It’s the most given piece of advice in this territory and it has a poor record.
You’ve had it from a manager, a book, and probably a friend. You’ve tried it. You produced a shorter list of vaguer things and you still couldn’t start, and you concluded that decomposition doesn’t work for you.
It does. The advice is aimed at the wrong variable, and once you see which one, the failure stops being mysterious and stops being about you.
Here’s the core of it, and everything else follows:
Smaller isn’t the target. Decided is.
The Wrong Variable
The common instruction to break a task into smaller steps identifies size as the relevant property, when the property determining whether an action can be executed is the absence of unresolved decisions. These are distinct: a very small step containing a decision cannot be started, while a large step containing none can be. Size functions as a rough proxy for decidedness, since smaller units tend to contain fewer decisions, but the proxy fails frequently enough to make the advice unreliable. The instruction fails in three further respects. It supplies no test for when decomposition is complete, so the person cannot tell whether they have gone far enough. It presumes that insufficient decomposition is the cause of the inability to act, which is one of several possible causes. And it is framed as a single act, when decomposition must be repeated as execution reveals what later steps should be.
Four failures. The first is the one that matters.
Failure One: Size Isn’t What’s Blocking You
Two examples make this immediately obvious.
“Choose the name.” That’s tiny. Three words. It could theoretically take four seconds.
It’s also completely unactionable, and you might sit with it for eight months. There’s a decision in it, and decisions can’t be executed — only made, and making one requires an input you haven’t got.
“Read this forty-page document.” That’s large. It’ll take an hour and a half.
It’s perfectly actionable. You could start now. There’s nothing to decide — you open it and read, and you’ll be finished when you’re finished.
So one tiny step is impossible and one large step is trivial to begin, which means size is not the variable.
The variable is whether any decisions remain.
Once you see that, a lot of your list re-sorts. The items that have been sitting there for months are almost never the biggest ones. They’re the ones with an undecided thing lodged inside them, frequently a small one.
Size correlates with decidedness loosely — smaller things do tend to have fewer decisions in them — which is why the advice works often enough to survive. But it’s a proxy, and it’s a bad one, and following a proxy while thinking it’s the target is how you end up with a list of small vague things you can’t touch.
Failure Two: No Terminating Test
The second problem is that the advice has no stopping condition.
Break it into smaller steps. How small? Until when?
Without a test you can’t tell whether you’ve finished decomposing, which means you can’t distinguish “I haven’t gone far enough” from “decomposition isn’t the problem here.” Both present identically: a list, and no action.
The test that fixes this is two-part:
Could you begin this within sixty seconds, and complete it without making another decision?
Both halves. If you can’t start within a minute, there’s a hidden step in front of it, and that step is your actual task. If you can start but you’ll hit a decision partway through, you’ll stall there — and the stall will feel like the original paralysis returning, because it is one, at smaller scale.
With that test attached, the advice becomes usable. Without it, it’s an instruction with no endpoint.
Failure Three: It Assumes the Diagnosis
Third, and it’s why the advice sometimes fails for reasons that have nothing to do with decomposition.
“Break it into smaller steps” presumes that underspecification is what’s stopping you.
Sometimes it is. Sometimes what’s stopping you is that the task carries identity weight and a bad first attempt would function as evidence about you. Sometimes it’s that you don’t actually want the goal. Sometimes it’s that you’re genuinely depleted.
Apply decomposition to any of those and you produce a beautifully specified list you still won’t touch — and then conclude that decomposition doesn’t work, when what happened is that you applied the correct fix for a different problem.
The four causes and how to tell them apart are in procrastination isn’t laziness. The relevant point here: decomposition is the fix for one of four things, and the advice is given as though it were the fix for all of them.
Failure Four: Given as a One-Off
Fourth and smallest.
The instruction implies a single act: sit down, break it down, now go and execute.
In practice decomposition is continuous. What you learn from doing the first action changes what the fourth one should be, and a plan produced in one sitting for a project you haven’t started is substantially fiction.
Worse, decomposing the whole project up front is absorbing, satisfying, produces a comprehensive document, and generates no output. It’s a sophisticated way of not starting.
Decompose one module at a time, and only the one you’re about to work on.
Why It’s Given So Confidently
Worth noting, because it explains the persistence of advice that fails this often.
