Every owner I talk to has a folder. Sometimes it is a Google Drive, sometimes a binder in the back office, sometimes a Notion workspace that got a burst of attention one January. Inside are documents with names like Opening Checklist v3 FINAL. Nobody opens them. Everybody still asks the owner.
That is not a discipline problem. It is a design problem, and it has a fixable cause: almost every small business SOP is written in the wrong order, at the wrong length, by the wrong person, at a moment when the work is not actually happening.
Here is what works instead.
Why most small business SOPs fail
Three failure modes, and they are nearly universal.
They are written from memory. You sit down on a Sunday to document how quotes get sent. You write eleven clean steps. What you leave out is the eight little judgment calls you make without noticing: which jobs get a site visit first, when you round up, which customer gets a call instead of an email. Those omissions are exactly where the handoff will fail, because they are the parts that were never in the steps to begin with. They were in you.
They are too long to survive a Tuesday. A fourteen-page document is a reference manual, and reference manuals are for looking things up, not for doing work. When someone is standing at a counter with a customer waiting, they will not scroll. They will ask you. Length is not thoroughness. Length is friction.
They describe tasks instead of decisions. This is the big one. A procedure that lists what to do when everything goes normally is only useful when everything goes normally, and the reason people interrupt you is that things are not normal. If the document does not say what to do when the customer pushes back on price, the customer pushing back on price becomes your phone ringing.
The underlying pattern
Documentation is the cheapest lever on owner dependency, because it is the one that makes every later move possible. You cannot delegate what has never been described, and you cannot automate what has never been delegated. Knowledge is the dimension that unblocks the other five.
What is actually worth documenting
The instinct is to document everything, which is why most owners document nothing. The pile is too big to start.
Use a cost filter instead. Survey data on how owners spend their week is bleak in a specific, useful way: a large share of the working week goes to administrative work like invoicing, data entry, and ordering, and most owners report spending roughly two thirds of their time inside daily operations rather than on growth. That ratio is the target. You are not trying to document the business. You are trying to document the handful of things that are eating the two thirds.
Three questions, in order:
- What breaks first if I vanish for two weeks? Not what would be inconvenient. What would actually stop revenue or lose a customer.
- What do I get asked about most? Every recurring question is an undocumented decision. Track the interruptions for one week and the list writes itself.
- What is expensive to get wrong? Anything touching money, safety, compliance, or a customer's first impression.
Whatever appears on two of those three lists is your starting set. For most small businesses that is five to eight procedures, not sixty.
The record-first method
Do not open a blank document. Blank documents are where accuracy goes to die, because you will fill them with the idealized version of the job.
Instead, the next time you actually do the task, record it. Screen recording if it happens on a computer, voice memo if it happens in a truck or a shop. Narrate as you go, especially the parts where you are choosing: "I'm calling this one instead of emailing because they've been slow to reply twice." Then get it transcribed and cut it down.
This has always been the right method. What changed recently is the cost. Accurate transcription used to be the expensive step, so the record-first approach stayed a nice idea. Now transcription is effectively free and near-instant on any phone or laptop, which moves the bottleneck entirely onto judgment: deciding what matters and cutting the rest. That is the part only you can do, and it is the part worth your time.
One warning about the transcript: it is a raw draft, not a document. Pasting a lightly cleaned transcript into a folder and calling it an SOP produces exactly the fourteen-page unread artifact described above. The cutting is the work.
The six-part SOP template
Cut the transcript into these six fields and nothing else. Copy this structure directly.
Notice what is missing: a revision history table, a scope statement, a glossary, an approvals block. Those belong in regulated manufacturing. In a nine-person plumbing company they are decoration that pushes the useful content below the fold.
Decision rules: the part everyone skips
If you write only one section well, write this one. Decision rules are what convert a set of instructions into someone else's ability to act without you.
A decision rule has a boundary and a fallback. It sounds like this:
- Spend. "Approve any part under $400 yourself. Over $400, text me a photo and the price."
- Discount. "You can take up to 10% off to close today. Beyond that, ask."
