Ask three people who owns onboarding and you get three answers: HR, the hiring manager, and "we have a checklist somewhere". That ambiguity is the actual problem, and no amount of content fixes it — a well-produced welcome module still leaves the new hire unsure who to ask about the thing that is blocking them.
What follows splits the first ninety days into three phases, each with a different owner and a different question it has to answer.
What onboarding is actually buying you
Two things, and it is worth being precise because they justify different budgets.
Time to productive: The gap between a start date and the point where someone contributes without supervision. This is the number that pays for onboarding, and it is measurable: ask managers when each new hire stopped needing daily direction.
Early attrition: People who leave inside a year. The reasons cluster around the first weeks — unclear expectations, no relationships outside the immediate team, and a role that turned out to differ from the one described.
Both are worth tracking before redesigning anything, for the same reason the training-plan guide insists on a baseline: without them, an onboarding programme cannot be shown to have worked.
Days 1–30: the company, and who to ask
Owned by HR. The question this phase answers is "how does this place work, and who do I go to".
- The required training: Anything the business or the regulator says every employee completes — conduct, data handling, health and safety. Build it once, assign it automatically on the start date, and stop rebuilding it per hire.
- Systems and access: Accounts, tools, and the policies attached to each. The most common first-week complaint is not culture, it is not having a login.
- The map: Who does what, and which name goes with which question. Worth writing down: the alternative is that the new hire builds this map by interrupting people.
Days 31–60: the role
Owned by the hiring manager. The question is "what does good look like in this job".
This is the phase generic onboarding content cannot touch, and where most programmes quietly stop. The specifics — the systems this team uses, the accounts it handles, the way work gets reviewed — are known by the manager and nobody else, which means the only way this phase gets covered is if someone is accountable for it and has time booked.
One thing worth formalising: written expectations for the first review, given in week five rather than discovered in month six. A new hire who knows what they are being measured on will usually get there.
Days 61–90: independence, and a check
Shared. The question is "is this working, for both sides".
Two conversations belong here. The manager assesses whether the person is where the role needs them, against the expectations set in week five. Separately — and this is the one usually skipped — someone who is not their manager asks the new hire what has been harder than it needed to be.
That second conversation is the only reliable source of onboarding improvements, because it is the only point where someone still remembers what was confusing and is not yet invested in defending the process. Six months later they will have forgotten, and a year later they will be running it.
What to build once, and what to keep human
The split is not about technology, it is about repetition. Anything identical for every hire — required training, policies, systems walkthroughs — should be built once and assigned automatically, because doing it live is a manager repeating themselves at forty minutes a head.
Everything that differs per person stays a conversation. Attempting to systematise it produces the onboarding module everyone clicks through, and the new hire still asks a colleague.