ManageTime

Article

Still polishing: when careful work turns into a stall

Careful work turns into a stall at the point where another pass stops changing what the reader gets. The research points at doubt rather than standards: across 67 studies, high personal standards went with better academic performance, while doubts about your own actions went the other way. Lower the doubt, not the standard. Those studies measured students, so read the direction as the useful part and the size with care. This page covers the stopping test, what the treatment trials found when they went after perfectionism directly, and how to close a revision loop that is already running.

How do you know when work is good enough to send?

Version four of the same deck, and the edits are getting smaller. You moved a heading up two lines. You swapped one word for a near-synonym, read it again, and swapped it back. The deck was usable at version two. The meeting is tomorrow, and the last three passes have now cost you more than the first draft did.

What ends this is a question about the person on the other end, not a feeling about the work. Does the next pass change a decision somebody makes, or does it only change how the work looks to you? Those two come apart fast. A number that is wrong changes a decision. A sentence that is fine but could be tighter does not, and that sentence is where your afternoon goes.

Most passes fail that question.

The objection to a test like that is fair, and worth naming before the test gets used as a licence to send sloppy work. Some work does improve on the fourth pass, and some readers notice the difference between fine and good. Craft is not the problem here. The problem is that the fourth pass feels identical to the second one from the inside, and the only way to tell them apart is from the outside, by what the work can now do that it could not do before.

Timeboxing carries the front half of that test already, in its instruction to define what done looks like before the clock starts. One page with the three numbers and a recommendation. Four slides a manager can present without you in the room. Written in advance, the definition is cheap and slightly boring, which is the point of writing it then. Written at version four, when you are tired and close to the work, it bends to whatever you have just produced.

When the test will not resolve, hand the decision over. Send the current version with one line naming what you were unsure about, and ask whether it matters. The answer costs a colleague half a minute and it ends an argument you could otherwise run alone until Friday. It also produces the one thing the loop cannot produce by itself: an outside view of what the work is for.

Is perfectionism why your work runs late?

Two things hide under that one word, and they pull in opposite directions. Researchers separate them. One half is what you are aiming at: high standards, personal standards, the version of the work you would be satisfied with. The other half is what you do about the gap: doubts about whether the work you just did was any good, and the distance you feel between the standard and what you actually produced.

In 2020 Osenk, Williamson and Wade published a meta-analytic review in Psychological Assessment that put numbers on that split. They pooled 67 studies and 378 effect sizes across the four perfectionism scales in common use, and set each subscale against academic outcomes: performance, procrastination, engagement, self-efficacy, test anxiety and several more. The subscales about standards came out on the helpful side. High Standards, Personal Standards and Self-Oriented Perfectionism were all positively related to academic performance and to the helpful outcomes, and High Standards, the one focused on striving for excellence, was the only subscale that related negatively to the unhelpful ones.

The doubt side went the other way, and it went there consistently. Two of the four subscales measuring perfectionistic concerns, Discrepancy and Doubts about Actions, were negatively related to academic performance, and every concerns subscale was positively associated with the unhelpful outcomes, procrastination among them. Same trait word. Opposite sign.

Read the sample before you carry that around the office. These were students, average age about 19, and the outcomes were academic ones: grades, test anxiety, coursework put off. Your quarterly report is not an exam, and a colleague waiting on the file applies a pressure that a coursework deadline does not. What transfers is the direction rather than the size. The half of perfectionism that costs you time is the doubt about whether the work is good, not the standard you set for it.

That changes what there is to fix. Advice on this subject usually asks you to lower the bar, which is unwelcome and, on this evidence, aimed at the wrong half. What runs the clock is the checking: the re-read, the second opinion nobody needed, the pass made to settle a doubt rather than to change the work. So the question is not how good this has to be.

It is how many times you have to look at it before you believe it is.

How do you stop revising the same piece of work?

Suppose your loop is already running. Version four exists, the test in the first section says stop, and you open the file again anyway. Two recent trials are worth knowing here, less for what they recommend than for what they rule out.

The first tested the most intuitive fix there is: get used to making mistakes by making them on purpose. Clark, Patel and Cougle randomly assigned 85 people with elevated perfectionism to one of two eight-session programmes, an exposure treatment built around computerised tasks engineered to produce mistakes, or a stress-management course covering sleep, diet and exercise. They reported it in the Journal of Behavior Therapy and Experimental Psychiatry in 2024, and it went against their own prediction. The stress-management group finished lower on overall perfectionism than the exposure group. The exposure group never settled into the tasks either: distress rose from the first session to the last.

That is a finding about a treatment rather than about your deck, and 85 people in a trial is a long way from your Thursday. Take the narrow lesson. Doing careless work deliberately, to prove that nothing bad happens, is not a technique with evidence behind it, and it can make your next pass feel worse instead of cheaper.

