You wrote the process down. The team doesn't open it, and the same questions still land on you. A better manual won't fix that. A system gets used when it is a short list of the steps people actually skip. It is written with the person who does the job and kept where the job happens. One person owns it, anyone may stop the work when a line is missed, and it changes every time someone has to ask. By the end you'll have three things: the six steps that turn one document into a list your team uses, a test for which kind of list fits each job, and one weekly number that shows whether it's working.
Six steps turn a written process into one your team uses
Do these for one job before you touch the next. Each step says what to do, how you know it worked, and why.
1. Start with the job that keeps coming back to you
Do this. Look back over the last 30 days. List every redo, every customer complaint that reached you, and every question your team brought you. Group them by job. Pick the job with the most.
It worked when you can name one job and say how many times it came back to you last month.
Why. A job that comes back to you is a job where the current way fails, and it's the cost you already feel. One owner put the pattern this way: "the customers would call me to complain," so "I had to check it last." That is the job to start with, not the one that is easiest to write.
Example. At Johns Hopkins, Peter Pronovost didn't start with a hospital-wide manual. He started with one failure: infections from central lines, the tubes put into large veins 3.
2. Write down only the steps people skip, in the kind of list that fits the reader
Do this. Watch the person who does the job do it once, start to finish. Don't correct anything. Mark every place they pause, guess, skip or redo. Those marks are your list. Leave out everything they did right without thinking.
It worked when the list fits on one page, has no more than nine lines, and every line is something someone has really missed.
Why. A long list is hard to use in the middle of a job. Daniel Boorman, who wrote checklists for Boeing, gave the test, as Atul Gawande reports it: bad checklists "turn people's brains off rather than turn them on," and good ones give "reminders of only the most critical and important steps" 89. Boorman keeps lists to "the killer items"; a common rule of thumb is five to nine 810.
Pick the kind of list. Boorman describes two kinds 810. A "read-do" list is followed line by line, like a recipe. A "do-confirm" list lets people work from memory, then stop at a set point and check that nothing was missed. A full manual works like a read-do list, so people who know the job tend to skip it, and with it the few steps that matter. That is one likely reason your best people ignore the manual. Match the list to the reader.
| Use it when | Who it's for | How it reads | |
|---|---|---|---|
| Read-do list | The order matters, or the job is rare | New people | Do each line as you read it |
| Do-confirm list | People know the job but skip steps when busy | Experienced people | Work from memory, then stop and check |
| Answer templates | The same few questions keep coming in | Anyone who answers customers | Sort the questions into a few groups, one answer each |
| An owned result | The job needs judgement every time | Senior people | One result they answer for, plus the few lines that must never be missed |
For most repeat work in a business with a team, the do-confirm list is the right default. It respects what people know and catches what they forget.
Example. In October 1935, Boeing's Model 299, the prototype of the B-17, crashed on a test flight. The crew had forgotten one routine step: the gust locks on the elevators and rudder were never released. The modern pilot's checklist grew directly out of that crash 7.
3. Write it with the person who does the job, and let them change their part
Do this. You bring two things: the result the job must produce, and the lines that must never be missed. The person who does the job writes the order and the wording. Then give them the right to change their part without asking you.
It worked when they can explain each line in their own words, and within two weeks they have changed at least one line.
Why. People defend what they helped write. And no list can cover every case, so the person doing the job has to own what the list leaves out.
Example. As the central-line checklist spread to other hospitals, Pronovost said the lists were probably "about 95 percent the same." The 5% each hospital adds is the "special sauce" 3. The WHO says the same about its surgical checklist: "Adaptation after local consultation is encouraged" 11.
4. Put the list where the work happens, at the moment it's needed
Do this. Find the moment just before the mistake usually happens: before the van leaves, before the quote is sent, before the job is marked done. Put the list there, in a form your team already uses. That might be paper on the van door, a required field in your job software, or a template in your quoting tool. Don't add a new app.
It worked when nobody has to go and look for the list. It is in their hands at the moment they need it.
Why. A list in a shared drive gets read when someone has time, and mistakes happen when no one has time. Boorman's first rule is a clear "pause point" in the work where the list is used 810. One owner put it this way, after a key person left: "I've tried: - Telling people to write things in Notion (they don't) - Screen recording tutorials in Loom (nobody watches them) - Writing SOPs myself (takes forever and I'm the bottleneck)." Another owner, whose staff mostly have near-zero computer skills, found that "tablet-based SOPs or checklists" face "a giant hurdle." A list that only some of the team uses is not a system yet.
Example. The WHO surgical checklist has 19 items spread over three pauses: before anaesthesia, before the first cut, and before the patient leaves the operating room 11.
5. Give the list one owner, and give everyone the right to stop the job
Do this. Write one name at the top of the list. That person doesn't have to do the job; they answer for it being done right, and they keep the list current. Under the name, write one stop rule in plain words, for example: "If a line is missed, the job does not leave. Call Sam, not the owner." The first time someone stops a job under that rule, back them in front of the team, even if it costs you a day.
