Most operations can tell you how long onboarding takes. Far fewer can tell you why that number hasn't moved in five years, through better hiring, better software, and a training program somebody rebuilt twice. The answer is usually sitting in the heads of the three people the new hire keeps interrupting.
Why does onboarding still take as long as it did five years ago?
Because the work was never written down, so every new hire has to reconstruct it by asking. You aren't training a person, you're re-transmitting a process from memory, and that takes the same number of weeks every time. Better candidates don't shorten it. They just ask sharper questions.
I watched this for years as the chief financial officer of a manufacturing company, and it looked identical on the floor and in the office. We'd hire someone capable, hand them a binder that described the plant as it existed three years earlier, and then lean on a supervisor to fill the gap in real time. The supervisor was the actual training program. The binder was decoration.
What makes it hard to see is that nothing fails. The work gets done, the new person gets there eventually, and the cost never appears on a line item. It appears in the calendar of the person who had to answer.
What is undocumented process actually costing you?
You're paying two salaries for the same output: the new hire's wages during the weeks before they're productive, and the experienced person's time spent answering instead of producing. Neither shows up as a training expense, which is exactly why the number never gets managed.
The cost hides in ordinary places:
- Wages paid through a ramp that's longer than it needs to be, every hire, every year
- Senior staff interrupted for questions they've already answered for someone else
- Errors caught late, because the new person guessed rather than ask a fourth time
- Two trainers teaching two versions of the same task, so the inconsistency outlives them both
- The exposure you carry when the one person who knows a process takes a week off
Run a rough number on the first two and it's usually larger than the cost of writing the procedures down. Take a role paying fifty thousand a year, hired three times in a year, reaching full productivity in eight weeks instead of five. Those three extra weeks per hire are roughly nine weeks of salary you paid for partial output, before you count a single hour of the supervisor's time. Cut the ramp by half through documentation that already needed to exist, and the arithmetic stops being close.
That calculation is the whole argument, and most owners have never run it, because the inputs stay invisible until somebody writes them in a column.
Why doesn't more training fix it?
Because training delivers information once, at the moment the new hire can least use it, and then expects recall weeks later under pressure. The failure isn't attention or effort. It's that working memory holds very little, and what it holds decays fast without the thing it attaches to.
A person on their second day has no structure to hang new detail on. Show them fourteen steps in a conference room and they'll retain the shape of it, maybe three specifics, none of the exceptions. Put the same fourteen steps in front of them on the day they perform the task, and retention stops mattering. They're reading, not remembering.
That's the quiet advantage of a written procedure. It moves the moment of instruction to the moment of use, which is the only moment that was ever going to work.
Which documents actually move the number?
Start with the task the new hire performs most often in week one, written so they can complete it without a supervisor standing behind them. Frequency beats importance here. The rare, high-stakes procedure feels more urgent to document, but it isn't what's eating the supervisor's day.
In practice, four documents carry most of the improvement:
- Procedures for the five most-repeated tasks in the first two weeks, each one written as a procedure someone can actually follow
- A one-page orientation to where things live, which is the question new hires ask most and the one nobody thinks to answer
- A short decision guide for the two or three judgment calls where a new person will otherwise guess or escalate
- A map of the forms and records the work produces, so nothing has to be reconstructed later
That's a modest amount of writing. It's also the portion that gets asked about forty times a year, which is why it returns faster than a full documentation system built end to end.
How do you get the process out of your experienced people's heads?
Watch them do the work and write down what they do. Asking them to write it themselves is the version that fails. Your best operator is usually your worst documenter, and it isn't a character flaw. It's that expertise compresses steps until they stop being visible to the person performing them.
They write "check the settings and run it," because to them that's one motion. To a new hire it's six decisions, two of which have consequences. I've written procedures beside people who were genuinely surprised at how many steps came out of something they'd described in a sentence. The compression is the expertise. It's also why the person who knows the most explains it worst.
The method that works is boring and reliable: watch the task performed start to finish, write it as observed, then hand the draft to a person who's never done it and see where they stop. Their confusion is the edit list. Nobody's memory needed to be reliable at any point.
How do you know it's working?
Count the interruptions and the days to first unsupervised task. Both are countable, both move within one hiring cycle, and neither depends on how anyone felt about the training. A satisfaction survey won't tell you whether the documentation carried the load.
Ask your supervisors to keep a tally for two weeks of questions they answer that a document could have answered. It's an uncomfortable list the first time, and it's also your writing queue in priority order. Run it again after the procedures exist. The gap is your answer, and it tends to be larger than people expect.
One objection deserves a straight response: some operations say their process changes too fast to document. Occasionally that's true, and the honest fix is a shorter, plainer document with a named owner and a revision date, not no document. More often the process isn't changing fast. It's changing untracked, which produces the same feeling and a different problem.
If you want a read on which documents would move your number first, that's a reasonable thing to work out on a call before anyone writes anything. I've also answered the questions that come up most often about scope, timeline, and how projects are priced on the FAQ.
Onboarding time isn't a training problem that documentation happens to help. It's a documentation problem that training has been compensating for, one hire at a time.
Frequently asked questions
Why does onboarding take so long even with a good training program?
Because undocumented process has to be transferred from memory each time someone new arrives. Training delivers that information once, before the new hire has any structure to attach it to, so most of it decays before the day they need it. A written procedure moves the instruction to the moment of use, which removes the need to remember it at all.
What does undocumented process actually cost a business?
Two salaries for the same output. You pay the new hire through a longer ramp, and you pay experienced staff for the hours they spend answering questions instead of producing. Neither appears as a training expense, so the cost is usually never measured, though it is straightforward to estimate from ramp length and interruption counts.
Which processes should be documented first to shorten onboarding?
The tasks a new hire performs most often in the first two weeks, chosen by frequency rather than importance. Add a one-page orientation to where things live, a short decision guide for the two or three judgment calls a new person would otherwise escalate, and a map of the forms and records the work produces.
How do you measure whether documentation is reducing onboarding time?
Track days to first unsupervised task and the number of questions supervisors answer that a document could have answered. Have supervisors tally those questions for two weeks before the procedures are written and again afterward. Both numbers are countable and both move within a single hiring cycle.
