Nobody follows your SOPs? The problem is where you wrote them

"Thirty-some pages, re-edited last year, and not one person on the floor has opened it." I've heard this many times. SOPs usually fail not because of discipline, but because you wrote them outside the workflow instead of inside it. This post breaks down three layers of writing SOPs: what should become fields and hard stops, what can only stay as text, which work should never get an SOP, and what to do now with that thirty-page document nobody reads.

Johnny Yang11 min read
Nobody follows your SOPs? The problem is where you wrote them

"We do have SOPs. Thirty-some pages. We even paid someone to re-edit it last year."

I've heard this line in meeting rooms many times. Usually the owner follows up with something angrier: "But go ask on the floor. Not one person has ever opened it." Then the conversation slides in the same direction. Should we tie it to performance reviews? Hold a meeting to push it? Haul the managers in and chew them out?

I'm Johnny Yang, Brightstream Technology, Taiwan. I've led implementation teams for seven years, across 252+ projects in 32+ industries. I've seen this scene enough times that I'll give you the conclusion first: When nobody follows an SOP, it is almost never a staff discipline problem. It's that you wrote the SOP outside the workflow. Between a document sitting three folders deep on the shared drive and the screen a person is working on right now, there are three clicks, one file search, and one "forget it, I roughly remember how." People don't leave the task in their hands to look up a document. That isn't bad attitude. That's human nature. So there is only one form of SOP that actually gets followed: it has to live inside the tool. It becomes a field, a required input, a dropdown, a hard stop, a reminder that goes out automatically when time is up. The parts that can't live inside the tool mostly shouldn't be SOPs either. Write them as principles, or admit honestly they don't need writing at all.

First, you're right: the instinct to write SOPs is correct

You usually start thinking about SOPs because you saw something specific: three people handle the same return slip three different ways; a senior employee takes three days off and the whole line stalls; a new hire is two months in and still asking "who do I ask about this."

That observation is right. The diagnosis is right too. Your company keeps too much in individual employees' heads, and heads take leave, quit, and misremember. Up to this point, your judgment is completely correct.

The next step is what goes wrong. Most owners' next step is "then let's write it down." They have someone spend two months compiling a manual, print it, hand it out, hold one meeting to push it, done. Three months later the manual is in a drawer, the actual process has changed twice, and item 14 in the manual stopped matching how things are done long ago.

The problem isn't writing. It's where you wrote it. You moved the knowledge out of people's heads, but you moved it somewhere nobody ever passes by.

Document SOPs always die, and they die the same way

Every unused SOP I've seen failed in ways that fall into almost exactly three points.

First, nobody can find it. The person doing the work is on their own screen. The SOP is in another folder. Jumping out to check once per action, for something they do thirty times a day: nobody jumps thirty times. You think they're lazy. Actually they did the math. The cost of getting it wrong from memory is lower than the cost of checking every time.

Second, versions drift. The process changes. People change how they do it, and nobody remembers to go back and update the document. After three months, the document and reality start to split. After six months, the document is a historical record of "how we used to do it." Then it gets worse: following the document actually produces mistakes, so experienced people deliberately stop reading it, and the SOP's credibility goes completely bankrupt.

Third, the level of detail is wrong. The document says "confirm inventory, then issue the shipping order." But where the person on the floor gets stuck is "this customer has two delivery addresses, which one do I pick." An SOP describes the skeleton of a process. People always get stuck at the joints. However detailed you write it, you can't keep up with how fast the floor changes.

These three have one thing in common: none of them gets fixed by "emphasizing it one more time." Ten more meetings and ten more review criteria won't change the structural problem that the document and the work aren't on the same screen.

What "SOPs living in the tool" actually looks like

Abstract talk is useless, so I'll break it into three layers. You can check your own processes against them directly.

Layer one is prompts. This is the lightest. It means putting the text in the person's line of sight. The small helper text next to a field, the notes area on a form, the names of the options in a menu: all of these count. Say the SOP says "for complaints, determine whether it's a quality issue or a usage issue." In the system, that's a "issue type" dropdown, and the options are those few types. Nothing to memorize. This layer is the cheapest to build and often works best, because it turns "have to remember" into "just pick from the list."

Layer two is constraints. This turns rules into the system blocking you. SOP item 4 says "a complaint must be assigned an owner within 24 hours." In the system: the ticket can't be saved without an owner, and once time runs out it jumps to the manager's to-do list automatically. The purchasing SOP says "amounts over a certain threshold need manager sign-off." That's a branch in the approval flow. No signature, no getting through. The power of this layer is that it makes "not following it" something you can't do, rather than something you shouldn't do.

Layer three is automation. The most complete SOP is one people don't even need to know exists. A form's status changes, and whoever needs to be notified gets notified automatically. Month-end close arrives, and the reports that need generating get generated automatically. This layer is the only SOP that truly never gets forgotten and never has version problems.

By now you should notice something: once you build these three layers, writing SOPs has actually become system configuration work. That's a good thing, but it has a cost, and I have to be honest about it. Rules written into the system are slower to change than a Word file, and you usually need someone's help to change them. So deciding which rules are worth locking into the system and which aren't becomes the core decision in writing SOPs, not "write everything down."

Which work should never get an SOP

This is probably the least popular section of the whole post, but I'm saying it anyway. Most of the projects I've worked on share the same waste: they wrote SOPs for things that shouldn't have SOPs, the whole manual got diluted, and the five items that actually mattered drowned among fifty.

For things done only a few times a year, don't write an SOP. A checklist is enough. Annual stocktaking, organizing records before tax filing. The problem with these isn't "the steps are hard." It's "it's been so long you forget." What you need is a list you tick off this year and pull out to check against next year, not a process document. The difference: a checklist only answers "did we miss anything," an SOP has to answer "how do we do it," and for a once-a-year task, how you do it will be different every year anyway.

