The Operator's System: running a business like a system
The Operator's System

How Do I Know If a Process Is Actually Being Followed?

Eleven procedures, all reporting full adherence, none of it checkable. A procedure that leaves no trace is unverifiable by design, and that is the defect.

The Main Street Operator · August 3, 2026 · 5 min read

You ask your service manager if the procedures are being followed. He says yes. You ask two techs in the parking lot. Also yes, and both of them mean it.

You've got 11 written procedures, and everyone you ask says they follow all of them.

And you still don't know anything.

Stop asking, start reading

Here's how you actually find out whether people follow a process: read the evidence the process leaves behind. Any procedure worth keeping produces a trace. A filled field, a photo, a signature, a timestamp. Reading 10 of those takes two minutes and doesn't depend on anyone's memory or goodwill.

Now go looking, and suppose nine of those 11 procedures leave no trace at all. Nobody in the building is lying. There's just no way to know.

Three ways to check, and how much each one lies

They aren't equally good, and the cheapest one is the worst.

The instrument The verdict Why
Asking Free and nearly useless People report their intention rather than their behavior, and nobody is lying when they say yes to a procedure they follow most of the time
Observing Expensive and self-defeating A crew being watched follows the procedure that day, which tells you what they can do rather than what they do
Reading the trace Cheap once designed, and honest The evidence was created during the work, before anybody knew it would be read

The middle one is worth a word, because it's where owners go once asking stops feeling good enough. Two hours of ride-alongs gets you one very compliant day and zero information about Tuesday.

In high-consequence work (aviation, say, or a hospital) the rule is blunter. A step with no recorded evidence is treated as a step that didn't happen, and that's a rule about how the procedure is built. Nobody there takes it as an accusation.

Build the evidence into the work

The trick is making the evidence something the next person in the chain already needs. Then it gets created because the work requires it, and nobody has to feel checked.

Your closeout procedure may already be traceable without anyone noticing. If the invoice can't go out without the parts list, a same-day invoice is proof the closeout ran.

A safety-check procedure is often another story. If it has no output at all, a tech doing it perfectly and a tech skipping it leave identical records. The fix is one field on the job ticket the dispatcher already reviews: ten minutes to add, and an uncheckable rule becomes a countable one.

So ask this of every procedure you've got: if it ran correctly, what would exist afterward that wouldn't exist otherwise? If the answer is nothing, the procedure isn't finished yet.

How it's written matters too. A procedure loose enough to be followed several different ways won't leave a consistent trace, and a process a new hire can actually run is the bar to hold it to.

Two minutes a week

You don't audit everything. You read one procedure a week, on rotation:

  1. Pick one procedure. Rotate through the list, so an 11-procedure business comes back around roughly quarterly.
  2. Pull the last 10 times it ran. Ten's enough to see a pattern and few enough to read standing up.
  3. Count them. How many have the trace and how many don't. Save the explanations for later.
  4. Read the misses, not the hits. Three misses out of 10 is a question about those three jobs, and the answer is usually one situation the procedure doesn't cover.

That's the whole thing, and it fits inside a weekly review you already hold. Two minutes a week beats a quarterly audit that eats an afternoon and gets rescheduled twice.

The misses are the useful part, and they usually point at the procedure more than the people. A miss that keeps coming back is a rule that stopped matching the work, which is how procedures actually die, more than neglect ever does.

Whether a procedure got adopted in the first place is a separate question with a human answer. Setting a standard that sticks happens before there's anything to check.

No trace is the finding

A procedure you can't verify has a design problem, and that's the more useful thing to find.

It means somebody wrote it as a description of good behavior instead of as a step in a chain of work. Descriptions of good behavior are exactly what gets ignored, since nothing downstream depends on them.

The repair's usually small. One field, one photo, one signature, one timestmp, hooked onto something someone else already reads. Do that and follow-through often goes up without a single conversation about follow-through, because skipping the step now blocks the next person.

There's an exit reason to care, too. A business where the owner is the only way anything gets verified runs on a person, and a buyer discounts a business that runs on a person.

The bigger frame is running the business as a system, and checking is the part most owners skip.

Some traces are worth reading for more than adherence. Customer records are the obvious one: they tell you whether the account procedure ran and also how concentrated your revenue really is.


Go down your written procedures this week and ask each one what would exist afterward if it ran. The ones with no answer are this month's work.

If you want to see how much of the business still gets verified by you personally, that's what Keystone's diagnostic measures, and it's closed while we finish building it. Join the waitlist at https://app.trykeystone.io.

When it opens, it's 18 questions and about four minutes, and it returns three scores and your value range with the most likely figure locked, calibrated against 10 years of BizBuySell Insight Reports and 1.6M+ SBA 7(a) loan records.

Adding a trace field is a ten-minute change. Rebuilding a whole set of procedures so each one produces evidence somebody downstream needs is the operating layer, and that's months of work.

A Full Operations Modernization installs that layer in a live system your team runs. The Full Operations Modernization page is where scoping begins.

FAQ

How do I know if my employees are following a process?

Read the evidence the process leaves rather than asking about it. Pull the last 10 instances of one procedure and count how many produced the expected trace, which takes about two minutes and does not depend on anybody's memory.

Why do employees say they follow SOPs when they do not?

Because they are reporting intention rather than behavior, and most people do follow the procedure most of the time. Asking is the cheapest instrument available and the least reliable, and watching is worse because a crew being observed complies that day.

How do I audit a small business process?

Do not run audits, run a two-minute weekly read of one procedure in rotation. Count the last 10 instances, read only the misses, and treat a recurring miss as a rule that stopped matching the work rather than a discipline problem.

What if a procedure leaves no evidence at all?

That is the finding. An unverifiable procedure is a design defect, and the repair is to add one field, photo, signature or timestamp attached to something somebody downstream already needs.

When reading is not the constraint

Some of this you can install yourself. Some of it takes two weeks and a second pair of hands.

Full Operations Modernization is one done-with-you engagement: every core process analyzed on-site or through a deep remote intake, the full document set built into your live Keystone tenant, the core systems wired, and two weeks of support while your team puts it in place. Selective, by application.

See how the engagement works

Prefer to start on your own, for nothing? The operator library is twelve tools, free to download.

The Main Street Operator covers the operating mechanics behind business value: what buyers actually pay for, what discounts a business, and the month-by-month decisions that compound.