I had the most beautiful GTD system in the world. It did nothing for me.

A few years ago I had the most beautiful GTD system I have ever come across. It ran in OmniFocus, and every task trailed a little row of emoji. I had just learned that there is more than one clock emoji: one for the full hour, one for quarter past, one for half past. So a task that takes fifteen minutes got the quarter-past clock. On top of that, symbols for whether something needs a lot of thinking, a little thinking, or none at all because it is quick. It looked colourful, it felt creative, and I liked showing it off.

Someone I demonstrated it to looked at it for a while and then asked: what does this actually do for you?

My answer came quickly. The emoji tell me how long a task takes and how much head it costs me. The follow-up was uncomfortably precise. If you already know while writing it down that vacuuming takes fifteen minutes, you will still know it later when you only read the line. And then the second one: how long do you need to pick the right emoji and stick it on? Ten seconds. On every task, every day.

The system collapsed in on itself during that conversation. It had a price per capture and a benefit of zero.

Why systems only ever grow in one direction

The overstuffed systems I get to see almost never came out of one wrong decision. They came out of a long run of right ones. You read about a trick, you try it, it helps you that week, and it stays. Nobody puts an expiry date on a new list.

On top of that, lists rarely start out wrong. They turn wrong. When context lists were invented, the computer stood in the office, which made it a place: if you wanted to be at the machine, you had to travel there. Today the phone in your pocket is a computer, and the context no longer answers any question. I kept an office list for years, even though I was self-employed and working from home. The list was not set up wrong, it simply became pointless at some point, and I carried it around anyway.

You pay for that in two places, and both are quiet. At capture, every extra decision costs you seconds. In your weekly review, every list costs you attention, even when there is nothing on it.

The question that decides every list

There is one question that settles this in a few seconds. When did I last go into my system deliberately, for this one list alone, to find something?

Not: when did I last see it. Not: when did I last write something into it. But: when did it take a decision off my hands? For the pretty tags, the honest answer is usually never.

Underneath sits a distinction that rarely gets said out loud. A context is a hard constraint. It answers whether you can do a thing here and now at all. An attribute is a preference. It answers whether you want to do it right now. Constraints earn their own list, because in a particular moment they do the searching for you. Preferences only earn one if you genuinely filter by them.

Some people do. If you regularly have twenty minutes of dead time before your next appointment and you actually want to use them, a time estimate is not decoration, it is the filter that makes the list usable. So the test is not a matter of taste. Two people with the same tag can come out of it differently.

List or tag	What it answers	Deserves its own list
Agenda per person	Who am I seeing or speaking to next?	Yes, the window is short
Errands per place	Am I there right now?	Yes, otherwise you go twice
Waiting For	Who owes me something?	Yes, otherwise it vanishes quietly
Computer	Am I sitting at a machine?	Once yes, now almost never
Low energy, 15 minutes	Do I fancy it right now?	Only if you really filter by it

For most people, a handful of context lists is what survives, plus one agenda for each person they deal with regularly.

Why “not used in three months” is the wrong test

The moment you start clearing out, you want a rule that makes the decision for you. The obvious measure is time: anything you have not needed in three months goes. I think that is wrong, and it is wrong precisely for the lists that are worth the most.

Take a list for the hardware shop. If you keep one, you might need it three times a year. By the time rule it would have to go, and that would be a mistake, because in exactly the moment when you are standing there anyway, it saves you the second trip. The same goes for everything that hangs on a rare opportunity: the tax paperwork, the appointment at the doctor’s surgery, the person you see twice a year.

The better measure has two parts. Does the list hold up in the moment it was built for? And what does it cost you in all the other moments? A place list costs almost nothing, because when you write the thing down you already know that drill bits come from the DIY shop. A time estimate costs you a few seconds of thinking at every single capture. Frequency is a measure for attributes, not for opportunities.

Simplifying a GTD system means deleting, not reshuffling

When a list disappears, its tasks do not disappear with it. That is where the clear-out usually gets stuck, and that is where it gets interesting what the list really was.

If every task immediately finds another list that already exists, the old one was a duplicate and you have lost nothing. If a task finds no home, that is the most valuable result of all: either you are missing a list for a situation that genuinely exists, or the thing has been dead for ages and has to beg for a place before anyone notices. And if you cannot decide, the task was never processed. It is sitting there as a topic, not as a next action, which is why it fits everywhere and nowhere.

