ManageTime

Article

Still doing it every week: the recurring work that stopped earning its place

To stop low-value recurring work, write down what you are stopping, tell whoever receives it, and set a date to come back and check what broke. Deciding is the easy part. Standing work survives because nobody ever re-decides it, so a stop that lives only in your head tends to undo itself. Most of the research on stopping work comes from healthcare, where it has its own name, and it points at one pairing: a commitment somebody else can see, plus a scheduled look back. This page covers how to judge a standing item, how to stop it, and what that evidence does and does not transfer.

How do you stop doing a recurring task at work?

Friday, late afternoon, and the weekly status document goes out the way it has gone out for two years. Nobody replies to it. Nobody replied last week either. Somewhere around the fourth month you stopped reading it back before sending, and the hour it costs has quietly become part of what Friday is.

Standing work survives for a dull reason: it only ever got decided once. A new request gets weighed, argued over, and either accepted or turned down. A recurring item was weighed on the day it started, under conditions that have since moved, and then it stopped being a decision and became furniture. The project it reported on finished. The manager who asked for it changed teams. The two people who read it now get the same numbers from a dashboard. None of that generates an email telling you to stop.

So the first move is to make the list exist. Twenty minutes with your calendar and your sent folder, going back a month, writing down every item that repeats: the reports, the standing checks, the updates, the reviews, the thing you do every Monday because you do it every Monday. Most people find the list runs longer than they expected, and that length is the finding rather than a side note. Nothing on it has been re-decided since the day it started.

A recurring task is a decision you made once and have been paying for ever since.

Then stop one of them properly, which means three things rather than one. Write down what you are stopping and from when. Tell the people who receive the output, before the first week it fails to appear. And put a date in the calendar, far enough out that a real problem would have surfaced by then, to ask what actually broke. Together those take about ten minutes, and they are the whole difference between a stop that holds and a drift back you will not notice until you are doing it again.

One case sits outside this page. If the recurring item is a meeting, the invite is the real object and the fix runs through other people, which the meetings pillar covers in its own right.

How do you tell whether recurring work is still worth doing?

Mostly you cannot tell from inside the task. The output leaves your hands and nothing reports back, so the absence of complaints reads as evidence that it is working.

Two questions get you most of the way. Who uses this, by name, and what do they do with it? And what happens if it arrives late, or never arrives? The first question is the one that fails, because the answer turns out to be a role rather than a person, or a person who left in March. The second thins the list again: a fair amount of standing work has no consequence attached to it anywhere, which is exactly why nobody ever reviewed it.

The Pareto Principle is the method on this site that comes closest to this job, since it asks which small share of your work produces most of the result you care about. Read it backwards for standing items. If a recurring task is not in the share that produces the result, it is in the share that is only producing itself.

The Eisenhower Matrix has a quadrant for this too, the one that is neither urgent nor important, where the instruction is to eliminate. In practice that quadrant is where recurring work hides, because a weekly item always feels mildly urgent on the day it is due. Sort by the week instead of by the day and the same task lands somewhere else entirely.

Stopping quietly to see who complains is the tempting version, and it is worse than it sounds. It works, in that you do find out. It also costs you the one thing you need if the answer comes back badly, which is a record that you made a decision rather than let something slide. A pause you announced is a judgment call. The same pause nobody heard about is an oversight with your name on it.

Say it out loud and the risk changes hands.

Why do you keep doing it after deciding to stop?

Deciding is not the part that fails. Six weeks on, the report is going out again, and you would struggle to say which week it restarted.

Most of what is known about getting anybody to stop doing something comes from healthcare, where persuading a clinician to stop ordering a test or prescribing a drug that no longer helps is a research problem with its own name: de-implementation. That setting is not yours and the stakes are not comparable, so what follows is evidence about the shape of a stop, not about your Friday report. The numbers belong to the trials. Nothing here says what a recurring task on your own list would do.

Start with the base rate. Heus and colleagues went through 121 randomised de-implementation studies published between 1990 and 2019 and reported it in the Journal for Healthcare Quality in 2023. Of the 109 that compared a stopping attempt against carrying on as usual, 75 produced a significant reduction in the targeted work, close to seven in ten. Across the 73 trials pooled for effect size, the median reduction was 17 percent, and the middle half fell between 7 and 42 percent. Read the median twice, because it sets what you should expect. Most attempts worked, and most of them trimmed the work rather than ending it.

What makes a stop hold is the more useful question, and the largest pooling so far takes a position on it. Raudasoja and colleagues screened just over thirteen thousand abstracts, included 140 randomised trials, and published in BMJ Medicine in 2025. One approach reached moderate certainty, and it was the only one that did: telling people why the work should stop, paired with somebody auditing what they actually did afterwards and feeding the numbers back. Its odds ratio was 0.73, with a confidence interval from 0.63 to 0.84.

