Marcelo had been running the same press for eleven years. I asked him to walk me through the changeover and he did it in nine steps, clean and fast, without looking at anything.

Then I pulled the SOP off the post behind him. Fourteen steps. Three of them referenced a fixture that had been scrapped two years earlier.

Nobody on that floor was being careless. The quality manager had a signed, revision-controlled procedure. The operator had a process that worked. They were simply two different jobs wearing the same document number.

If you want to understand why SOPs fail in manufacturing, start there. Not with compliance. With the distance between the paper and the press.

The comfortable explanation, and why it’s wrong

Ask around and you will hear the same diagnosis: people do not follow the SOP because they were not trained properly, or because they think they know better, or because the culture is loose.

That explanation is comfortable because it puts the problem in the operator.

It also does not survive contact with the floor. The operators I meet are not improvising out of arrogance. They are working around documents that stopped describing reality somewhere between the last process change and the last revision date. When the written standard and the real standard disagree, an operator who follows the document produces scrap. So he follows the process that works, and the document becomes decoration.

The Manufacturing Institute’s 2026 workforce survey found that manufacturers spent roughly $32 billion on training last year, an average of 47.6 hours per employee, up from $26.2 billion and 42.9 hours in 2019. Investment is not the missing piece. We are pouring record training hours into instructions that are quietly going out of date.

Standards drift: how the gap actually opens

Standards drift is rarely a single event. It is an accumulation.

A tooling change gets made on a Tuesday. It works better, so it stays. Nobody updates the procedure, because updating the procedure means a form, a review, an approval, and two weeks of somebody’s attention.

A month later an operator finds a sequence that cuts a minute off the cycle. Also better. Also undocumented.

A new machine arrives with a different control interface. The old work instruction still shows the old screen.

Each individual change is small and sensible. Together they open a gap that nobody can see, because the only two things being compared are the document to itself, revision to revision. The real process is never in the comparison.

Then one of three things happens. An auditor finds the gap and it becomes a finding. A new hire follows the document and produces a defect. Or the operator who carries the real process in his head takes a job across town, and the gap becomes permanent.

The three failures behind almost every dead SOP

It was written by the wrong person. Most procedures are written by an engineer or a quality specialist who observed the job, or by a consultant who interviewed someone about it. Observation catches sequence. It does not catch judgment. The operator knows when to back off the feed, what the machine sounds like an hour before it drifts, which fixture reads true and which one lies. That is the content that prevents defects, and it almost never makes it onto the page, because nobody asked in a way the operator could answer.

It was built to be filed, not used. A fourteen-page controlled document written in passive voice, laid out for an audit binder, is not a tool anybody reaches for at a workstation. If following the standard is slower than asking the guy next to you, the standard loses. Every time.

Updating it costs more than ignoring it. This is the one that does the real damage. When the cost of correcting a procedure is a form plus three signatures plus a two-week wait, the rational move for a busy supervisor is to leave it alone. Multiply that decision across a year and the document set silently detaches from the plant.

Notice that none of these three is a discipline problem. They are all design problems.

What actually fixes it

The fix is not a better template, and it is not more enforcement. It is changing who the standard comes from and how it stays current.

Capture from the people who do the work, in their own words. The knowledge that keeps a process stable is tacit. It lives in habit and in judgment, and it does not come out through a form. It comes out when someone experienced is asked to explain what they are doing while they do it, and asked follow-up questions when they skip a step because it is obvious to them. That is what tribal knowledge capture actually looks like: a conversation, not a questionnaire.

APQC’s The Great Retirement research, published in 2025, found that 41% of organizations rarely or never try to collect know-how from departing employees, and that 52% name day-to-day deadlines as the top barrier to knowledge transfer. The intent exists. The mechanism does not.

Make the standard cheap to change. If a procedure can be corrected in the moment it is found to be wrong, by the person who found it, drift stops accumulating. The moment updating costs more than working around, you are back where you started.

Close the loop where the work happens. A standard that only gets reviewed on an annual cycle is guaranteed to be wrong for most of that year. The review has to happen continuously, at the workstation, as part of doing the job.

At SenseiLab this is the whole design of our Smart Living SOPs. Operators talk directly to our AI agent about how they actually run the job, and that conversation becomes the standard. When the floor changes, the standard changes with it. Drift gets caught inside the work rather than discovered during an audit, and the procedure keeps evolving after we leave instead of freezing on our last day on site.

Something you can do this week without buying anything

Take one SOP. Not your worst one, and not your showcase one. A normal one.

Hand it to the operator who runs that job and ask him to mark every step he does differently, and every step he does that is not on the page at all.

Do not make it an audit. Make it clear you are checking the document, not the person.

Then count. That number is your real measure of process control, and I have never once seen it come back as zero. What comes back in the margins, the steps he adds that were never written down, is the most valuable documentation your plant does not currently own.

FAQ

Why do employees not follow SOPs? Most of the time because the SOP no longer matches the process. When the written procedure and the working method disagree, following the document produces worse results, so operators follow the method that works. Discipline is rarely the root cause.

What is standards drift? Standards drift is the widening gap between a written procedure and the way work is actually performed. It builds up through small undocumented improvements, tooling changes, and equipment swaps that never make it back into the document.

How often should SOPs be reviewed? Annual review cycles guarantee the document is wrong for most of the year. The practical answer is continuously: the standard should be correctable at the moment somebody discovers it is wrong, by the person who discovered it.

What is a Living SOP? A Living SOP is a procedure that updates as the process changes, captured directly from the operators who run it, rather than a static document frozen at its last revision date.

Who should write an SOP? The operator who performs the work should be the source. Engineering and quality shape structure and compliance, but the judgment that prevents defects only exists in the person doing the job.

If you want to see how far your written standards have drifted from your real ones, that is where our 30-day SOP Sprint starts. Book a call.

Diego Echenique is the founder of SenseiLab. He has spent more than twenty years in manufacturing operations across the Americas, Europe, the Middle East and Asia.