What Deep Work Actually Is (And What It Costs)

The term has become a synonym for working hard, which has cost it most of its usefulness. Working hard is common. What the term originally pointed at is not.

Here is the definition this site uses, and everything after it is an argument for taking the definition seriously enough to exclude things.

Deep work is cognitively demanding work, done without interruption, on one thing, for long enough that the output is better than a fragmented version would have been.

The load-bearing word is continuity. Not difficulty, not duration, not importance. A four-hour session that is broken every twenty minutes is not deep work that got interrupted; it is a different and worse activity that happens to have taken four hours.

The four characteristics

It is cognitively demanding. The difficulty is in the thinking. If the hard part is coordination, chasing, or the number of small steps, that is a different kind of hard.

It is continuous. The value comes from not being broken. This is the characteristic that makes it fragile and the one everything else follows from.

It is single-threaded. One problem, held. Not three problems visited in rotation, however smoothly.

It produces something that did not exist. A judgement, a draft, a design, a proof, a decision. Deep work has an artefact, even a rough one. Time spent thinking that leaves nothing behind was probably rumination.

The Depth Test

Here is the part almost nobody does. Before you protect an hour for something, check whether it can use one.

A task can reward depth only if all four of these are true.

1. Improvement. Does uninterrupted time make the output better, not just arrive sooner?
2. Cognitive load. Is the difficulty in the thinking rather than the logistics?
3. Compression penalty. Would a ten-minute version be materially worse — not merely shorter?
4. Advanceability. Can one sitting either finish it or move it somewhere it could not otherwise reach?

Fewer than four means the task is shallow. Not unimportant — shallow.

Three examples, one of which fails

Writing the strategy section of a proposal. Improvement: yes, the argument gets better with continuity. Cognitive load: yes. Compression penalty: yes, a ten-minute version would be a list of assertions. Advanceability: yes, one sitting produces a draft. Four out of four. Protect it.

Reviewing a colleague’s technical design for flaws. Improvement: yes, noticing what is wrong requires holding the whole design at once. Cognitive load: yes. Compression penalty: yes, a skim finds typos and misses the architecture. Advanceability: yes. Four out of four. Protect it.

Processing forty emails, three of which need real answers. Improvement: no — a longer runway does not make the replies better. Cognitive load: mostly no. Compression penalty: no, the ten-minute version of most of them is the correct version. Advanceability: it can be finished, but that is one out of four. Shallow. Batch it.

The third example is the one that matters, because it is the one people protect time for. It feels demanding. It generates a sense of having worked. And a morning given to it is a morning that could not be given to the first two.

Shallow work is not the enemy

This needs saying plainly, because the deep-work conversation has developed a tone about it.

Shallow work is most of what keeps an organisation upright. Answering people, scheduling, filing, reviewing the small thing, unblocking someone. It is real work and someone has to do it.

The error is not doing it. The error is protecting it — giving it the treatment that only depth benefits from. Shallow work should be batched: gathered into a defined window, done at speed, at a time of day when your thinking is worst anyway, and then closed.

Batching works because shallow work is interruption-tolerant. You can be interrupted mid-inbox and lose almost nothing. That property, which makes it shallow, is also what makes it safe to put in the least protected part of your day.

Being busy is neither

There is a third category, and it is the largest.

Busyness is activity that feels like work, generates a record of having worked, and produces nothing that would be missed. Meetings that exist to demonstrate alignment. Documents nobody reads. Threads about a decision that was already made.

The Depth Test does not catch busyness, because busyness usually fails on all four criteria and so gets correctly sorted into shallow — where it is then done efficiently, at speed, forever. The only test that catches it is a different one: if this did not happen, who would notice?

That is outside the scope of this piece and mostly outside your control. But it is worth knowing that a depth practice can be working perfectly while the shallow batch it protects you from is itself largely unnecessary.

What depth actually costs

Advice about deep work tends to present it as free — as though the only obstacle were insufficient discipline. It is not free. Here is the bill.

Recovery. Ninety minutes of genuine cognitive load leaves you diminished for a while afterwards. Two such sessions in a day is achievable. Four is a claim people make and rarely a schedule people keep.

Availability. For the length of the session you are not reachable. If your role includes being reachable, that has a real cost that falls on real colleagues, and pretending otherwise is how people end up resented rather than respected for it.

Displacement. The hour comes from somewhere. Every protected hour is an hour of something else not happening, and the something else does not disappear because you have decided it is shallow.

The social cost of the boundary. Being unavailable is a statement, and it is read as one. In some teams it reads as seriousness. In others it reads as grandiosity. You do not get to choose which, and the piece on protecting time takes that seriously rather than pretending the cost away.

When depth is the wrong call

There are weeks when the correct answer is no depth at all.

A hard deadline where the work is known and the only variable is throughput. A coordination-heavy week where your value is being responsive. A period where nothing on your list passes the Depth Test, which happens more often than the genre admits and is not a personal failing.

Insisting on a deep session in those weeks produces a protected hour spent resenting the protection. Better to batch everything, get through it, and put the practice down for a few days. A focus practice that cannot be put down is a system designed for a week you do not have.

Where the term comes from, and where we depart

Deep work in its current sense comes from Cal Newport, whose book made the economic case: that the ability to concentrate without distraction on a demanding task is becoming both rarer and more valuable, and that this combination is unusual.

We think that case was made and has largely been won. Where we part company is not with the argument but with what it left open. Newport’s contribution was establishing that depth is valuable and schedulable. The question it did not have to answer — because it was not the book’s question — is how you decide what deserves it.

That decision is where most people’s practice actually fails. They know depth is valuable. They have the calendar. And they spend the protected hour on the third example above.

Depth is not scarce because people lack discipline. It is scarce because deciding what deserves it is genuinely hard, and nothing in the working day is designed to help you do it.

The test above is our attempt at making that decision mechanical enough to actually happen. Run it on tomorrow’s list before you protect anything.