Each half on its own fell short, which is the part that should change what you do. Education alone came out at 0.86, interval 0.72 to 1.03, which crosses one. Audit and feedback alone came out at 0.82, interval 0.67 to 1.00, which touches it. In practice that maps onto two familiar failures: you know perfectly well why the report should stop and never check whether it did, or you keep checking without a written reason and cannot remember what you were testing.

The reason needs the re-check.

A trial from the year before shows what that looks like when the commitment is the lever. Kullgren and colleagues ran a stepped-wedge cluster randomised trial across 8 primary care clinics, published in JAMA Internal Medicine in 2024, covering 81 clinicians and 8,030 older patients. Clinicians were asked to commit in writing to three specific things they would stop doing.

Those who committed had their photograph put on a poster in the clinic and received a weekly email listing the alternatives. The targeted services showed up in 20.5 percent of control patient-months against 16.0 percent of intervention patient-months, an adjusted odds ratio of 0.79 with an interval from 0.65 to 0.97. Note where the commitment lived: not in the head of the person who made it.

Two limits on that one, and both matter here. The effect appeared in the pooled result only; taken one condition at a time, no single difference reached significance. And the whole package ran at once, so nothing in the design separates the written commitment from the poster or from the weekly email. What survives is narrower than the headline: the commitment was written, somebody else could see it, and something arrived every week to put the decision back in front of the person who made it. Copy that structure if you like. Do not copy the number.

The two reviews also disagree, which is worth knowing before you build anything on either. Heus found that effectiveness did not depend on how many strategies were used or which ones. Raudasoja found exactly one combination that cleared moderate certainty and rated the rest low. Different trial sets and different methods produce that gap and it is not settled. The thin overlap is the part that travels: a reason, and something that comes back to check.

For one person and one recurring task, that is unglamorous. You write down what you are stopping and why, somewhere another person can read it. You tell whoever was receiving it. And you book the look-back before you need it, because the version of you in six weeks has no memory of this decision and a calendar full of other things.

How does your working style change the answer?

Where this goes wrong depends on how you already work. The four names below are this site's own, from the working-style self-check on the home page.

Architects carry the most standing work and have the least trouble stopping any one piece of it, because their recurring items were set up deliberately and are written down somewhere findable. The exposure runs the other way. A system you designed well three years ago keeps running perfectly on a purpose that expired. Sprinters stop things easily and restart them without noticing, since a burst of work picks up whatever sat in front of it last time. If that is you, the written note is doing almost all of this and your memory is doing none of it.

Visionaries will cut a recurring task the moment it stops being interesting, which is usually earlier than it stops being useful, so if that is your pattern, hold onto the second question about who is actually waiting. Improvisers often have no list at all, which makes your twenty minutes of inventory the exercise itself rather than the preparation for it. Once the repeating items sit visible on one page, an Improviser will frequently stop three of them in an afternoon.

No style makes the look-back optional.

Some recurring work this page does not reach. Anything a regulator, an auditor or a contract requires is not yours to stop however little anybody reads it, and your honest move there is to make it cheaper rather than to end it. Work a senior person asked for personally needs a conversation rather than a note, because stopping it by email reads as a decision you did not have. And a standing item that three different people each depend on in a different way is a prioritization problem with other people in it, which gets settled out loud.

So open the calendar, go back four weeks, and write the repeating items down in one place. Against each one put a name, the person who uses what it produces, and leave the line blank where you cannot think of one. The blanks are your candidates. Pick the cheapest blank to stop, send the one message that says you are stopping it, and put the look-back date in before you close the calendar. That last step is the one the trials keep pointing at, and it is the one nobody does.

Frequently asked questions

How do you stop doing a recurring task at work?

Write down what you are stopping and from when, tell the people who receive the output before the first week it is missing, and book a date to come back and check what broke. All three take about ten minutes together. The written note is what stops the task reappearing by habit, and the booked date is what catches the case where somebody did quietly need it. Skipping the message is the expensive shortcut, because an announced pause is a judgment call and a silent one is an oversight.

How do you tell whether recurring work is still worth doing?

Ask who uses the output by name and what they do with it, then ask what actually happens if it arrives late or not at all. The first question fails more often than people expect, because the answer turns out to be a job title or somebody who left. The second thins the list again, since a fair amount of standing work has no consequence attached to it anywhere. Where both answers come back empty, you are looking at a candidate.

Who do you tell when you stop producing something?

Everyone who receives the output, plus whoever originally asked for it, and in writing rather than in passing. Keep it to three lines: what you are stopping, from when, and how to reach you if they needed it. You are not asking permission and you are not apologising, so neither belongs in the message. The reply is useful either way, because the person who speaks up is the reader you did not know you had.

Should you pause recurring work or end it outright?

Pause it, with a date to decide properly, because a pause is easier to announce and easier to reverse than an ending. It also matches what the evidence points at. In the de-implementation trials, the approach that reached moderate certainty paired a reason with an audit afterwards, and a pause with a review date is the small version of that pairing. If nothing broke by the review date, the pause becomes the decision without you having to defend it twice.

Sources