Notes · Operating

The bottleneck you're too good at handling

It isn't the thing that broke last month. It's the job you've never missed, and your perfect record is why it's still there.

PAUL BELLINGER · NOVEMBER 2026 · 17 MIN READ · SOURCES AT THE END

There's a thing you do every week that isn't on any to-do list. Maybe it's the Friday half hour moving numbers from one screen into another. Maybe it's the string of texts it takes to get four people to the same place at the same time. Maybe it's the question only you can answer, so everyone waits on you.

It isn't hard, and you're good at it, and that's the problem. Because you're good at it, no one ever sees it fail. So it never gets written down as something to fix. That's a bottleneck, the kind that's hardest to see.

The best research I've found on why this happens is about hospital nurses. I've also lived it, in a weekly job I never once missed and came to dread. This research note explains why a bottleneck like this lasts and how I got out of mine. Then it gives you a twenty-minute way to find yours.

The claim

The better you get at handling a small problem, the longer it stays broken. That's how a bottleneck survives: your skill hides it, rewards you for keeping it, and uses up the time a fix would take.


01 · The findingYour workaround is what keeps it broken

A hospital seems a long way from your business, but this note is built on a study that followed nurses. The problems they ran into lasted because of what made them good at their jobs.

Anita Tucker and Amy Edmondson, then at Harvard Business School, followed 26 nurses at nine hospitals for 239 hours and logged every time something got in the way of the work. They counted 194 failures.

The linen run

One nurse on a night shift ran out of clean linen for her patients' beds. So she walked to another unit and took some, and brought back extra for the other nurses. It took a few minutes and cost almost nothing. Nurses at other hospitals did a pricier version of the same thing: they often paid for a taxi to bring supplies from the off-site laundry.

By any ordinary standard, she'd handled it. That's a workaround: it gets you through the shift and leaves the cause where it was. Now look at what didn't happen. Nobody who ran the linen supply heard it had come up short, so nothing about the supply changed, and the unit she'd taken from was now short itself. The next shortage would get the same walk.

That was the pattern for 93% of the problems the researchers watched. The nurse did what it took to keep going, no more, no less, and told no one. When she needed a hand to keep going, she asked a coworker, not someone who could fix the cause. Only 7% of the nurses' responses got a problem in front of someone who could. And the bar for that was low: any such call counted, whether or not anything changed. Averaged out, the workarounds cost about 33 minutes of every eight-hour shift. The authors priced that, on their own assumptions, at about $256,000 a year in nursing time for a 200-bed hospital.

Out of every 100 problems

What nurses did with the problems they ran into.

Tap a dot for the kinds of problem the study saw.

worked around, told no onethe other 7

FIG 1 · Out of every 100 problems, nurses worked around 93 and told no one (Tucker and Edmondson, 2003, p. 61).
How this figure was made

The dots show a share, not 100 problems counted one by one. The paper gives a percentage: nurses did what it took to keep going, no more and no less, and told no one, for 93% of the problems (p. 61). It logged 194 failures, 166 of them problems (p. 58), and lists no count for the 93%, so none is drawn. The other 7 dots are what is left of 100. The paper doesn't say what nurses did with those problems. They are not the 7% in the text, which counts something else: the share of nurses' responses that called attention to a problem in a way that could start a look into its cause (p. 62). A tap shows examples of the kinds of problem in the study, each with its page: linen (p. 60), a late medication (p. 68), the wrong meal and missing orders (p. 62), missing or broken equipment (p. 58). The paper gives the linen run as its example of this rule and doesn't sort the others into either group. Tucker and Edmondson, California Management Review 45(2), 2003. Drawn by the page from the JSON build_hundred_problems.py wrote.

Good nurses, easy problems

You might guess the problems were hard or the nurses careless. Neither holds. The authors describe the problems as straightforward and routine, like missing medications, supplies running short, or the wrong meal delivered to a diabetic patient.

And the nurses weren't careless. All but one of the hospitals were picked for their reputations for nursing excellence, and every nurse had more than three years on the unit. Of the ten nurses followed for a full shift, nine stayed about 45 minutes afterward, unpaid, to finish their patients' care. One woke up at 4 AM and called the unit about something she'd forgotten to pass along.

