Some version of this story comes up again and again in the communities we follow.
A team grows from four people to twelve, and the review process that worked fine when everyone sat together starts falling apart. Feedback arrives in task comments, email threads, chat messages and the occasional screenshot with circles drawn on it. Two reviewers ask for opposite changes. Nobody is sure which version the client actually signed off.
Other people describe the same problem from different angles. An agency wants clients to request work, comment and approve it, but every client login on its project tool costs a paid guest seat. A studio embeds a staging site inside a client portal so feedback happens "in one place", and then discovers the comments still end up in email. People building approval flows for publishing and for local-business profiles keep landing on the same uncomfortable detail: the record says "approved", but not what was approved. And as AI tools draft more of the work, the "agent drafts, a person approves" step has become the part teams most want to get right.
We build and look after sites and community platforms for a living, so review rounds are a big part of our week too. Here's how we think a review round could work, step by step, without buying a new tool first.
Step 1: Make each round a thing with edges
The biggest single fix is treating a review round as its own item, not an endless comment thread on the original task.
A round has:
- One version under review. A link, a file name or a staging URL that won't change during the round.
- A named set of reviewers. Not "the client", but the two or three people whose input counts.
- A due date for feedback. After it, the round closes and late notes go into the next round.
When feedback has nowhere to "live" except the original task, every new comment reopens everything. A round with edges lets you say "that's round three, we'll pick it up in round four" without anyone feeling ignored.
Step 2: Turn feedback into a checklist, not a conversation
Free-text comments are fine for collecting feedback, but they're hard to close. Before you start changing anything, convert each piece of feedback into a single checklist item.
Then every item gets one of three replies:
- Done, with a link or screenshot.
- Won't do, with one sentence explaining why.
- Needs a decision, because it conflicts with something else.
The "won't do" reply is the one most teams skip, and it's the one that stops the same note coming back three rounds later. In our experience it's also the first thing that makes clients feel heard, even when the answer is no.
Step 3: Name who breaks ties before the round starts
Conflicting feedback isn't a tooling problem. It's an ownership problem.
If marketing wants the headline shorter and legal wants a disclaimer added under it, someone has to decide. Agree up front who that is for each project: usually one person on the client side and one on yours. When a "needs a decision" item appears, it goes to them, not back into the group thread.
Step 4: Approve the thing, not the task
This is the detail that keeps coming up in conversations about approval portals. An approval that only records "Approved" and a first name is weak evidence later. A useful approval records:
- What was approved: the exact version, file or staging snapshot.
- Who approved it and when.
- Any conditions: "approved, pending final logo".
If the content can change after approval, keep a copy or a version reference. Otherwise "you approved this" turns into "that's not what I saw".
Step 5: Decide where clients actually log in
This is where the cost question shows up. Many project tools charge for every guest, viewer or client account, so teams either pay for seats that click "approve" twice a month, or they move approvals back to email, which undoes steps 1 to 4.
Practical options people use:
- Guest access in the existing tool, if the plan includes enough free guests and permissions are tight enough.
- A shared form or approval page per round that writes back to the task.
- A client portal on a site you already run, so clients see only their projects and the cost doesn't grow per person.
- Email approvals with a reference number, logged by someone on your side. Low-tech, but it works if one person owns the logging.
There's no single right answer. A five-person studio with two clients needs something different from an agency with forty.
Disclosure: we're the Wbcom Designs team, and we build Kanbanomy, a self-hosted project-management plugin for WordPress with a client portal and approvals. It's one option for the third approach, alongside hosted tools with generous guest plans and plain forms.
Step 6: Close the round out loud
When the checklist is done, send one short summary: what changed, what didn't and why, what's waiting on a decision, and what the next round covers. It takes five minutes and saves the "wait, did you see my note?" message that otherwise arrives two days later.
Where it usually goes wrong
- Rounds with no end. If feedback is accepted forever, every round becomes the same round.
- Too many reviewers. Each extra person adds opinions faster than they add clarity. Invite people to watch, but only named reviewers count.
- Approvals in chat. A thumbs-up in a busy channel is almost impossible to find six months later.
- Tool first, process second. Switching tools to fix messy reviews often just moves the same mess into a new interface. The steps above work in a spreadsheet; the tool only makes them easier.
- AI drafts with no named approver. If something drafts the work automatically, the approval step needs a human name on it, not "auto-approved after 48 hours".
A short checklist to steal
- One version per round, frozen until the round closes
- Named reviewers and a feedback deadline
- Every note becomes a checklist item: done, won't do, or needs a decision
- One tie-breaker per side, agreed up front
- Approvals record the version, the person, the time and any conditions
- Clients log in somewhere that doesn't cost more every time you add one
- A closing summary for every round
Your turn
We'd love to hear how this works in your team:
- How do you stop feedback arriving in five different places? Did a rule work better than a tool?
- When two reviewers disagree, who decides in your projects, and did you agree that before the work started?
- Do your approvals record exactly what was approved, or just that someone said yes? Has that ever come back to bite you?
- Do you pay for client or guest seats on your project tool, or have you found another way to let clients approve work?
Top comments (0)