Blog
7 min read
A design change after sign-off is a new instruction, not a revision. That is the whole argument. Charging for design changes after client sign-off turns awkward for one reason. Most practices never recorded what the sign-off covered, so there is nothing to measure the new request against.
Stage 3 closes. The client approves. Three weeks later they want the kitchen moved and the roof pitch altered, and somebody in the office starts redrawing before anyone has said a number out loud.
Anything the client now wants that the approved deliverable did not promise. That is the test. It is narrower than it sounds.
Sign-off is not approval of an intention. It accepts a defined set of outputs: a stated number of drawings, at a stated scale, describing a stated arrangement.
Your client accepted a plan with a single-storey rear extension. Now they want two storeys. The plan you agreed no longer exists, so redrawing it produces a different deliverable rather than a revision of the old one.
Most practices lose this argument before it starts, because they described the stage as "developed design" and stopped there. Write it the way defining deliverables so a client cannot expand them sets out, with counts and acceptance criteria, and the boundary draws itself.
Two things sign-off does not mean, and both cause trouble. It does not mean the client has stopped having ideas. It does not mean the design is finished, because stage 4 will change plenty of things the client never sees. The meaning is narrower and more useful than either: the arrangement is fixed, and moving it now costs money.
Do this before you reply. It takes twenty minutes and it sets the fee.
You are reconstructing the boundary from whatever record exists: the appointment, the fee proposal, the drawing register, the sign-off email. Four rows are enough.
| Record | Stage 3 example |
|---|---|
| The outputs | 12 drawings at 1:50, one general arrangement per floor, one section |
| The arrangement they fix | Single storey, flat roof, kitchen on the north elevation |
| The acceptance test | Client signed the drawing register on 14 May |
| Revisions included | Two rounds within the approved arrangement |
That last row is where most disputes live. A round of revisions inside the agreed arrangement is included work. A change to the arrangement itself is not, however small the drawing edit turns out to be.
Do not skip the acceptance test row. A stage that closed on a phone call and a nod is hard to defend, and you want to know how thin the record is before a client tests it.
Where no signed register exists, use the dated email that transmitted the drawings. Weaker evidence. Still a boundary.
Run the request through these, in order, before you price anything.
Three yeses and a no is a new instruction. Price it.
Whichever rate you can evidence. There are two defensible answers and one bad one.
Your stage rate works when the change re-runs a recognisable slice of a stage. Say stage 3 was a third of the fee, and the client has asked you to redo a third of stage 3. Now the arithmetic is visible, and they can check it themselves. The bad answer is a round number picked in the moment, because a number nobody can trace is a number the client will argue with.
A published fee split shows what a stage is worth. The US practice Neumann Monson publishes its phase weighting: 20% for schematic design, 20% for design development, 30 to 35% for contract documents. Different names, same idea. Each phase carries a share of the fee, and that share is the price of doing the phase once.
Practices forget a third component. Disruption. Reopening a closed stage is not the same work as running it in sequence.
The team has moved on. Consultants have issued. Somebody now has to unpick coordination that everyone treated as settled a month ago.
Price that unpicking as hours in its own right, and name it up front, because absorbing it in silence is how a practice ends up subsidising its own coordination.
An hourly additional-services rate works when the change does not map onto a stage cleanly. The Designing Buildings wiki's summary of fee considerations for architects names three things a fee proposal should define for the client. The fee for basic services, the fee for additional services, and a time charge provision. Exactly the split you need here. Check which of the three your own appointment names.
You quote before you draw. The request arrives, and your first response is a price and a programme effect, not a revised drawing.
This is the part practices skip. Someone opens the model to have a quick look, four hours disappear, and by the time anyone raises the fee the work is visibly half done. The client sees a drawing. You see an invoice they never agreed to.
Say it plainly, in writing, the same day. Name what changed against the signed-off deliverable. Give the fee, the rate it came from and the effect on the programme. Then wait for an answer before anyone opens a file.
Barbara Kaucky chairs the RIBA Small Practice Group. She puts the principle bluntly in RIBA guidance on setting out fees clearly for clients: revisions and variations are design work that takes staff time, so they should have a cost attached.
That same guidance names the trouble it heads off. A client treats a variation as fine-tuning, part of the service, not something that should carry a fee. Saying the cost out loud, before the work, is what corrects that.
Your own time records, broken down by stage.
A practice that books hours against "the job" cannot price a reopened stage, because it has no idea what the stage cost the first time round. A practice that books hours against stage 3 can answer in an hour. That difference is worth more than any published fee scale. The guide to tracking time against RIBA stage fees covers how to set the recording up without adding admin to every timesheet.
Two projects of history is enough to start. You are not after a benchmark. You are after the last comparable stage you ran, and what it took.
Those same records settle the harder question afterwards. Say reopening stage 3 cost you 40 hours against the 12 you quoted. That gap belongs in the next fee proposal, not in a grievance. Practices that price the next job from the last one stop losing the same argument twice.
Show them what they paid for. Not the contract clause, the deliverable list they signed.
Go in assuming the client is not arguing in bad faith. They have a picture in their head of what stage 3 bought, and nobody ever corrected it. A one-page list of outputs, with the new request sitting visibly outside it, ends the disagreement faster than any appointment document.
Where it does not end it, look at the appointment. It promised revisions and never bounded them. So the fix sits upstream, in the proposal. Charging for extra revision rounds deals with the same failure on a smaller scale. The scope creep and change control archive covers the wording that prevents it.
Pick the next stage you close. Before you send the sign-off request, attach the list of what that stage delivers and mark the arrangement it fixes as approved. The next time a client asks for something outside it, you will not be reconstructing the boundary. You will be quoting from it.
COMMON QUESTIONS:
RELATED READING:
Unscoped integration work is any interface the signed scope never named. Here is how to count them all before signature, and what to do the day an undocumented one appears.
The money is lost in the ninety seconds between a client asking for a change and a trade starting it. Here is the pause, price and confirm sequence that closes it.
Turn implementation unknowns into testable scope exclusions, then route each failed assumption into discovery or a priced change request.
Korrel turns briefs into structured proposals, then tracks what each job actually costs.
START FREE TRIAL