It worked when the first stop happens without you, and the problem it caught gets fixed without you.
Why. A list nobody owns goes stale. A list nobody may enforce becomes a suggestion. The stop rule is also what sends the customer's call to the list's owner instead of to you.
Example. At Toyota, every team member "is permitted to stop the production line" by pulling the andon cord. That alerts the team leader, and the line stops if the problem can't be fixed in time 12. At Hopkins, nurses ran the checklist, and leaders agreed to back them if doctors balked 3.
6. Turn every question into a line on the list, and count what still reaches you
Do this. Make one rule. Questions about the job go to the list first, then to the list's owner, and only then to you. If the list doesn't have the answer, the owner adds a line that day. Once a week, count the questions and redos about that job that still reached you.
It worked when that weekly count falls, and the list has a few more lines than it started with, each one written after a real question.
Why. A question is not proof that someone didn't read. It is a report that the list has a gap. If the answer goes back into the list, the same question is less likely to return. One service-business owner described the time before their procedures existed: "I spent most of my time answering the same questions all day." The weekly count is the number that tells you whether the system is used. A document that exists proves nothing.
Example. GitLab, which runs on a public handbook, tells staff that once a question is answered they should "discuss where it should be documented and who will do it" and ask, "Who will document this?" Its co-founder Sid Sijbrandij is frank about how hard that is: "The biggest problem is GitLab not working handbook first" 13.
Those are the six steps. Run them on one job, and you have a list your team uses, and a count that shows it.
How you bring a list in seems to matter as much as what is on it
The usual advice is to write every task down, train people on it, and hold them to it. Michael Gerber's The E-Myth Revisited (1995) taught owners to build the business like a franchise, so it runs the same whoever does the work 14. Gino Wickman's Traction asks for each core process to be "followed by all" 15. That is right about the writing. It says less about the getting. One reader of Traction wrote that "this book would not be sufficient for implementing it." For why reading them rarely changes the week, and what does, see You Read the Business Books. Why Didn't Anything Change?.
Medicine ran the test. In a study for the World Health Organization, eight hospitals in eight cities brought in a 19-item surgical checklist. Deaths fell from 1.5% to 0.8% after it came in, and complications from 11.0% to 7.0% 1. Then Ontario made every hospital report publicly on its checklist use. Across 101 hospitals, about 109,000 operations in three months before and 106,000 in three months after, deaths went from 0.71% to 0.65% and complications from 3.86% to 3.82%. Neither change was statistically significant 2. The same kind of list, with no clear effect.
Here is the idea the usual advice leaves out. A document does not create a habit. Taken together, these studies suggest a list gets used when the people using it help shape it, someone checks it at the moment it counts, and leaders back that check. Steps 3 to 5 build those three things.
The owner writes a full manual of the job, saves it in a shared drive and tells the team to follow it.
The team works from a short list of the steps that get skipped. It sits where the job happens, one person owns it, and it changes every time someone has to ask.
It runs as a loop, not a launch. The job runs with the list at hand. Something is missed or someone asks. The list's owner fixes the list that day, and the next run uses the better list. Fewer questions reach you each week. That is the move from level 2 to level 3: from an owner who checks everything to repeat work that runs from lists the team uses.
Johns Hopkins and Michigan show all six steps at work
The central-line checklist is the clearest worked example, because it was measured from the start.
At Johns Hopkins, Pronovost wrote five lines: wash hands, clean the skin with antiseptic, cover the patient in sterile drapes, wear sterile gear, put a sterile dressing on the site. Then the work was watched. "In one case out of three, the doctor skipped at least one of these essential steps" 3. Nurses ran the list at the bedside and were told to insist on every step. Hospital leaders agreed to back them if doctors balked. Reporting errors came to be rewarded 3. A year in, the infection rate had fallen from 11% to zero, and it held for another 15 months 3. Pronovost and his colleagues judged that "nearly 60 percent of their impact could be ascribed to a change in the 'safety culture'" 3.
Then the program went to Michigan: 103 intensive care units in 67 hospitals 34. A later study found it worked through, among other things, a close network of units, a culture of commitment, infection data used as pressure, and "hard edges" 5. Infections fell by up to 66% and the drop held through 18 months 4. The National Academy of Sciences account credits the program with an estimated 1,500 lives saved 3.
Read it against the steps. One failure first (step 1). Five lines, then watching what got skipped (step 2). Each hospital's own 5% (step 3). The list at the bedside, at the moment the line went in (step 4). Nurses who could insist, with leaders behind them (step 5). Errors reported and infection rates counted (step 6).
Lists fail when they arrive as an order, replace judgement, or cover for a missing person
This method has limits, and owners who have read the books raise them for good reason.