These were good people at good hospitals with easy problems, and the problems stayed. The authors blame three things that sound like virtues: personal responsibility for whatever comes up, units staffed with no slack, and teams left to run without a manager nearby. Each one keeps problems quiet.

Why it feels right

Seven of the nine nurses they asked said working around a problem felt good. The one who made the linen run put it this way:

Working around problems is just part of my job. A nurse, quoted by Tucker and Edmondson · 2003

That feeling keeps the problem alive. Tucker and Edmondson drew this as a loop, in a model built from what they saw and heard: handling a problem well leads to feeling competent, and feeling competent leads to less effort spent getting anyone else involved.

Why the problem comes back

Part of Tucker and Edmondson's model in plain words, with their own label under each step.

Tap a step for what the paper says about it.

leads to, in their modelthe loop back to the first step

FIG 2 · Feeling competent at handling a problem cuts the effort to get anyone else involved, so the cause stays and the problem comes back (after Tucker and Edmondson, 2003, p. 65).
How this figure was made

This is part of Figure 1 in the paper, the authors' model of how nurses respond to problems (p. 65), redrawn in plain words with their label under each step. It draws the part the note describes, the loop through feeling competent. Two steps stand for two of theirs: "You handle it well" is both first-order problem solving effort and its effectiveness, and between "Less effort" and "The cause is never fixed" their model has the effectiveness of second-order problem solving. Left out: a short loop in which the patch removes the problem for now (p. 64); a delayed link from the patching to burnout, which cuts the effort further (p. 66); and six outside pushes, among them management support and psychological safety, which raise second-order effort (pp. 65 to 67). The model shows which way each link pushes, not how hard. Tucker and Edmondson, California Management Review 45(2), 2003. Drawn by the page from the JSON build_chain.py wrote.

Managers push the same way. Most of the twelve nurses the researchers interviewed believed their manager expected them to work through daily disruptions alone, and speaking up was likely to read as incompetence. One nurse explained, "My manager is not interested in hearing about things if they are small. If I went to her with a small problem, she would say, 'Solve it yourself.'"

Most managers, the authors write, "would identify an ideal employee as one who can handle with ease any problem that comes along, without bothering managers or others." If you want the organization to learn, they argue, the ideal employee is instead a "noisy complainer," who fixes the immediate thing and then tells the manager and the supply department that the system failed.

The researchers watched one noisy complainer at work. A new ICU nurse had brought a sleeping two-year-old to an oncology floor in an ICU bed and left the bed with him. The ICU wanted it back, and the oncology nurse returned it. But she also did something to fix the cause: she called the ICU's nurse manager so it wouldn't happen again. She opened with "I don't want to get anyone in trouble." One phone call, and she apologized for making it.

That's the trap. A system that's quietly broken looks, from the outside, like a person who's good at their job. The nurse who never complains about linen is the reason nobody orders more linen. And the workaround isn't free for her. To the nurses living it, the authors write, this all seems "stressful, but basically in balance." They call the balance "illusory," because the cost shows up later as burnout.

A system that's quietly broken looks, from the outside, like a person who's good at their job.


02 · My own caseThe job I never missed

Tucker and Edmondson suspected the pattern went well past hospitals, and they named help desks, repair technicians and airline crews as work that looks similar. I can add a subscription community with a weekly deadline, because I was the nurse in it.

Every Saturday night I spent at least an hour, often two, getting our contributors' content together, organizing it, structuring it, and publishing it to our community platform. It took intense focus the whole time. I dreaded it, and it hung over me all week.

I also did it well. I never missed one. Every Sunday morning our members had their new content, like clockwork. I got good at it, and I became the only person in the company who could be trusted to do it correctly.

That's the linen run with a calendar on it. From where a member sat, nothing was wrong, because I made sure nothing was wrong. A job that never fails never comes up as a problem. And the problem got stickier, because the only person who could fix the job was the person busy doing the workaround.

