. State which screens or flows are approved and which remain open.
- It carries its exceptions. Parked comments and accepted risks travel with it, not in someone's memory.
- It has a reopening rule. New information reopens a decision; a change of mind on its own does not.
- It reaches the Informed. Tell the affected teams once, with the decision record, so the approved design is the one that gets built.
Name the decider, review against the goal, and write down why. Those three habits turn five opinions into one decision with an owner.
About the authors
Yash Gidwani
Product Designer
More from YashYash Gidwani is a Product Designer at Tcules who writes about getting people into a SaaS product and keeping them there. Yash's guides cover SaaS onboarding strategies, onboarding and engagement software, and redesigning a live product without upsetting the people who use it. Other articles go into specific workflows: lending and loan origination apps, insurance quoting and claims, staffing and recruitment software, and a search-first create flow that stops duplicate records. Yash has also co-written the SaaS UX audit checklist and a 20-question checklist for legacy modernisation readiness.
Anshu Jain
Business Development Specialist
More from AnshuAnshu Jain is a Business Development Specialist at Tcules, the digital product design studio in Ahmedabad. At Tcules, Anshu works on how SaaS teams choose a design partner, and co-writes about questions such as what a product design agency should do, how to hire a product designer, and what to ask a UX audit agency before signing. The pieces below speak for the studio, and this page grows as Anshu writes.