Most onboarding fails in the first four days, and almost nobody notices until month three. The new person is polite. They nod in meetings. They say “still getting up to speed” when you ask. Then the ninety-day mark arrives, they are producing about a third of what you hired them for, and you cannot tell whether you hired badly or trained badly.
You trained badly. Almost always. Hiring is a bet placed once; onboarding is a process you run, and processes that nobody owns produce whatever they produce.
This is the version that works for a small team — one to fifty people, no HR department, no learning-management system, the manager doing it personally. It takes about six hours of your time spread across a month. It is not about making someone feel welcome, though it does that. It is about getting them to the point where work leaves your desk and does not come back.
Before day one: the four things that must exist
If these are not done before they walk in, the first week is spent on logistics and you have burned 25% of the ramp.
- Access. Email, chat, the shared drive, the two or three tools they will live in, and the password manager. Every account created and tested by someone logging in as them. “IT will sort it Tuesday” costs you a week of momentum and signals that they are not expected.
- A named first project. Not a category of work — a specific, finishable thing with a deadline inside the first three weeks. More on choosing it below.
- A written scorecard. One page: what “good” looks like at 30, 60 and 90 days, in outputs rather than behaviours. “Runs the Monday numbers unaided” not “demonstrates ownership.”
- A buddy who is not you. Someone at their level who answers the questions people are embarrassed to ask a manager. Ask that person directly and make it an actual assignment, not a favour.
The scorecard: write it before they start
The scorecard is the whole playbook compressed onto one page. Write it before the offer goes out if you can, because if you cannot write it, you do not yet know what you are hiring for — and no amount of onboarding fixes that.
Structure it in three columns.
- Day 30 — can do it with help. They can complete the core recurring task with you reviewing it. Example: “Drafts the weekly client update; I edit it before it sends.”
- Day 60 — can do it alone. Same task, no review. “Sends the weekly client update. I read it after it goes out, like everyone else.”
- Day 90 — can improve it. They have changed something about how the work is done and can explain why. “Has proposed one change to the update format based on what clients actually reply to.”
Three to five lines per column. Any more and it stops being a scorecard and becomes a job description, which nobody reads twice. Hand it to them on day one and tell them plainly: this is the document we will both be looking at in a month.
Choosing the first project
The single highest-leverage decision in onboarding is what you hand them first. The right first project has four properties:
- Finishable in under three weeks. They need to complete something. A person who has shipped one thing behaves differently from a person who is still “ramping.”
- Real, not a sandbox. Practice work is transparent and demoralising. Give them something that would otherwise have to be done by you.
- Low blast radius. If they get it wrong, it costs a day, not a customer. Internal-facing, or external with a review gate.
- Touches four or five people. The fastest way to learn an organisation is to need something from it. A project that forces them to ask five colleagues for input teaches the org chart better than an org chart does.
The classic good first project: take a process you currently do by hand and have them document it, then run it once under supervision. They learn the work, you get documentation you have been avoiding writing for two years, and the output is genuinely useful. This is the mirror image of documenting your own job so you can leave the room — and it works for the same reason.
The shadow–pair–own ladder
Do not explain a process. Nobody learns a process from an explanation. Run this three-step ladder on every recurring task you want them to take over:
- Shadow. They watch you do it once, start to finish, while you narrate the decisions — especially the ones you make without thinking. Their only job is to write down every moment they did not understand why you did something. That list is the real curriculum.
- Pair. They do it, you sit alongside. You are allowed to answer questions. You are not allowed to take the keyboard. This is harder than it sounds and it is the step most managers skip.
- Own. They do it alone and send you the output. You review the output, not the method.
One rung per week per task, roughly. If a task is still on rung two after three weeks, the problem is usually that the task has undocumented judgement calls in it — which is worth knowing, because it means you cannot delegate it to anyone yet.
The meeting cadence — and the script
Front-load contact time and taper it. Daily 15 minutes for week one, twice a week for weeks two and three, weekly from week four. Ending a daily check-in is a promotion, so say that out loud when you do it.
The check-in script is three questions, in this order:
- “What did you finish since we last spoke?” Outputs first. It sets the norm that finishing is what counts.
- “What are you stuck on that you have not asked anyone about yet?” The second half of that sentence is the important half. New people sit on questions for days because asking feels like admitting something. Naming the behaviour makes it askable.
- “What did you see this week that seemed odd?” A new hire has roughly six weeks of useful outsider vision before they normalise to your dysfunction. Harvest it deliberately. Write the answers down. Some of them will be the best process feedback you get all year.
The 30-day conversation
Book it on day one for day 30. Thirty minutes, scorecard on the table, and you go first.
Open with a direct read: “Against the day-30 column, here is where I think you are. Two of these are solid, this one is not there yet, and here is what I think is in the way.” Then hand over: “Where do you think I am wrong?”
Then the question that does most of the work: “What is slowing you down that I control?” Access they still do not have, an approval they are waiting on, a meeting they should be in, a document that does not exist. Fix at least one thing from that answer within 48 hours, visibly. It teaches them that raising problems produces results, which is the habit you actually want.
If someone is clearly behind at 30 days, say so at 30 days. The kindest version of a bad hire is a short one, and the cruellest is silence until month five. If you have run the ladder properly you will have evidence rather than a feeling, which makes the conversation survivable for both of you.
Failure modes
- The firehose week. Eleven meetings, forty documents, six tool demos. Retention is near zero and the new person concludes that looking busy is the job. Cap week one at three meetings a day.
- The invisible manager. You hired them because you are overloaded, so you have no time to onboard — and then you stay overloaded, permanently. Six hours in month one buys back hundreds later. Put it in the calendar before they start.
- Vague first assignments. “Have a look at our marketing and share your thoughts” is a test with no answer key. It produces a deck nobody uses and a person who has learned that their work does not matter.
- Taking the keyboard. Doing it yourself because watching is painful. You will save eleven minutes and lose the transfer.
- Waiting for the annual review. Feedback that arrives more than a week after the work is archaeology.
- Onboarding only to the job. Nobody explains who decides things, which meetings are real, which policies are enforced. Say the unwritten rules out loud in week one; they will otherwise take four months to infer and get two of them wrong.
Our take: Onboarding is the highest-return management work there is and it never gets scheduled, because it has no deadline and nothing breaks visibly when you skip it. The cost shows up as a person who takes six months to reach a level they could have hit in eight weeks — and, often, as a resignation at month seven that everyone calls a “bad fit.” Six hours, one page, one real project, three rungs. That is the whole system, and it beats every onboarding portal ever built.
The one-page version
- Before day one: all access tested, first project named, scorecard written, buddy assigned.
- Week 1: shadow the core task. Three meetings a day maximum. Daily 15-minute check-in. First project starts by Wednesday.
- Weeks 2–3: pair on the core task. Check-ins twice weekly. First project ships.
- Week 4: they own the core task. Weekly check-in. The 30-day conversation with the scorecard on the table.
- Every check-in: what did you finish, what are you stuck on that you have not asked about, what looked odd.
- Every 30 days: what is slowing you down that I control — then fix one of them inside 48 hours.
Print the scorecard. Book the 30-day conversation before they start. Everything else is detail.