I got out by building a set of tools to handle it for me. Now I just kick off the tools and check on them when they're done. A job that took four to eight hours a month takes about 15 minutes or less, and all of the stress went with it.


03 · What holds it in placeWhy nobody comes to fix it

Skill hides the problem and rewards you for keeping it. Two more things hold it in place. It never hurts enough to force a fix, and when the week is full, working harder always looks cheaper than working smarter.

It never hurts enough

Even when you ask owners directly, the time these jobs take doesn't rank high. When about 2,900 members of the NFIB, the small-business trade group, rated 75 problems in 2024, "controlling my own time" came 38th. About one in ten called it critical. The cost of health insurance came first. I don't think that low rank means owners' time is fine. I think it goes in pieces too small to rate as a problem.

75 problems, ranked

One dot for each of the 75 problems owners rated, in their order, most serious first, reading across.

Tap a dot for its problem and the share of owners who called it critical.

38th: controlling my own time1st: the cost of health insurancethe other 73, in rank order

FIG 3 · Small-business owners ranked controlling their own time 38th of 75 problems in 2024; the cost of health insurance came first (NFIB, 2024, pp. 9 to 10).
How this figure was made

NFIB mailed its questionnaire to 40,000 members and 2,873 owners returned usable ones (p. 1). Each rated 75 problems on a scale of 1 to 7 running from a critical problem to not a problem (p. 1). NFIB ranks the problems by their mean rating, and the dots follow its order, left to right and row by row. Controlling my own time is 38th, with a mean of 4.27, and 10.4% called it critical (Table 1, p. 10). The cost of health insurance is 1st, with a mean of 2.58 and 40.5% calling it critical (Table 1, p. 9). The share who said critical is a separate column and doesn't set the order, so some problems below 38th have a higher share. This shows what owners rated, not what time costs them. NFIB's members are not every small business: the report says they lean toward interior states and include farms (p. 2). Holly Wade and Madeleine Oldstone, NFIB Research Center, "2024 Small Business Problems & Priorities", eleventh edition. Drawn by the page from the JSON build_nfib.py wrote.

Daniel Gilbert and three colleagues argue that small pains can outlast big ones, in a paper titled The Peculiar Longevity of Things Not So Bad. Dealing with pain takes effort, so we only switch that effort on once the pain crosses a line. Their experiments tested hurt feelings: people stayed angry longer at someone who hurt them a little than at someone who hurt them a lot. They make the same point about the body. You can put up with a sore knee for years, but tear your ACL and you're in front of a surgeon that week.

People rationalize divorces, demotions, and diseases, but not slow elevators and uninspired burgundies. Gilbert, Lieberman, Morewedge and Wilson · 2004

In their lab studies, students who were stung harder by an insult ended up thinking better of the person who wrote it than students who were stung less. The harder sting crossed that line, so they did the work of getting over it. Those were small studies, measured over minutes, so I'm borrowing the idea and not the numbers.

Getting good at a job keeps it under that line. Every time you get better at it, the job hurts a little less and sinks a little further down. Nobody stops to deal with it, not even you. The big breakdown got fixed, because it hurt. The thing that's been slightly annoying for three years is still there.

Work harder or work smarter

Nelson Repenning and John Sterman at MIT spent a decade studying why companies skip improvement work that would pay them back for years. A manager who's behind, they found, has two basic choices: get people to work harder, or get them to work smarter. Working harder means more hours and more effort on the process you already have. Working smarter means spending time on the process itself, so every hour after it gets more done. "Of course, everyone knows that it is better to work smarter than to work harder," the authors write. In most of the companies they studied, harder won anyway.

The reason is timing. Working harder is the workaround: it helps right now and hurts later, when the neglected process starts to wear out. In their words, it's "better-before-worse." Working smarter runs the other way, "worse-before-better," because the time comes out of this week and the payoff comes later. So under pressure, people pick the option that looks free.

Better before worse

Performance over time after each choice, from the same start. A sketch of a shape, so no numbers.