- Scheduling. "If we are booked past Thursday and the customer has no heat, bump the maintenance visits, not the installs."
- Refunds. "Under $150, just refund it and note why. Over $150, call me."
Most owners resist writing these, and the reason is usually honest and a little uncomfortable: the number in their head has never been fixed. They decide case by case, by feel. Writing the rule forces you to pick a number, and picking a number feels like giving something up.
It is. That is the trade. A rule you set deliberately, in advance, applied consistently by someone else, beats a perfect judgment call you have to be awake to make. And you can always change the number next quarter once you see how it lands.
A useful test
Read your SOP and count the sentences that contain a number, a threshold, or the word "if." If there are none, you have written a description of a task, not a system that runs without you.
The silent test
Here is the step that separates SOPs that work from SOPs that sit in a folder.
Hand the document to the person who will actually run the procedure. Ask them to do the job from the document alone, while you sit nearby. Then stay quiet. Do not help. Do not clarify. Do not do the thing where you say "here, just let me" and take the mouse.
Every question they have to ask, write it down. Every place they hesitate, write it down. Those are the gaps, and they are almost never the gaps you predicted. When they are finished, you fix the document. You do not coach the person.
Twenty silent minutes will teach you more about your own process than a weekend of writing. It is also uncomfortable in a way that is worth naming: watching someone do your job slightly differently, slightly slower, and get an acceptable result is the exact experience that owner dependency is built to protect you from feeling. Sit in it. That discomfort is the price of the handoff, and it is a one-time cost.
Keeping SOPs alive
Documents rot. Yours will too unless three things are true.
- Every SOP has a named owner who is not you. Unowned documents have no one to notice when reality moves. The owner does not have to be a manager. They just have to be the person who runs it most.
- Every SOP has a review date on it. Quarterly is plenty for most. A stale SOP is worse than none, because it teaches people the document lies, and once they believe that, they stop opening any of them.
- Changing it is easier than working around it. If updating the procedure requires asking you, nobody will, and the workaround becomes the real process while the document quietly becomes fiction. Give the owner permission to edit it directly.
One more habit worth building: when someone asks you a question that the SOP should have answered, answer it, and then update the SOP in the same five minutes. Not later. Later is where the folder came from.
Where this fits
Documentation is step one of three. You map where the business routes through you, you rank those places by what they actually cost in hours and dollars and risk, and then you fix them in order by delegating, documenting, or automating. Writing SOPs before you have done the mapping is how owners end up with sixty documents covering the wrong sixty things.
If you want the number first, the free 18-question assessment scores your dependency across all six dimensions in about five minutes and tells you whether knowledge is even your binding constraint. It often is not. Sometimes the documents are fine and the problem is that nobody is allowed to decide anything, which is a different fix entirely.
Common questions
How long should a small business SOP be?
One page, or one screen of scrolling. If a procedure needs more than that, it is usually two or three procedures wearing a trench coat. Split it. Long SOPs do not get read, and an SOP that is not read is not a system, it is a document.
Who should write the SOP, the owner or the employee?
The person who does the work most often should write the draft, and the owner should edit it. Owners write what the job should look like. Doers write what it actually looks like, including the workarounds. Documenting the real process is what makes the handoff survive contact with a Tuesday.
What should an SOP include?
Six parts: the purpose, the trigger, the owner, the steps, the decision rules, and the definition of done. The decision rules matter most, because they are what let someone act without asking you.
How many SOPs does a small business need?
Far fewer than most owners assume. Start with the five to eight procedures that carry the most revenue, risk, or repetition, and stop. A business with eight living SOPs that people follow is in better shape than one with sixty in a folder nobody opens.
Do this week
- Keep a note on your phone for five working days. Every time someone interrupts you with a question, write it down. Do not analyze it yet.
- On Friday, find the question that came up most. That is your first SOP, chosen by evidence instead of by guess.
- Next time that task comes up, record yourself doing it. Ten minutes, normal speed, narrating your decisions.
- Cut the transcript to the six fields. Stop at one page.
- Hand it to whoever will run it, watch them use it, and stay silent. Fix the document, not the person.