The second trial is more useful for what it failed to find. Rozental and colleagues gave 138 people seeking help for perfectionism one of two eight-week internet programmes, one built specifically for perfectionism and one a general-purpose protocol, and published the comparison in Cognitive Behaviour Therapy in 2024. Both groups ended far better on the perfectionism measure than they started, and were still better at six and twelve months. Neither programme beat the other. Procrastination moved in both arms, by similar amounts.

The authors read that as a possible lack of specificity, and so can you. When two different structured programmes produce the same result, the working part is probably not the theory inside either of them. It is more likely the structure: eight weeks, a defined task each week, and somebody expecting an answer.

Your deadline already does most of that.

So the fixes this site offers you are structural rather than attitudinal. Timeboxing gives your revision a slot and a hard stop, and that is the version of it that matters here. Twenty minutes, one pass, then send. Parkinson's Law applies the same compression from the other end, shortening your runway before the work expands to fill it. The expansion it describes is a heuristic, not a measured effect, so the shorter runway is what you borrow. Its own page puts it in one line: a good enough version delivered on time is often better than a perfect one delivered late.

One more move costs nothing and works on the days the others fail. Put the send in your calendar as an event with a time on it, and tell the person expecting the work what that time is. A deadline you have announced behaves differently from one you are holding privately, because the cost of another pass is now visible to somebody else. This site's page on which half-finished tasks to push through and which to kill runs that logic across a list of separate tasks. Here it runs inside one.

How does your working style change the answer?

Which part of the loop catches you 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, and matching them to a revision problem is editorial judgment rather than a research finding.

Architects write the definition of done and then quietly revise the definition. The plan said four slides and a recommendation; by Thursday the plan says four slides, a recommendation and an appendix, because the appendix would make it complete. If that is you, freeze the definition in writing before you start, and treat every addition as a new task with a slot of its own. The appendix is not wrong. It belongs to a different decision, made on a different day.

Sprinters rarely stall at version four, because the burst that produced the draft is over and the file is shut. The exposure sits at the other end: work sent at the peak of a burst, unchecked, with the error still in it. Your timebox runs the other way. Book one short check after the burst, with a named thing to look for, rather than leaving the question of whether to re-read to whatever mood Friday brings.

Visionaries stall on the framing rather than the finish. The deck is done, and then the angle could be better, and a better angle means most of it gets rebuilt. That pass costs real time. It differs from moving a heading, which is exactly what makes it hard to refuse. Set your cutoff by the clock instead of by satisfaction: reframing is allowed inside the timebox, and after it the work ships as it stands.

Improvisers do not usually polish, so the failure here wears the opposite clothes. Work goes out fast, comes back with a question you could have answered in the first place, and the second round costs more than the check would have. One pass before sending, against a list you wrote once and reuse: the numbers, the names, the ask, the date. What matters is that the list lives somewhere other than your memory.

None of that is a character fault.

Whichever of those is yours, the next step is the same size. Take the piece of work sitting at version three right now and write one sentence naming what it has to do for the person receiving it. Give the check twenty minutes on the calendar. When the time is up, send it, and keep the sentence, because the next piece of work will need one too. If the loop closes but the same task keeps reappearing untouched on tomorrow's list, that is a different problem, and the procrastination pillar starts where this page stops.

Frequently asked questions

How do you know when work is good enough to send?

Ask whether the next pass changes a decision somebody makes, or only changes how the work looks to you. A wrong number changes a decision. A sentence that is fine but could be tighter does not. Write that definition of done before you start, because at version four it will bend to whatever you have just produced.

Should you lower your standards to finish faster?

No, and the evidence points elsewhere: the doubt costs you the time, not the standard. A 2020 meta-analytic review of 67 studies found the subscales about standards related positively to academic performance, while doubts about your own actions and the felt gap between standard and result related negatively. Those were students, so treat the direction as the part that transfers. In practice you keep the bar and cut the number of times you check that you cleared it.

Is perfectionism why I miss deadlines at work?

Sometimes, though the delay usually comes from the checking at the end rather than the standard you set at the start. Count the passes on your last late piece of work and ask what each one changed. If the answer for the last three is wording, the standard was already met before they happened. Giving the check a fixed slot and a hard stop is the cheapest fix, and Timeboxing is the method on this site that does it.

How long should you spend checking work before you send it?

Long enough for one pass with a named purpose, which for most desk work is twenty minutes or less. Book it as a timebox with a hard stop rather than reading until it feels right, because that feeling arrives late or not at all. Give the pass a list: the numbers, the names, the ask, the date. When the box ends, the work goes out.

Sources