Why To-Do Lists Fail (21,655 Tasks of Data)

TL;DR. I counted every task in Loggd. Across 21,655 one-off tasks from 1,375 people, 7,774 were completed, which is 35.9%. The shape matters more than the rate: 40.8% of completed tasks were finished the same day they were created, the median completed task took 1 day, and the ones that did not move quickly now sit a median of 56 days past their own planned date. A to-do list is not a reservoir that drains slowly. It is a queue that either moves today or turns into archaeology.

I build a habit and task tracker, which means I can do the thing most productivity writing cannot. I can open the database and count.

So I did, in August 2026. The raw population is 51,201 tasks across 1,375 users, with my demo account excluded.

Then I threw out more than half of it, and you should know why before you trust a single number below.

The 29,546 tasks I threw out

57.7% of those tasks, 29,546 of them, are not decisions anybody made. They are auto-generated instances of recurring tasks: the "take the bins out" that reappears every Tuesday whether you want it or not.

Recurring instances behave nothing like the things people type into a list at 11pm. The app generates them, not a person having a thought. Counting them would either flatter the completion rate or wreck it, depending on how many instances got spawned for a routine somebody quit in March. Either way the number stops meaning what the headline claims.

So everything from here on is one-off tasks only: 21,655 tasks across 1,375 users. Things somebody deliberately wrote down once.

The uncomfortable number

7,774 of those 21,655 tasks were completed. That is 35.9%.

Most of what people write down never gets ticked off.

The obvious objection is that this is an accounting artefact: plenty of tasks are simply not due yet, and any snapshot catches a pile of healthy pending work. It does not hold here. 13,880 one-off tasks are still open, and this is what they are:

Fewer than one in a hundred open tasks is legitimately pending. Recompute the completion rate over settled work only, meaning completed plus overdue plus never-scheduled, and it barely moves: 36.1%.

There is no escape hatch in a "cancelled" bucket either. Exactly 1 task in the entire snapshot is cancelled. Tasks do not get retired, they just sit there.

How old the rot is

This is where it stopped being an abstract percentage for me.

Among the overdue tasks, the median is 56 days past its planned day and the 90th percentile is 132 days. 826 distinct users are carrying at least one.

Two months past a date the person picked themselves. The top tenth is more than four months gone: things written in spring, still nominally scheduled, still producing a small pulse of guilt every time the app opens.

I have several of those. One of them has moved house with me.

The actual finding: tasks are done immediately or never

Here is the part that changed how I think about lists.

Of the 7,774 completed tasks:

• 40.8% were completed the same day they were created.
• The median time from creation to completion is 1 day.
• The 90th percentile is 11 days.

Read those three lines next to the 56-day median rot and the shape falls out. There is no slow, steady drain where old tasks get picked off over weeks. Nine in ten completed tasks were done inside 11 days, and 40.8% before the day was over. Anything still open past that window is not in progress. It is archaeology.

That reframes what a to-do list actually is. It is not storage. It is a queue with a very short effective memory, and writing something into it does almost nothing to raise the odds that it happens next month. If it does not move now, or nearly now, the honest forecast is that it never will.

Even the tasks people finished missed their date

One more number, and it is the one that makes me most sympathetic to everyone here.

Take only completed tasks that had been given a planned day: 7,506 of them, across 853 users. How many landed on the day they were planned for?

55.1%. Another 38.7% were completed late and 6.2% early.

So even inside the population of pure successes, nearly half missed their own date. That is not an effort failure, it is an estimation failure, and it is consistent enough that Kahneman and Tversky named it: the planning fallacy, our habit of predicting how long something takes from the version of the story where nothing goes wrong.

The date you assign a task is not a plan, it is a wish with a timestamp. Adding precision to your scheduling is mostly wasted effort. Reducing the number of things you schedule is not.

The aggregate is not you

Now the honest correction, because "35.9% of tasks get done" is the kind of stat that gets screenshotted without its context.

That aggregate is dragged down by one pattern: somebody signs up, dumps forty things they have been carrying around in their head, feels enormously better, and never comes back. Those forty tasks sit in the denominator forever.

Filter to people who used the thing more than once. Among the 340 users who created at least 5 one-off tasks, the median personal completion rate is 54.5%.