Tap a line, or the shaded stretch, for what Repenning and Sterman found.

working harder: better, then worseworking smarter: worse, then betterwhere performance stood before the choice

FIG 4 · Working harder helps now and hurts later; working smarter costs now and pays later, so at the moment of choice working harder looks free (after Repenning and Sterman, 2001, p. 74).
How this figure was made

A sketch of the idea with no numbers on either axis, because the paper prints none. The two lines are traced from Figure 6 in Repenning and Sterman (p. 74), two runs of their simulation of a made-up process that start from the same level: in one, people work harder, putting more effort into the process they have; in the other, they work smarter, spending time on improving the process (p. 73). The paper draws them side by side; here they share one chart, each keeping the paper's proportions. The dotted line is where performance stood before the choice. Now is where the first line leaves it, and the shaded stretch runs until the two lines cross. The paper calls the first shape better before worse and the second worse before better (p. 73), and says people who don't see the trade-off pick working harder (p. 75). Repenning and Sterman, California Management Review 43(4), 2001. Drawn by the page from the JSON build_bbw.py wrote.

And it feeds itself. Every week spent working harder skips a little upkeep. The process slips, and next week takes even more effort. They call it the capability trap. An auto plant supervisor told them about team leaders who couldn't stop a line making 20% garbage, because they had to "run like crazy" to hit their counts. "They could never get ahead of the game."

Their best case of working smarter is British Petroleum's refinery in Lima, Ohio, which was burning off butane as waste. By the paper's account, a team of its own people stopped it, with two weeks of work and $5,000, saving about $1.5 million a year. Most of the equipment and materials were already on site. Members of the team had known the problem, and the fix, for eight years. At that rate, about $12 million burned off while a $5,000 fix waited.

What stopped them, the authors argue, was the belief that there was never time or money for improvement. A refinery isn't a small business, but I'd bet every business has a fix like this at some scale: two weeks of work that's been waiting for years.

And when someone does make one, nobody rewards them for it. The paper's title quotes an engineer at an auto company: "Nobody ever gets credit for fixing problems that never happened." In companies stuck this way, the authors argue from their interviews, the people who get promoted are the ones who come to the rescue when something big breaks, not the noisy complainers who get small problems fixed before anything breaks.


04 · How to spot yoursYou have to go looking

If you want to find your bottleneck, you can't wait for it to hurt, and you can't count on anyone telling you about it. Ask an owner what eats their time and you'll hear email, admin, scheduling. True, and useless. Ask about last week instead, because it actually happened.

Then count, don't guess. Nobody can tell you how long your weekly jobs take, and your own guess can be off in either direction. People tend to remember a long task as shorter than it was. Ask them about weekly chores, though, and they guess far more hours than they write down in a time diary: in one study, men guessed 23 hours a week of housework and logged 10. Managers misjudge their whole operation, too. Researchers in the World Management Survey asked managers of mid-sized manufacturers to rate how well their own firm was run, not counting themselves, out of 10, with 5 as average. The managers gave their firms 7.1, well above average, and their ratings bore no relation to the researchers' scores or to how the firms actually performed.

The signs

  • You've never missed it. Nobody has ever had to chase you for it, which is why nobody thinks about it.
  • You do it every week, the same way. If you could write the steps on a card, it's a candidate.
  • It starts with "I'll just." I'll just text everyone. I'll just copy it over. Those are the words of a workaround.
  • You're the only one trusted with it. It lives in your head or a text thread, so everything behind it waits for you.
  • It hangs over you before you start. The job takes an hour. The dread takes the week.
  • You have a workaround you're a little proud of. A color-coded spreadsheet, a note pinned on your phone. Remember the nurses: feeling good about a workaround is how the problem stays.
Twenty minutes, with last week's calendar and your phone
  1. Scroll back through last week: your calendar, texts, sent email and call log. Write down the tasks that come up every week. Stop at ten.
  2. Next to each one, write how many times and roughly how long. For the two biggest, time them once this week instead of guessing.
  3. Ask yourself, if you were out sick for two weeks, what falls apart first?
  4. Then, if you could hire one person tomorrow, what would they do all day?
  5. Circle the task on your list that comes up in your answers to those last two questions. If more than one does, pick the one that hangs over you before you start. That's usually your bottleneck.

