Three weeks into a website build, the client's feedback starts to feel like a test. The homepage “feels flat.” The marketing lead wants the headline shorter and the founder wants it longer. A new page turns up in the list the week before launch, filed as a revision.
That's usually when people search for how to handle difficult clients. On a website project, most of them aren't difficult people. They were never given a clear way to say what they want, so it comes out vague, contradictory, late, or aimed at something already settled. Each of those is a process problem with a specific fix. A few clients are a different matter, and they get their own section near the end.
Why do clients get difficult on a website project?
A client hires you because they can't build the site themselves. That usually means they can't describe one precisely either. Ask for feedback with no structure and what comes back is shaped by whoever had time to reply, in whatever words they had.
All of it happens in good faith. It's what review looks like when nobody said how review works.
How to handle difficult clients, one kind of feedback at a time
Vague feedback: “make it pop”
Agree before the review: comments go on the page, one issue each, in a fixed shape. What you see, what you expected, and why. Our guide on how to annotate a website walks through that shape with examples.
When it happens anyway: don't guess, and don't answer with three new versions. That spends a round on a comment nobody understood. Ask which element they mean and what a visitor should do there that they aren't doing now. Then offer two concrete readings (“bigger headline” or “more contrast on the button”) and let them pick.
Contradictory feedback from two people at the client
Agree before the review: one named person signs off. Anyone at the client can comment. When two comments disagree, that person decides, and you build what they decide.
When it happens anyway: don't pick a side, and don't build a compromise neither of them asked for. Put both comments in front of the person who signs off, say what each one costs, and ask for one answer in writing.
Late feedback, after sign-off
Agree before the review: every round has a closing date, and the contract says what happens to comments that arrive after it.
When it happens anyway: sort it. Something broken is yours to fix whenever it's found. A late opinion is a change, and it goes to the next round or gets a price. Refusing it outright starts a fight. Pricing it starts a decision.
Scope creep dressed as a revision
Agree before the review: a revision changes something inside the agreed scope. A new page or a new feature is a change request, with its own cost and its own date.
When it happens anyway: say yes, with a price. “That's a new page. Here's what it costs and what it does to the launch date. Do you want it in this phase or after launch?” Most scope creep stops being a problem once it has a number next to it. Some of it turns out to be a good idea for the next phase, which is a better outcome than an argument.
The reopened decision
Agree before the review: decisions get written down when they're made, with who made them, in a place both sides can open.
When it happens anyway: answer with the record, not your memory of it. Then ask what has changed. New information makes it a change request. No new information, and the link to the earlier decision is usually the whole reply.
The record protects the client as much as you. One of our customers, an agency, spent six months on a product redesign. When the client's team first clicked through the production build on November 2, two of their first three comments asked nearly the same thing: “I thought we opted to pull this?” They were right. Both had been settled in July, in threads on the prototype. The decisions existed. They just weren't where the people building production were reading.
What should the contract and kickoff settle?
This is practical advice from the review side of the work, not legal advice. Have a lawyer read your contract. These are the things worth having in writing before the first review:
- How many rounds are included, and what a round is: one consolidated set of comments on one version, by a date.
- Who signs off. One name, not a committee.
- What counts as a change request, and how it gets priced and scheduled.
- When payments are due. Tie them to dates or to work you control, not to a final approval the client can withhold.
- How either side ends it, and what gets handed over when they do.
Then say all of it out loud at kickoff. A clause the client signed and never read doesn't change how they give feedback. A five-minute walkthrough of how review will work does.
Where should the decisions be kept?
“We already agreed this” only works if you can show it. On most projects the agreement is spread across an email thread and a call nobody recorded. A shared document can do the job if someone keeps it up every week. Few people do.
The record holds up best when it's made in the same place as the feedback. In Skyelight, the website feedback tool we make, a comment is pinned to an element on the live page, and the replies stay with it as a thread. Each thread has a status: open, resolved, or deferred for the good idea that belongs in a later phase. At handover, a coding agent connected to the Skyelight MCP can write a project review or decision log from those threads, quoting each one with its id so every line can be checked.