The people who give it are frequently people for whom the decomposition already happened invisibly.
If you work in an environment where tasks arrive specified — where somebody upstream converted an outcome into an instruction before it reached you — then you’ve never experienced the failure mode. Your tasks are already at moment level when you receive them, so breaking them down further genuinely does help, and you’ll recommend it in good faith.
It’s advice from people who have never had to do the step they’re recommending, which is why it names the wrong variable — they’ve only ever seen the version where size was the only thing that varied.
What to Use Instead
Same instruction, corrected in three places:
Resolve until nothing is left to decide, not until it’s small. Size will come down as a side effect. It isn’t the target.
Apply the sixty-second test. Both halves, honestly. That’s your terminating condition.
Check the diagnosis first. If you can already name the exact next physical action and still aren’t doing it, decomposition isn’t your problem and more of it will waste a fortnight.
The full method, including the three points where decomposition reliably stalls and how each converts into a task, is in breaking a goal down far enough to act on.
What This Is Not
It is not a claim the advice is worthless. It points in the right direction and it works for people whose only issue is scale. The problem is that it names a proxy rather than the variable, carries no test, and is applied to problems it doesn’t fit.
It is not an argument that decomposition is hard. It isn’t. It’s twenty minutes with a timer. What’s hard is that it’s boring and that most people stop one level too early.
It is not permission to skip the work. “That advice is flawed” is not a reason to do neither version. The corrected instruction is still an instruction and it still has to be carried out.
The Cost
The price here is small and slightly annoying.
You lose “I’ve tried breaking it down and it doesn’t work for me.”
That sentence has been doing quiet work. It’s evidence-based — you genuinely did try, and it genuinely didn’t work — and it closes the question. If the standard technique fails on you, then the problem is something deeper and more interesting than technique, and there’s nothing obvious to do next.
Once you see that the technique names the wrong variable, the failure is reclassified from a fact about you to an error in the method. Which is better news and it comes with an obligation: the corrected version now has to be tried, and there’s no established record of it failing.
Second cost, minor: the corrected version is duller. “Resolve until no decisions remain, then apply a two-part test” is administrative in a way that “break it into smaller steps” isn’t. It works, and it doesn’t sound like advice worth passing on at a dinner party.
Common Questions
Why does “break it into smaller steps” fail?
Because it names size as the relevant property when the property that matters is whether any decisions remain. A tiny step containing a decision can’t be started; a large step containing none can. Size is a loose proxy and it fails often.
Why can’t I start even after breaking a task down?
Most likely because the smaller items still contain undecided things. Check each one against the sixty-second test: could you begin within a minute, and finish without making another decision? Items that fail the second half will stall you at the decision.
Does task size matter at all?
Loosely, because smaller units tend to contain fewer decisions. But it’s a correlation rather than the mechanism, and following the proxy while believing it’s the target produces a list of small vague items you still can’t touch.
How do I know when I’ve decomposed enough?
Apply the two-part test. That’s the terminating condition the standard advice doesn’t supply, and its absence is why people can’t tell whether to keep going or whether decomposition isn’t their problem.
What if I decompose properly and still can’t start?
Then underspecification wasn’t your cause. The other common ones are identity weight on the task, not actually wanting the goal, and genuine depletion — each needing a different response, and each made worse by the wrong one.
Why is this advice so widely given if it fails so often?
Because it’s frequently given by people who receive tasks already specified by someone upstream. They’ve never experienced the failure, so they’ve only ever seen the version where size was the only variable.
Where to Start
Five minutes. Your existing list. One pass.
Take whatever list you’re currently not executing and go through it marking each item with one letter:
D if there’s a decision left inside it.
A if there isn’t.
Don’t rewrite anything yet. Just mark.
You’ll find that the items that have been sitting longest are almost all D, and that several of them are among the smallest things on the list — which is the whole argument of this article, visible in your own handwriting.
Then take one D and convert it. The decision inside it needs an input, so write the input-gathering task instead. Not choose the platform but ask two people who’ve done this which they used and why.
That’s now an A, and A items get done.
I’ve spent twenty-five years working on how ordinary people develop capabilities that look impossible from the outside, and one of the more common things I meet is somebody who believes they’re bad at a technique they’ve never actually used.
They used a version of it that names the wrong variable, and then drew a conclusion about themselves.