In short
Design the ending as deliberately as the beginning. A working offboarding sequence covers five things: hand over everything the client will need without you, close the commercials, remove access in both directions, ask for feedback and a referral while goodwill is at its peak, and record what the engagement actually cost. The ending is when clients are most willing to refer and least likely to be asked, which makes it the cheapest source of new work most service businesses ignore.
Key takeaways
- The last two weeks disproportionately shape what a client says about you afterwards.
- Hand over assets and access in a form they can use without you. Ownership disputes surface months later.
- Ask for the referral and the testimonial at completion, not later. Goodwill decays fast.
- Remove your own access too. Lingering credentials are a security exposure with no upside.
- Record what the engagement cost while people remember, or the next quote repeats the mistake.
Why endings matter more than they get credit for
Service businesses invest real effort in beginnings. There is a kickoff, a welcome pack, a considered first fortnight. Endings get improvised: the final deliverable is sent, the invoice goes out, and the relationship trails off through diminishing email.
This is backwards in terms of return. The ending is when the client forms the account they will give to other people, when they decide whether to come back, and when they are most willing to say something useful about you. It is also, in most businesses, entirely unowned, because delivery considers itself finished at the last deliverable and account management considers the work delivery's until it is done.
The specific costs of a vague ending:
- A referral that never happens, because nobody asked in the fortnight when the answer would have been yes.
- A client who returns to a competitor because the ending gave no reason to think of you.
- Files and access disputes that surface months later and cost hours to untangle.
- A final invoice that arrives after attention has moved on, and is paid late as a result.
Nobody plans the end of an engagement, which is why it consistently underperforms the beginning that everyone planned carefully.
The handover pack
The test is whether the client could continue without you if they had to. Not because they will, but because a handover that leaves them dependent is a handover that generates awkward requests for months, all of which are unbillable.
| Include | Why |
|---|---|
| Final deliverables in an editable format | A flattened output is a dependency, not a delivery |
| Source files and working assets | Ownership disputes almost always start here |
| Access to anything created in your accounts | Domains, analytics, ad accounts and tools set up in your name are the classic trap |
| Documentation of anything ongoing | What runs on a schedule, what needs renewing, what breaks if ignored |
| A short note on what needs attention next | Generous, useful, and the single most common reason a client comes back |
That last row is worth the twenty minutes it takes. An honest paragraph on what you would do next, offered with no immediate commercial ask, is remembered. It also tends to produce the follow-on engagement that a proposal would not have.
Check your contract on ownership before assembling the pack, and resolve any ambiguity now rather than when the client asks in nine months.
Closing the commercials
Financial closure is easier at the moment of completion than at any point afterwards, because everyone still remembers what was agreed and the client's attention is still on the work.
- Invoice immediately on completion, triggered by the final delivery rather than by a monthly cycle. The mechanics are in getting paid is an operations problem, and the delay is worse at the end of an engagement because the relationship no longer generates natural contact.
- Reconcile any additions before the final invoice, using the record kept during delivery. If you tracked additions as described in scope creep, this is a summary. If you did not, it is an argument.
- Confirm what is now out of scope. One sentence. It prevents the slow drift where small requests continue for months after the engagement ended.
- State what continues, if anything. Retainer, support window, or nothing. Ambiguity here produces unbilled work and a client who believes they still have a supplier.
Access, in both directions
Access is the part everyone intends to handle and nobody schedules. It runs in two directions and both matter.
Your access to their systems. Client portals, admin accounts, shared drives, analytics. This is a live security exposure with no remaining purpose: your team retains credentials to a business you no longer work for, and if anyone leaves your business afterwards, those credentials leave with them, as described in employee offboarding.
Their access to yours. Shared project boards, internal channels, document folders. Leaving these open creates a client who can see work in progress for other people and who believes they still have a route in for questions.
Agreed date for mutual access removal, set during closeout rather than afterwards. Access left to be tidied up later is access that persists for years, and the discovery usually happens during a security review or an incident.
Asking, while goodwill is high
The fortnight after a successful delivery is the highest-goodwill moment of the entire relationship, and it is almost universally wasted. Three asks, and they cost one conversation.
- Feedback, asked properly. Not “how did we do,” which produces politeness. Ask what was frustrating, what they would want done differently, and what nearly stopped them hiring you. The third question is the most commercially useful sentence you will hear all year.
- A testimonial, requested specifically. Ask about a particular aspect rather than in general, and offer to draft something for them to amend. Both dramatically raise the response rate.
- A referral, asked directly. Name the kind of business you are looking for rather than asking whether they know anyone, which is too broad a question to answer.
Response rates on all three fall steeply with time. A month later the result is no longer fresh; three months later it competes with everything that has happened since. The reason this gets deferred is discomfort, and the reason it should not be is that the answer at this moment is usually yes.
The internal close
The final step happens after the client is gone, takes half an hour, and is the one that improves the next engagement.
| Record | Use |
|---|---|
| Actual elapsed time versus what was quoted | Feeds the estimation ratio described in why your estimates are always wrong |
| Hours consumed including meetings, revisions and chasing | Feeds client profitability, and this is the only moment anyone remembers |
| What was added and never charged | Tells you what should be priced into the next proposal rather than absorbed again |
| What went wrong and why | The retrospective input, while it is still specific |
This is also the natural moment to archive the engagement, moving it out of the active area with the structure intact, which is the trigger-based archiving described in where things live. Closing an engagement is a trigger; a quiet month never arrives.
Assign one owner to the whole sequence, in the same way that client onboarding needs one. Both transitions fail for the same reason: they sit between two functions, each of which reasonably believes the other has it. The Mayim Ops assessment looks specifically at these unowned transitions, because they are where growing service businesses lose the most value with the least visibility.
Frequently asked questions
What should a client offboarding process include?
A handover of everything the client needs to continue without you, final invoicing and commercial closure, removal of access in both directions, a feedback and referral conversation, and an internal record of what the engagement actually cost to deliver.
When should you ask a client for a testimonial or referral?
At completion, while the result is fresh and goodwill is at its highest. Asking three months later means competing with everything that has happened since, and the response rate falls sharply. The best moment is immediately after a successful final delivery.
What should you hand over at the end of a project?
Final deliverables in an editable format, source files, access to anything created in your accounts, documentation of anything ongoing, and a short note of what would need attention next. Handing over only the finished output creates dependency the client did not agree to.
Should you remove client access when a project ends?
Yes, in both directions and on a defined date. Your team's access to their systems is a security exposure with no remaining purpose, and their access to your internal tools creates confusion about what is still supported. Agree the date during closeout rather than leaving it.
How do you end a client relationship you no longer want?
Give notice with a clear final date, complete outstanding work properly, and hand over thoroughly. A relationship ended cleanly rarely damages your reputation; one ended by declining responsiveness does, because the client narrates it as being dropped.