Article
Six things started, none finished: which to push through and which to kill
When six things are open and none is finished, rank them on what each still costs against what it still returns, and ignore the hours already spent. Those hours are gone whichever way you decide. Then finish one before you start anything new, because the number of open tasks is what created the problem. Past effort pulls harder than it should, and there is now direct evidence that it pulls on mental effort and not only on money. This page covers the rule, what the research does and does not support, and how to tell someone you are dropping their task.
How do you decide which unfinished task to finish first?
Wednesday afternoon, and the list reads like an accusation. A draft at eighty percent. A script that runs but has never been tested properly. Two reviews you opened and left. A deck somebody will ask about on Friday, and the piece of work you started the week meaning to do, still sitting at a heading. Nothing there is abandoned. Nothing there is finished either.
The instinct in that moment is to pick up whichever one you have already put the most into, because stopping it now would waste the work. That instinct is the one thing on this page worth arguing with. The hours are spent. They stay spent whether you finish the task this afternoon, park it until next month, or delete it, and nothing you decide now buys a minute of them back.
So price what is left instead. For every open item, two estimates: how much work it still takes to reach a state somebody else can use, and what it is worth once it gets there. Add a third question wherever other people are involved, which is who stays blocked while it sits. Rank on those and the order that comes out will not match the order of your investment. That mismatch is the reason to run the pass on paper rather than in your head.
The hours you already spent are evidence about your estimates. They are not an argument about this afternoon.
Ten minutes is usually enough, and it thins a list further than people expect. Two or three items turn out to be nearly done and worth closing on the spot, which is the cheapest result on offer. Some need a real block and should get one on the calendar rather than another line on a list. And one or two will not survive the second question at all, because the work still costs a day and nobody is waiting for what it produces. Those are the ones to kill deliberately, while killing them is still your decision.
Should you finish a task you have already spent hours on?
Sometimes, and the hours are still not the reason. Your case for continuing is that what remains is small and somebody wants the result. Your case against is that neither of those holds, and the task has been running on momentum since about its second week. Telling the two apart from inside the task is hard, which is why the pull has a name and a research literature behind it.
Until recently that literature was mostly about money. Sunk-cost effects are well documented for economic decisions, and the evidence for non-monetary costs such as cognitive effort had been mixed. Bogdanov and colleagues went at that gap directly in a 2026 paper in Cognition, with two experiments in self-reported healthy adults, 36 in person and 89 online. Participants worked through an effort investment task built to test the question you are sitting with: does past effort change a present choice about continuing or abandoning a course of action?
Across both experiments, prior investment substantially increased how willing people were to invest further effort. The second result is the one that describes your Thursday: when continuing demanded higher additional effort, participants were more likely to keep going after a large prior investment than after a small one. The pull got stronger exactly where carrying on cost you the most.
Which is the shape of a half-finished task nobody can quite let go of.
Now the limits, because they matter here. That was an effort task in a lab, not a migration with a manager attached to it, and the paper reports nothing about how much a real backlog shifts. This page will not invent a figure for it. What transfers to your own list is the direction of the pull, not its size.
A second caution comes from inside the same field. Bialek and Biesiada tested the sunk-cost scenarios that decades of research have leaned on, across two experiments with 395 participants, and published the result in Brain Sciences in 2025. Internal consistency ran between 0.14 and 0.57, correlations between scenarios were weak, and even closely similar scenarios correlated only moderately. Their conclusion is about measurement rather than about the bias itself: the standard scenarios do not reliably measure one thing.
So treat the pull as something that happens to people, yourself included, rather than as a score you could measure and then manage.
How do you tell your manager you are dropping a task?
Dropping it quietly is what most people do, and it is the version that costs you in a month. The work stays live on somebody else's mental list, gets asked about at the worst possible moment, and by then you have neither the output nor any record that a decision was made. Say it while it is still a decision rather than an absence.
The message has three parts and fits in four lines: what you are stopping, what finishing it would still cost, and where that time is going instead. I have stopped the vendor comparison. It needs about two more days, and those days are going into the launch checklist, which lands first. Tell me if that is the wrong trade. No apology in it, and no argument that the earlier work was wasted, because neither helps the person reading it decide anything.