The hire question does the most work, because it turns a vague wish for help into a job description. If you'd like a second pair of eyes on your list, email it to me at [email protected] and I'll go through it with you.

If it's your best person's job

If you run the place, the linen run you most need to find may not be yours. It may belong to the person who never bothers you. That's exactly who won't tell you, for the same reasons the nurses didn't: handling it feels good, and speaking up can look like incompetence.

Tucker and Edmondson saw that managers who were physically on the floor heard about more problems. The researchers themselves found the workarounds only by standing next to nurses for 239 hours. A bottleneck like this won't come to your desk. Go stand next to your own people for a morning, and ask what they handle that you never hear about.

Then do something with what you hear. One nurse told the researchers, "I know nurses on our floor used to come up with suggestions for change. No one seems to listen and now no one bothers trying." So when someone brings you a small problem, don't tell them to solve it themselves.


05 · What to do nextProve the fix small, on your own floor

Once you've found your bottleneck, your first reaction will probably be that the job isn't worth fixing. That's a belief, and a randomized trial in Indian textile mills, published in 2013 by Nicholas Bloom and colleagues, shows how wrong that belief can be and what changes it. The researchers gave some plants free help adopting basic practices, like recording defects and keeping stock in order, and those plants raised productivity 17% in a year. The changes themselves cost less than $3,000 per firm, though the consulting that found them would have run about $250,000.

Why didn't the owners make these cheap changes on their own? Money wasn't the problem. Some owners had simply never heard of a practice, and took it up once it was explained. The harder cases had heard of it and decided it wasn't worth doing. They were profitable, and they figured their quality was as good as the next firm's, and that belief was slow to move. What changed their minds was trying it on a few machines first, then rolling it out once it worked.

The authors describe their evidence on this as "speculative," and those were big firms, about 270 people each, with consultants on the floor for months. Trials with much smaller firms are less encouraging. In Ghana, tailors who got advice from an international consulting firm changed how they ran their shops, then went back to their old ways. That means a small test on your own floor won't settle it for good. It's still the cheapest way to find out you were wrong about a job being fine.

Belief wasn't the last thing in the way. Once owners were convinced, the changes piled up waiting for them. They were already working about 68 hours a week, and over nine months, the share of unadopted practices waiting on the owner's time grew from 3.7% to 14%. It's my Saturday night again: the person who'd have to make the fix is too busy to make it.

Take something away

Size the fix to the pain. A job this small doesn't need a big project. The first question is: what can you remove? When people finally do fix something, they tend to add a new app, a new step, a new form. A 2021 study in Nature found people overlook changes that take something away, and more so when their minds are on something else.

My smallest example is Chalk, a phone app I built for capturing my thoughts and passing them on to the AI assistants I work with, without having to type a word. I have ideas while I'm driving or walking the dog, and most of them never reached a keyboard. My workaround was voice memos, which meant recording, remembering, moving the file to my Mac and transcribing it. The process was fiddly and often broke. I once lost hours trying to get a long voice memo into my transcription tool, because AirDrop would send only its first 30 seconds. But it mostly worked, which is exactly why I kept doing it. That was my linen run, a way of coping that fixed nothing.

I took the steps out. Chalk is one red button on my phone. Tap, talk, tap. It saves every second while I talk and sends the audio to my Mac, where it's transcribed automatically into text my assistants can read. Nothing to type, no files to move.

A couple of my other small tools came from the same place. Drillbag, which I use all the time, took over the texting it takes to schedule pickleball practice with my friends. Moneybag keeps score and runs the website for a weekly pickleball game I organize. I build these apps with AI, which has made the building cheap for anyone. Seeing the job in the first place is still the hard part.

Expect it to get worse first

While you're fixing the job, you're still doing it. Back at the Lima refinery, maintenance costs went up 30% in the first six months of the improvement program, and the managers stuck with the program because a training game had shown them the jump was coming. A small fix costs far less, but it runs the same worse-before-better course, so don't judge it by its first two weeks.

Here's how you'll know the fix worked. Some jobs you'll forget you ever did. Others you'll be grateful for every week you don't have to do them. I am, every Saturday night, when the evening is mine again.


