Offboarding

Why does one offboarding checklist fail across three kinds of exit?

Because a resignation, a dismissal and the end of a fixed term are three different events wearing the same paperwork. A resignation gives you a cooperative person and a period of notice in which to use them, so the priority is handover and keeping the relationship warm. An involuntary exit gives you neither: the conversation is often the last one, access has to be considered immediately, and the sequence is dictated by dignity and by documentation rather than by convenience. The end of a fixed term or a contract engagement is a scheduled event nobody treats as one, which is why it is the track most often missed entirely, with access left live and equipment unreturned months later. A single checklist written for the resignation case silently misapplies its timing to the other two.

When does offboarding actually start?

At notice, not at the last working day, because both of the things offboarding depends on decay from that moment. Leverage decays first. While someone is still employed, a request to write down how a process actually works is simply part of the job and it gets done; the same request a month after the last day is a favor asked of a stranger, and it is answered thinly or not at all. Goodwill decays alongside it, as the departing person's attention moves to what comes next and their manager's attention moves to arranging cover. The window between notice and departure is the only period in which the person is simultaneously available, still informed and still obliged, and it is shorter than it looks. Starting then costs no more effort than starting later.

What actually leaves the building with the person?

Not the documented work, which survives in the repository or the file share. What leaves is the undocumented layer: why a workaround exists, which vendor contact answers the phone, which client will escalate if a date slips, which of two apparently identical reports the board actually reads. None of it is secret and none of it was ever written down, because writing it down was nobody's task while the person was there to be asked. The exit is also the moment an employer discovers its single points of dependency, since the questions that suddenly have no answer map precisely onto them. That discovery is useful and badly timed. Treating handover as a transfer of relationships and reasoning rather than of files is what separates an exit the team absorbs from one it is still recovering from a quarter later.

How do you revoke access without stopping the handover?

The dilemma is real and it is usually resolved badly in both directions. Revoke everything at notice and the handover stops: the person cannot open the systems whose contents you are asking them to explain, cannot forward the thread that contains the decision, cannot show the successor where anything lives. Leave everything live and an account belonging to someone halfway out of the building stays authenticated to production and to the customer record for the whole notice period. Neither extreme is a policy so much as an avoidance of the judgment involved. The workable position separates access by consequence rather than by date. Read access needed for handover can survive to the last day; the ability to change, export, delete or approve rarely needs to. Splitting the two is a configuration exercise done once.

The harder problem is knowing what to revoke. Most employers can name the systems provisioned centrally and almost none can name the ones a team adopted on a card, the shared credential nobody rotated, the vendor portal where the departing person is the named administrator. What is not in the inventory does not get revoked, and it does not announce itself; it is found later, usually because an invoice keeps arriving. Keeping the record of what a person holds in one place, alongside the employment record rather than in a spreadsheet a team maintains privately, is the only version that survives a reorganization. Where an [HR system](/hr-software) already carries the person's assets and entitlements, the exit query becomes a lookup instead of an investigation.

Why does knowledge transfer fail even when a handover document exists?

Because the document answers the questions the departing person thought to ask themselves, and the ones that matter are the ones they no longer notice. Expertise compresses. After a few years the reason a report is filtered a particular way stops being a decision and becomes a habit, invisible to the person performing it and impossible to write down because it does not feel like knowledge. A handover written unprompted therefore captures the procedures and omits the judgment. The intervention that works is not a better template but a successor: someone who has to do the work asks the questions the incumbent cannot generate, and each one extracts context that would otherwise have left silently.

Relationships are the part no handover document has ever transferred. A vendor who answers within the hour does so because of a person, not a contract; a counterpart in another department who agrees to a change without a ticket is extending credit to an individual. When the individual leaves, the successor inherits the formal version of every relationship the incumbent held informally, which is slower and reads to everyone else as a decline in service. The only thing that softens it is a deliberate introduction while the incumbent is still there to make it, which is almost never scheduled. Whatever the exit fails to transfer becomes something [the onboarding process](/employee-onboarding-software) has to reconstruct from scratch, and the two halves are usually owned by people who never compare notes.

What does the way an exit is handled tell the market?

More than the recruitment marketing does, because it is the part people talk about unprompted. A candidate weighing an offer will ask someone who left, and what that person says is shaped far more by how the ending was handled than by everything that preceded it. The asymmetry is unfair and it is stable: a good tenure ending badly is remembered as a bad employer, while an ordinary tenure ending well is remembered as a decent one. Very little of what produces the good version is expensive. Being told clearly what happens next and when, having the final conversation held by a person rather than delivered by a system, having questions answered after the last day by someone who still replies. What produces the bad version is usually not malice; it is silence.

The commercial argument sits alongside the reputational one. A person who left on good terms is a shorter hire than a stranger: they need no cultural assessment, they arrive knowing the systems, and their reasons for leaving are usually knowable. They are also a referral source, and referrals from former employees carry weight precisely because the person has nothing to sell. Both disappear if the exit was handled as an administrative disposal. Employers that keep any of this record, at exit, whether the person would be considered again and why, in a form a future recruiter can find. Where a departing person can retrieve their own documents through [a self-service portal](/employee-self-service-portal) rather than by asking a favor of an ex-colleague, one common source of late friction disappears.

Which parts of an exit are governed by rules rather than by preference?

Several, and they are the parts most often handled by inherited habit. Final pay carries obligations about what must be included in it and by when it must reach the person, and those obligations differ between countries and frequently between states inside one country. Records relating to a departing employee carry a second set: how long they must be retained, what the person is entitled to receive, and what must be deleted afterwards. Restrictive covenants sit on a third: whether notice of them was properly given, how far they reach, and whether they bind at all once the relationship has ended. None of this is a matter of policy preference and all of it changes. Confirm the current position for each place you employ people with a qualified advisor or the relevant authority.

The involuntary track is where this bites hardest, because it is the one where a procedural shortcut is most tempting and most expensive. An exit that is contested later is reconstructed from what was recorded at the time, which means the record has to have been made contemporaneously rather than assembled afterwards from memory. That is a discipline about the months preceding the decision rather than about the exit itself. The second failure mode is a workflow with a jurisdiction's rule hard-coded into it, written once by somebody competent and then applied unchanged to a new location or a rule that has since moved. Encoding the sequence is sound; encoding the substantive answer is not. Keep the checkpoints generic and keep the local determination with a named person responsible for knowing it is current.