There is a short window, right at the start of someone’s employment, when they can see your plant more clearly than anybody else in the building. It lasts about ten days.

During that window they have not yet absorbed the workarounds. They do not yet know that the sheet says two hundred but everyone runs one eighty. They have no idea which steps are load-bearing and which are ceremony. So they ask. Constantly, and often at inconvenient moments.

Every one of those questions is a gap in your documentation with a timestamp on it.

Then the window closes. By week six they have stopped asking and started compensating, exactly like everyone else on the line. They have joined the club. And the single most valuable diagnostic instrument that will ever walk through your door has quietly turned into another operator who knows things that are not written down.

Why the window closes, and why nobody notices

Acclimatization looks like progress, and mostly it is. But two things are happening at once, and only one of them is good.

The first is knowledge acquisition. The new person learns the job. That is what you hired them for.

The second is gap blindness. They learn to route around the places where the document is wrong, and once routing around a gap becomes automatic, the gap becomes invisible to them. Ask a six-month employee what confused them in week one and you will get a vague, generous answer. Ask a six-day employee and you will get a list.

The information does not survive. It decays fast, and it decays into competence, which is why nobody experiences the loss as a loss.

The numbers make this expensive, not just interesting

A new production worker operates at 40 to 60 percent of an experienced worker’s output for the first 60 to 90 days, according to FirstHR’s 2026 manufacturing benchmark. Replacing one manufacturing employee costs $20,000 to $40,000 when recruiting, ramp and lost output are counted. And 33 percent of new-hire turnover in manufacturing happens inside the first 30 days.

That last number is the one to sit with. A third of the people who leave, leave before they have finished learning the job.

Separately, and this is a cross-industry HR survey rather than manufacturing-specific data, Enboarder found the single most-cited reason people quit inside 90 days is misalignment between the job as described and the job as it actually is. Thirty point three percent. Not pay, not the commute, not the supervisor.

Read that back through a plant lens. “The job as described” is your work instructions, your training packet, your qualification sheet. “The job as it actually is” is what happens at the station. The distance between those two things is being described by HR as an expectations problem. On a factory floor it is a documentation problem wearing an HR label.

Only 12 percent of employees strongly agree their company onboards new people well, per Gallup. Structured onboarding delivers 82 percent better retention, per Brandon Hall Group. Both numbers are usually read as an argument for a better onboarding program. They are also an argument for something cheaper: listening to what the new person cannot find.

Run the ten-day question log

This costs nothing and takes a supervisor about twenty minutes total. Do it with your next hire.

Day one. Hand them a small notebook with one instruction: every time you have to ask a human being something because the document did not answer it, write down the question. Not the answer. The question. Do not filter, do not worry about sounding slow, do not skip the obvious ones. The obvious ones are the most valuable.

Day ten. The supervisor collects the notebook. That is the whole collection protocol.

Then sort what came back into three buckets:

• Missing step. Something is done at the station that appears nowhere in the document. This is a same-week fix. Add it.
• Wrong number. The document says one thing and the floor does another. This is a revision, and it needs a decision: which one is correct? Do not assume it is the document.
• Unwritten judgment. “How do I know when it’s ready?” “How do I tell if that sounds wrong?” These cannot be fixed by editing a sentence. This bucket is your knowledge-capture backlog, and it is the bucket that matters most.

The three-way sort is what makes the exercise useful rather than merely interesting. Missing steps and wrong numbers are administrative. The judgment bucket is where the actual expertise of your plant lives, undocumented, in a small number of heads.

The harder half of the exercise

Run the same interview retroactively on your last three hires. Ask each of them what confused them in week one.

Then compare how much less they can remember than the person currently in week one.

That decay is the finding. It is not a story about memory. It is a demonstration that your plant generates a precise, free audit of its own documentation several times a year and throws every copy away.

What the log actually catches

The questions that come back are rarely dramatic. They are small, specific, and they are exactly the things nobody thinks to write down:

“When do I know the changeover is finished? Nobody starts the next run at the same moment.”

“The sheet has four checks before startup. Everybody on this shift does six. Which is right?”

“What is the noise it makes when it’s about to jam? You said I’d learn it.”

None of these are training failures. All of them are documentation failures that training has been quietly compensating for, sometimes for years.

Why better delivery is not the same as better content

A 2026 randomized field study in the Journal of Manufacturing Technology Management found that digital and projection-based work instructions improved task completion times by 16 to 20 percent, cut errors by 46 to 60 percent, and reduced help requests by 69 to 83 percent compared with paper.

Those are real numbers and the mechanism is real. Better delivery works.

But look closely at the third one. Help requests dropped by up to 83 percent. That is a gain in productivity, and it is also the sound of the question log going silent. If the underlying content is wrong, a better delivery system distributes the wrong content faster, more legibly, and with fewer interruptions to tell you so.

The industry buys delivery. Almost nobody audits accuracy. Onboarding platforms measure completion; learning systems measure hours; qualification matrices measure signatures. Not one of them records what the new person could not find.

What to do with the judgment bucket

The first two buckets you can close yourself this month. The third one is harder, because you cannot fix unwritten judgment by editing the document. You have to get it out of the person who holds it, and that person usually cannot simply write it down, because to them it does not feel like knowledge. It feels like doing the job properly.

That is the work the SOP Sprint is built for: thirty days, on site, five to ten procedures captured from the people who actually run them, with an internal champion trained to keep the process going after we leave. The ten-day question log is the five-minute version you can run yourself, on one hire, starting Monday. The Sprint is the same instinct pointed at the whole line.

Either way, the direction is the point. Stop asking whether your people are following the document. Start asking what the document is missing, and ask it while somebody in the building can still see it.

Frequently asked questions

Why do new manufacturing hires take so long to become productive? A new production worker runs at 40 to 60 percent of an experienced worker’s output for the first 60 to 90 days. Most of that gap is not physical skill; it is the accumulation of undocumented knowledge, the steps, checks and judgment calls that live in experienced people’s habits rather than in the written work instructions. The new hire is effectively learning two jobs at once: the documented one and the real one.

What should be in manufacturing work instructions for new hires? Beyond the sequence of steps: the checks people actually perform before starting, the sensory cues that signal a problem, the decision points where judgment is required and what informs it, and the current correct values rather than the last-revised ones. If a new hire has to ask a human for it, it belongs in the document.

How do you find gaps in your SOPs without running a formal audit? Give your next new hire a notebook and ask them to record every question they have to ask a person because the document did not answer it, for their first ten days. Sort the results into missing steps, wrong numbers, and unwritten judgment. The exercise costs nothing, takes about twenty minutes of supervisor time, and produces a gap list that no internal audit will surface, because everyone else in the building has stopped being able to see the gaps.

How long is the window before a new hire stops noticing gaps? Roughly ten days for the sharpest observations, and largely closed by week six. Once routing around a documentation gap becomes automatic, the gap becomes invisible to the person routing around it.

About the author: Diego Echenique is CEO and co-founder of SenseiLab, which does knowledge capture and Living SOPs for manufacturers in regulated, knowledge-intensive industries. He has spent over twenty years in manufacturing operations across the Americas, Europe, the Middle East and Asia, as a Lean and Six Sigma practitioner and in supplier development and plant leadership roles.