In my own system the test caught a list that stands in every textbook: Out and about. I kept it because it gets recommended. At some point I noticed that I have to look at it every single day, because I am out and about every single day, and that it holds exactly one task. I need to print a page and I have no printer. My idea behind it was that while I was out, I would happen to walk past a printer that was desperate to print something. That is not going to happen. The task was on the wrong list; it belongs to the person who has a printer, and therefore on their agenda. Without it, Out and about has no content left and no reason to show up every morning.

With projects the same move has a different aim. Anything that is guaranteed to sit untouched for the next month or two does not get deleted, it goes on the Someday/Maybe list, and once a month you check what is allowed back.

The meta layer costs you at capture, not at search

The second price of all that superstructure is not paid when you search. My rule of thumb: if capturing something takes longer than ten seconds, the friction is too high. Honestly, I now think ten seconds is generous.

The moment in question is always the same one. I sit down on the sofa in the evening, and the second I am sitting, something good occurs to me. All that matters is holding on to that thought in a way that lets me find it again tomorrow. “Mum” is entirely enough. You will still know tomorrow why she came to mind.

If you have to decide in that moment which list the thing belongs to, how long it takes and which tag it gets, two things happen. Sometimes you just do not do it, and the thought stays in your head. And when you do do it, you rob yourself of the chance to process it properly.

The obvious objection is fair: you cannot assign anything in ten seconds. True, and that is exactly why assigning does not belong there. Capturing is not clarifying, and clarifying is certainly not organising. If you type the idea straight into the right list, you skip the step where the real question gets asked: what is this, and what is the next action?

Where this usually goes wrong

• The clear-out happens in the middle of the work. On a Tuesday morning you notice that a tag does nothing for you, and you strip it out everywhere on the spot. It feels productive, and it is the reason nothing is finished by the evening. The decision belongs in the weekly review, the tidying up in an appointment of its own.
• Tinkering gets banned instead of scheduled. Building on your system is fun, and there is nothing wrong with that. Ban it and you will do it anyway, just secretly, at nine o’clock on a Monday morning. Set up a project that is named for what it actually is, give it a fixed slot, and put it behind the work, not in front of it.
• The tasks from the deleted list end up in a leftovers box. A catch-all list with no criterion is not a simplification, it is the old state under a new name. When a task finds no home, that is a result, not a problem to be filed away.
• The big cull is followed immediately by the next rebuild. Whether a structure holds up shows after a few weeks at the earliest, and most clearly in the weekly review. Bring in the next concept in the meantime and you will never know which one did the work.
• Afterwards the system looks better than it works. Tidy tiles, good names, matching colours: the effect is real, it just sits in the wrong place. The test stays the same. Which decision did this list take off your hands this week?

Common questions

How do I know it is time to simplify my GTD system?

The clearest sign is that you push the weekly review back even though you have the time. Then there is the hesitation at capture: if you would rather keep a small thing in your head than type it in, the friction is too high. And if you have lists you only look at because they exist, the system is no longer carrying you, you are carrying it.

How many context lists make sense?

There is no right number, but there is a test that holds: every list has to take a decision off your hands in a specific situation. For most people a handful survives, plus one agenda for each person they deal with regularly. Anything that only describes how a task feels is an attribute, not a list.

What do I do with the tasks when I delete a context list?

You move them one by one before you remove the list, and only onto lists that already exist. If a task finds a home straight away, the deleted list was surplus. If it does not, that is the interesting exception: either a list is missing for a situation that genuinely exists, or the task is done and nobody ever ticked it off.

Is it wrong to put a time estimate on tasks?

No, as long as you filter by it. The estimate costs you a few seconds at every capture and only pays off if you really do go looking for short tasks while you are waiting around. If you do that regularly, keep it. If you cannot remember a single time you did, you are maintaining it for the look of it.

When should I tinker with my system, if not on the side?

In a planned slot that does not sit at the start of a working day. A fixed slot after lunch works well, or a monthly appointment, so the urge has somewhere to go. What matters is the order: first the work that is due, then the system you enjoy fiddling with.

---

Open your system and find the one list you have only been looking at for weeks because it is there. Do not delete it. Write down what is on it instead, and next to each item, where that thing would go if the list no longer existed. Whatever you cannot place is your answer.

If the clear-out leaves you with the question of whether a list can really go, that is rarely about the list and usually about the structure underneath it. That is exactly where the Workflow Setup starts: one pass through your system, until all that is left in it is what you actually use. Write to me if you have spent years working with a system that grew along with you and now weighs more than the work inside it.