That is a very different picture: the typical engaged person finishes a bit more than half of what they write down, which sounds about right for a human with a job.

The tail is real too. 45.3% of those users finish under half of what they write down. So the aggregate is pessimistic, and the individual reality is still that a large minority of people are running a permanent deficit against their own list.

What I would actually change

All of this points at the same four moves, and none of them are "try harder".

1. Cap the day, then let the cap do the deciding. Since almost everything that gets done gets done within a day or so of being written, the useful question is not "what needs doing" but "what am I doing today, given that today is roughly all I get". A fixed small number is the crudest and most effective version of that. The 1-3-5 task planner makes the constraint concrete: one big task, three medium, five small, an optional time estimate per task, and a prompt to carry over yesterday's leftovers. Nine slots, and when they are full they are full. I covered the method itself in the 1-3-5 rule guide.

2. Decide the first action before the day starts. A task that reads "taxes" has no entry point, so it loses to every task that does. The 55.1% on-time figure suggests the weak link is specification rather than willpower. Pick the one thing that must move and name its first physical step. Eat the frog is the least complicated version of that ritual.

3. Let the backlog go, deliberately. Given the 56-day median, a task that is two months overdue is not waiting. It is dead and unburied. Once a quarter, move the whole overdue pile somewhere that is not a to-do list. A brain dump suits this: write everything out without filtering, optionally on a short timer, then tag each line as a task, idea, worry, or reminder. Most of it turns out to be worries and ideas wearing a task costume. The survivors go back on the list with a date. The rest get deleted, which is a decision, not a defeat.

If you would rather sort than dump, the Eisenhower matrix does the same triage across four urgency and importance quadrants, and lets you drag things between them as you change your mind.

4. Plan fewer things rather than better things. Every real improvement I have made to my own list was a subtraction. How to plan your day covers the ritual side, but the arithmetic is blunt: if 35.9% of written tasks get done, writing twice as many does not double your output. It doubles your overdue queue.

You may have seen the Zeigarnik effect cited around this topic, the idea that unfinished tasks stay active in memory and nag at you. It matches how an overdue list feels, but its replication record is mixed rather than settled, so treat it as a decent metaphor for why a stale backlog feels heavy, not as a mechanism you can bank on.

Methodology, and what this data cannot see

The figures come from a production snapshot queried on 2026-08-10, covering data through 2026-08-05. Population: 51,201 tasks across 1,375 users, demo account excluded. Auto-generated recurring instances, 29,546 tasks or 57.7% of the raw count, are excluded from every figure, leaving 21,655 one-off tasks. Sub-samples are stated inline where they differ: 7,506 completed tasks that had a planned day, across 853 users, and 340 users with at least 5 one-off tasks.

Four limits, in descending order of how much they should bother you.

A task app cannot see work done elsewhere. This is the big one. Plenty of these incomplete tasks were done in real life and never ticked off, because the person did the thing and got on with their day. I am measuring the list, not the person, and every completion rate here is a lower bound.

This is correlation, not an experiment, and the population is self-selected. Nobody was assigned to a condition. People who install a tracker are not a random sample: they already write things down, which cuts both ways, more organised than average and more prone to over-capturing.

Survivorship pulls the aggregate down. Users who abandoned the app entirely still have their unfinished tasks counted. That is honest, since dropping the quitters is how you manufacture a flattering number, but it means the 35.9% headline includes many one-visit lists nobody intended to work through. The 54.5% median among engaged users is the fairer comparison.

Rounding and scope. Everything is reported to one decimal place, no individual user or task text is included, and nothing is extrapolated. This is one app's data, not a claim about humanity.

The one sentence version

If a task is not going to move in the next week or so, writing it down more carefully will not save it. The only lever with real evidence behind it is writing down less.

I have never met anyone whose problem was a to-do list that was too short. Including me.

Last updated: December 2026.

Written by Eusebiu, the solo founder building Loggd in public. I build the app in the evenings around contract work and a small daughter, which makes my own overdue queue a live research subject. More aggregate findings in what 4,000 habit trackers reveal, and the running commentary on Threads.

See your real completion rate instead of guessing at it. Loggd keeps tasks, habits and goals in one place and shows honestly what got finished. Start free.