Often the reply changes your decision, which is the point of sending it. Managers rank across work you cannot see. The piece you were ready to abandon is sometimes the one somebody downstream has been waiting on quietly, and the piece you were about to rescue occasionally stopped mattering a fortnight ago. Neither of those facts reaches you while the call stays inside your own head.
Then fix the reason there were six of them.
Working on one thing at a time is not the answer, and nobody works that way. A cap is. Personal Kanban is built for this exact state: three columns, and a hard limit on how many cards may sit in the middle one, so nothing new starts until something leaves. The cap is the method, and it is the part that gets quietly treated as a suggestion the first busy week. If a board is more apparatus than you want, the Ivy Lee Method enforces the same thing through order alone, because its rule is that you do not touch the second task until the first is finished.
Both are cheap enough to test against the list you already have open. Neither helps if the six were started because six different people each asked for one and none of them knows about the other five. That is a prioritization problem with other people in it, and it gets solved out loud rather than on a board.
How does your working style change the answer?
Which part of this 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.
Architects rarely start six things by accident, so their exposure is the kill rather than the count. A task that was planned, scheduled and justified is much harder to stop than one that drifted in, because the plan itself becomes the argument for finishing it. Sprinters have the opposite trouble. A burst opens a task quickly and the next burst opens another, and what gets left behind is a row of items at eighty percent that each need a dull hour nobody feels like spending.
Visionaries stall in their own particular way here. The interesting phase of a piece of work is over by the time it is three-quarters done, and what remains is tidying, so their abandoned pile fills with things that were genuinely worth doing and are now merely worth finishing. Improvisers keep the most open, since starting is cheap for them and a shifting week gives every start a reason. For them the cap does more than any ranking rule, because ranking assumes a list that holds still long enough to be sorted.
Some weeks this page does not describe. If all six are mandatory, funded and owned by different people, no finish-or-kill pass gives you the authority to kill any of them, and the honest move is to show the six to whoever believes you have room for them. If the same task has sat at ninety percent for two months while you finish everything around it, what is missing is more likely a definition of done than any pull from sunk cost. That one needs a finish line written down, not a better ranking.
So open whatever holds your work in progress and count what is genuinely started. Against each one write two numbers, the hours it still needs and the hours it has already had, then cover the second column with your hand. Finish the cheapest useful thing today, put the largest survivor on the calendar, and send one message about the one you are stopping. The hours under your hand are not coming back either way. The only question they can still answer is how good your estimates were.
Frequently asked questions
How do you decide which unfinished task to finish first?
Rank your open tasks on what each still costs and what it still returns, and leave the hours already spent out of the comparison. Add a third question wherever other people are involved: who stays blocked while the task sits. The pass takes about ten minutes. Expect it to close two things on the spot, put one on the calendar, and leave you holding one that nobody is actually waiting for.
Should you finish a task you have already spent hours on?
Only when the work left is small enough and somebody still wants the result; the hours behind you are not a reason either way. That is harder than it sounds, because past effort pulls. In a 2026 Cognition paper, Bogdanov and colleagues ran two experiments, 36 participants in person and 89 online, and found that prior investment substantially increased willingness to invest further effort. The pull was stronger when continuing demanded more additional effort, which is the case where carrying on costs you most.
Why do I start more tasks than I finish at work?
Because nothing limits how many can be open at once, and starting is cheap while finishing is the part that takes a dull hour. Personal Kanban attacks that directly with a cap on how many items may sit in progress, so nothing new begins until something leaves. The Ivy Lee Method gets there through order instead, since you do not touch the second task until the first is done. Either way it is the count you are fixing, not your willingness.
How do you tell your manager you are dropping a task?
In writing, in three parts: what you are stopping, what finishing it would still cost, and what you are doing with that time instead. Leave the apology out, and do not argue that the earlier work was wasted, because neither helps them decide. Often the reply changes the decision, which is the reason to send it. Managers rank across work you cannot see, so the task you were ready to drop is occasionally the one somebody downstream is waiting on.