06 · The catchNobody will notice when it's gone

A job like this will never force your hand, because you won't let it fail. And fixing it won't earn you anything. When I stopped spending my Saturdays on the community's content, no member had any reason to notice. The new content still showed up every Sunday morning, like always. Remember the engineer Repenning and Sterman quoted in their title: "Nobody ever gets credit for fixing problems that never happened."

No customer will complain about this job, no employee will flag it, and no number will turn red. The push has to come from you, on a quiet week when nothing is wrong.

The nurses were good at their jobs, and their problems stayed. I was good at my Saturdays, and that job stayed mine until I built my way out of it. The better you are at a job, the less anyone will ever ask you to fix it. Your bottleneck is hiding behind your proudest workaround.

SourcesWhere all of this comes from

Researched in October 2026 in three parallel passes using AI research assistants, then a second, separate round that checked each claim against its source. That round confirmed most of it and corrected the rest, among them a cost figure, a date, and a finding I'd stated too broadly. For this version in November 2026, an AI assistant went back through the papers again. Then, on November 10, a third round went looking for evidence against the note, and I corrected six passages and added one study. I decided what stayed. The claim in the box at the top of this note is my reading of this research, not a measured finding. Two separate searches by AI research assistants found no study that times people as they get better at a workaround and tracks whether the problem gets fixed. The closest test is a 2016 lab experiment by Anita Tucker: nurses who had to walk farther to work around two missing supplies were more likely to suggest a fix. The community story is my own account, and its numbers are my estimates.

  1. NFIB Research Center, Small Business Problems and Priorities, 2024 edition, a mail survey of NFIB members. nfib.com (PDF)
  2. Anita L. Tucker and Amy C. Edmondson, "Why hospitals don't learn from failures: Organizational and psychological dynamics that inhibit system change," California Management Review 45(2), 2003. utsouthwestern.edu (PDF)
  3. Daniel T. Gilbert, Matthew D. Lieberman, Carey K. Morewedge and Timothy D. Wilson, "The peculiar longevity of things not so bad," Psychological Science 15(1), 2004. dtg.sites.fas.harvard.edu (PDF)
  4. Nelson P. Repenning and John D. Sterman, "Nobody ever gets credit for fixing problems that never happened," California Management Review 43(4), 2001. web.mit.edu (PDF)
  5. John P. Robinson, Steven Martin, Ignace Glorieux and Joeri Minnen, "The overestimated workweek revisited," Monthly Labor Review, June 2011. bls.gov (PDF)
  6. Michael M. Roy, Nicholas J. S. Christenfeld and Craig R. M. McKenzie, "Underestimating the duration of future events: Memory incorrectly used or memory bias?," Psychological Bulletin 131(5), 2005. pubmed.ncbi.nlm.nih.gov
  7. Nicholas Bloom and John Van Reenen, "Why do management practices differ across firms and countries?," Journal of Economic Perspectives 24(1), 2010. researchonline.lse.ac.uk (PDF)
  8. Gabrielle S. Adams, Benjamin A. Converse, Andrew H. Hales and Leidy E. Klotz, "People systematically overlook subtractive changes," Nature 592, 2021. nature.com
  9. Nicholas Bloom, Benn Eifert, Aprajit Mahajan, David McKenzie and John Roberts, "Does management matter? Evidence from India," Quarterly Journal of Economics 128(1), 2013. povertyactionlab.org (PDF)
  10. Dean Karlan, Ryan Knight and Christopher Udry, "Consulting and capital experiments with microenterprise tailors in Ghana," Journal of Economic Behavior & Organization 118, 2015. doi.org
  11. Anita L. Tucker, "The impact of workaround difficulty on frontline employees' response to operational failures: A laboratory experiment on medication administration," Management Science 62(4), 2016. doi.org
  12. My own tools: pbhq.app/chalk, pbhq.app/drillbag and pbhq.app/moneybag.

Related notes: "Nobody is lying to you," on what you can and can't learn by asking people, and Is The Lean Startup still right, fifteen years on?, on why the proven benefit of testing is quitting sooner.