For things where the conditions differ every time, write principles and boundaries, not steps. Whether to give a discount on a custom-built quotation, whether to make an exception for a long-time customer's special shipment, whether this rush order jumps the queue. These are judgment calls, not process questions. Write them as steps and you'll force one of two outcomes: everyone follows it and makes dumb decisions, or everyone goes around it and never trusts SOPs again. What you should set here is the boundary: within what range sales decides on their own, and past it, who they ask. Three lines are enough for a boundary, and it's more useful than three pages of process.

For a process that changed three times in the last six months, don't write it yet. Writing an SOP while the process is still moving is taking a photo of a moving object. Let it stabilize first, or more honestly, first confirm whether it should even look like this. I've seen plenty of companies carefully write a clearly broken process into an SOP, which amounts to stamping the chaos with an official document.

For work only one person does, with no plans to change that soon, the return is very low. Saying this will offend people, but it's the truth. Unless that person is about to leave, or you plan to move them to other work, spending two weeks writing their job into an SOP gets you a document they won't read and nobody else will use. If you're genuinely worried about the risk, what you should do is keep that work's data in the system, not keep that person's method in a document. If the person leaves, it's fine if the method leaves with them. As long as the data is still there, you can rebuild it.

Type of workWhat should carry itReason
High frequency, fixed rulesSystem fields, required inputs, hard stops, automatic notificationsDone every day, worth locking in; once locked in, no memory needed
Low frequency, fixed rulesA checklist, kept where that task actually happensThe problem is omission, not not knowing how; a checklist costs the least
High frequency, judgment each timeBoundaries and authority ranges, done in three to five linesWriting steps only pushes people to go around them
Processes still changingDon't write it yet; let it stabilize firstPhotographing a moving object is pointless

The right order for writing SOPs: walk the process, set up fields, write text last

If you buy the logic above, the order of work for writing SOPs is the exact opposite of what you originally had in mind.

Pick a process first. Don't do everything at once. Pick the one that goes wrong most often, or gets stuck most often, and touch only one at a time. I break this down in more detail in "Which process should you digitize first?". The key point: don't try to standardize the company's entire set of processes in one go. Every project like that I've seen fizzled out after six months.

Next, walk it in person. Don't draw a flowchart from memory in a meeting room. Go sit next to the person who does the work and watch them do it. Note two things: where they pause to think (those are judgment points), and where they copy data from one screen to another (those are break points). These two kinds of points are what an SOP actually needs to deal with. The other steps, you honestly don't need to teach.

Only then do you sort. Among judgment points, the ones with fixed rules become dropdowns or hard stops, and the ones without fixed rules get written as boundaries. For break points, find a way to have the system connect every one of them. Every manual copy is a place that will definitely go wrong in the future, and when it does, you won't even be able to trace who copied it wrong.

Writing text comes last. And don't put that text in a separate document. Stuff it back into the system: into field descriptions, into form notes, into the names of menu options. There's only one principle: the text has to appear on the screen where the person will need it, not in a file they have to go open separately.

One related trap while I'm at it. Some companies have complete SOPs and the system is live, but the data is still wrong, because people on the floor backfill entries after the fact to make the numbers look good. That has nothing to do with SOPs. The reporting mechanism was designed wrong. I covered how to diagnose it in "Everyone logs production, but the numbers are still useless?".

What to do with that thirty-page old SOP now

Don't rewrite it. Take it apart first. Rewriting is the move that feels most satisfying and does the least.

Take a pen, start at page one, and sort every item into three piles. Pile one is "this can become a system setting." This pile will later disappear into fields and workflows and no longer exist as text. Pile two is "this can only stay as text." Usually judgment principles, external messaging, things required by regulation. Keep this pile, but move it somewhere people can see it. Pile three is "nobody actually does this, or doing it makes no difference."

In most of my projects, pile three ends up being more than half. The first time owners see this, they're usually uncomfortable. It's something they spent money and time on in the past. But look at it from another angle: that half of the items existing is one of the reasons the other half never got followed. A thirty-page document nobody reads versus three pages of boundaries everyone can remember: the second has far more real binding force.

Once it's taken apart, you'll have a very short list that says "these items need to become this in the system." That's the work you actually need to do, and that work has an end point.

So what should you do tomorrow

Tomorrow, find someone who runs the process that goes wrong most often at your company, and ask them one question: "The last time you did this, whose instructions did you follow?"

If the answer is "I remember," that SOP is dead. If the answer is "I ask so-and-so," then your company's SOP is actually that person, and when they take leave, you stop. If the answer is "the system blocks me" or "the system reminds me," congratulations, that one is alive. Go see how it was made alive, then apply the same approach to the next one.

After asking, you'll have a very honest list: which items are alive, which are dead, and which don't need to exist at all. That list is worth more than any re-edited manual, and you can start asking this afternoon.

If you're halfway through taking apart your SOP and aren't sure what to lock into the system and what to let go, you can talk with me for thirty minutes and I'll help you sort the three piles.

Found this useful? Pass it on
30 minutes · free

Bring the one thing you are stuck on — a free 30-minute diagnostic

You do not need to know what system you want. Tell Johnny Yang about the one thing that keeps blocking you, and in 30 minutes you will know whether to skip it, build it with coaching, or pay for a custom build. No pitch — if we cannot help, you will hear that straight.

Johnny Yang replies himself · limited spots each month · or just book a free 30-minute diagnostic

Johnny Yang (熊董)
Johnny Yang (熊董)
Founder, Brightstream Technology

Serial founder leading a small team that builds systems into an owner's everyday work.

As a founder
7 yrs
Projects delivered
252+
Industries served
32+
About Johnny