"I'll just make it mandatory." In England, researchers from Imperial College ran 119 interviews with operating room staff at 10 NHS hospitals 6. The checklist was often rushed or only partly done, and key people were often missing during the checks. Staff saw it as "just another piece of government-mandated paperwork" 3. An order from the owner gets you the paperwork. Steps 3 to 6 get you the habit.
"It turns people into machines." One reader of Traction said it treats an organisation as "a linear 'machine' that can be controlled."
I don't see myself hiring other people to serve as automatonsa reader of The E-Myth Revisited
That fear is right about long manuals. A short do-confirm list is the opposite. It leaves the skill to the person and only guards the steps that hurt when missed. If a list starts telling skilled people how to do what they already do well, cut it back.
"It won't work in a skilled or service business." One reader of The E-Myth Revisited doubted it transfers to "service businesses, and highly skilled technical businesses." Surgery and intensive care are highly skilled work, and that is where the studies above come from 14. For judgement-heavy jobs, use the last row of the table in step 2: one result the person answers for, plus the few lines that must never be missed.
"The person can't do the job." A list catches skipped steps. It does not teach a missing skill or fix a wrong hire. If the same person misses the same lines after the list is in place, that is a people decision.
"I don't have a team yet." One solo owner said it plainly: "I just don't feel like I can afford to pay someone to do that work yet." Lists still help, but the first question is what to hand off and to whom. That is covered in How to Replace Yourself in Your Business.
"I'll give the list to an AI agent instead." An agent can follow a list, but following a list is not the same as moving a result. That trap is the subject of Your AI Agents Are Busy. That Doesn't Mean They're Useful.
This week, build one list for one job and count what comes back
Don't systemise the whole business this week. Build one list for one job, with the person who does it, and measure it over ten runs.
One job, one list, ten runs
It's working if, by the tenth run, that job comes back to you less often than in the ten runs before. Then pick the next job.
You now have the steps, the right list for each job, and the number that shows it's used
You started with a process written down that nobody opens, and questions that still land on you. Here is what you have now.
- The six steps: start with the job that comes back to you, write only the skipped steps, write them with the doer, put the list at the pause point, give it one owner and a stop rule, and turn every question into a line.
- The test for which list fits: read-do for new people and rare jobs, do-confirm for people who know the job, answer templates for repeat questions, and an owned result for judgement work.
- The weekly number: the questions and redos about that job that still reach you. When it falls, the system is in use.
Read before: how to record yourself doing a job once and turn it into a written process, and which processes to write down first. If the work still runs through you alone, start with How to Replace Yourself in Your Business.
Read next: who keeps the list up to date when the work changes, how to get a key person's knowledge written down, and how to prove the business runs for 90 days without you.
Where you stand: this is item O-4 People on the Operator 2030 Scorecard. Level 2 reads "Good people, but I check their work and retrain them again." Level 3 is where every seat has a written outcome and one number its owner reports.
Sources
- A surgical safety checklist to reduce morbidity and mortality in a global population · Haynes, Gawande et al., New England Journal of Medicine, 2009 (abstract)
- Introduction of surgical safety checklists in Ontario, Canada · Urbach et al., New England Journal of Medicine, 2014 (abstract)
- The Hospital Checklist: How Social Science Insights Improve Health Care Outcomes · Robin Marantz Henig, From Research to Reward, National Academy of Sciences
- An intervention to decrease catheter-related bloodstream infections in the ICU · Pronovost et al., New England Journal of Medicine, 2006 (abstract)
- Explaining Michigan: developing an ex post theory of a quality improvement program · Dixon-Woods, Bosk, Aveling, Goeschel, Pronovost, Milbank Quarterly, 2011
- A Qualitative Evaluation of the Barriers and Facilitators Toward Implementation of the WHO Surgical Safety Checklist Across Hospitals in England · Russ, Sevdalis, Moorthy, Mayer et al., Annals of Surgery, January 2015
- Namesakes: Pete Hill · Robert S. Dudney, Air & Space Forces Magazine, 1 April 2020
- The Checklist Manifesto: How to Get Things Right · Atul Gawande, Metropolitan Books, 2009 (Daniel Boorman on checklist design)
- Book review: The checklist manifesto · Scott Berkun, 18 November 2010
- Boeing Checklist: How It Saves Lives, What It Can Teach You · Shortform summary of Atul Gawande, The Checklist Manifesto
- Safe Surgery: tool and resources · World Health Organization
- Andon: Toyota Production System guide · Toyota UK, 2016, updated 2026
- Handbook Usage · The GitLab Handbook
- The E-Myth Revisited: Why Most Small Businesses Don't Work and What to Do About It · Michael E. Gerber, 1995 (revised edition of The E-Myth, 1985)
- Traction: Get a Grip on Your Business · Gino Wickman, BenBella Books, 2011
