Focus: protecting deep work in a distracting day

The common belief is that some people can genuinely do two demanding things simultaneously. What actually happens is rapid alternation, and each alternation has a cost — the state needed for one task has to be set aside and the state for the other rebuilt.

The cost is not the moment of switching. It is the reconstruction: where you were, what you had decided, what you were about to do next. For anything requiring held context — writing, analysis, planning, debugging — that reconstruction can take several minutes, which means a task interrupted every ten minutes is spending a large fraction of its time being restarted.

The reason this is easy to miss is that fragmented work feels productive. Every one of those ten-minute segments involved effort and attention, so the day reads as full. What is invisible from the inside is the comparison: the same total minutes, uninterrupted, would have produced noticeably more and better work. You cannot observe the counterfactual, so the cost never presents itself as a cost.

Quality degrades alongside speed, and less visibly. Work assembled from fragments tends to be locally sensible and globally incoherent, because holding the whole shape requires exactly the continuity that got broken.

Notifications off is the widely given advice and it is insufficient on its own, because the phone interrupts in two distinct ways and silencing addresses only one.

The first is external: it makes a noise and you look. Turning that off works.

The second is internal: the phone is visible, you wonder whether anything has arrived, and you check. No notification was involved. A silent phone face-down on the desk still generates this, because the prompt is the object itself. This is why out of the room outperforms silent, and it is the single highest-value change available for most people.

The same reasoning extends to the computer, where the equivalents are more legitimate and therefore harder to remove: the open mail tab, the chat application, the browser window with fifteen tabs. Each is a place attention can go without any external trigger.

The part worth being honest about is that this is not primarily an interruption problem. A meaningful share of checking is self-initiated, and often initiated precisely when the work becomes difficult. Removing the option matters most at the moment the work stops being enjoyable, which is also the moment the work most needs doing.

Making yourself unavailable without explanation invites a second attempt, and the second attempt usually succeeds — someone who cannot reach you and does not know when they can will simply try again, or come and find you.

'Back at eleven' resolves this almost entirely. It gives the other person a decision they can make: wait, or judge it urgent enough to interrupt anyway. Most things wait. The absence of that information is what turns a non-urgent request into an interruption, because waiting for an unknown period is not something anyone will choose.

For colleagues this is a stated time and a visible signal — a closed door, headphones, a status message. For customers it is response expectations set in advance, which most customers accept readily provided the stated window is honoured. Predictable responsiveness within known hours is easier to work with than sporadic instant replies, from both sides.

And the arrangement needs a defined exception, or it collapses. Naming what counts as urgent enough to interrupt makes the boundary usable rather than absolute, and people respect a boundary with a documented override far more reliably than one that appears to admit none.

Demanding work has a warm-up. The first ten or fifteen minutes are spent reloading context — what this is, where you left it, what the next move is — and produce little. That overhead is roughly constant regardless of how long the session lasts, which is what makes short sessions inefficient rather than merely small.

A forty-five minute block might yield thirty minutes of useful work. A two-hour block yields something closer to an hour and three quarters. The second is not slightly better, it is more than three times the output, and the difference is entirely the ratio of warm-up to work.

This is also why three short blocks are not equivalent to one long one, and why the reflex to slot deep work into whatever gaps exist between meetings produces so little. A day with three fifty-minute gaps has no space for demanding work in any practical sense, even though it contains two and a half hours of nominally free time.

Two hours is not a magic figure. It is roughly the point where warm-up becomes a minor fraction, and it is short enough to sustain concentration. Anything under an hour is mostly warm-up for genuinely hard work, which is worth knowing before scheduling it there.

A protected block only stays protected if it survives a meeting request, and that is where most of them are lost. A block visible on a shared calendar but treated as available is not protected; it is a preference, and preferences lose to requests.

The workable position is that the block is a commitment with the same standing as any other calendar entry, and moving it requires the same consideration as moving a meeting. That is a defensible stance and it becomes easier to hold if the block is consistent, because a recurring entry at the same time each day gets learned by the people around you and stops being proposed.

Offering an alternative does nearly all the diplomatic work. 'Not at ten, but I have two o'clock free' declines nothing in substance and reads as scheduling rather than refusal.

And the block has to be genuinely used, or the position becomes untenable — for you before anyone else. A protected two hours regularly spent on email is a habit that dissolves within a month, because you will stop believing the block matters, and that is the belief the whole arrangement rests on.

It cannot supply attention that is not there. All of this assumes that with interruptions removed you can direct your concentration at the work. If focus is absent even in genuinely quiet, protected conditions, the problem is upstream — sleep, workload, or something worth discussing with a doctor rather than solving with a stricter phone policy.

It cannot fix a poorly defined task. A great deal of what presents as a focus problem is not knowing what to do next: an unclear objective produces avoidance, and avoidance seeks distraction. The diagnostic is quick — if you cannot state the specific next action in one sentence, the difficulty is definition rather than concentration, and more discipline will not touch it.

It cannot create time in an overcommitted week. Protecting two hours in a week that requires more hours than exist just relocates the shortfall.

And it will not make the work enjoyable. Difficult work is often unpleasant in the middle, and the point of removing distraction is that there is nowhere easier to go when it becomes so. Expecting focus techniques to make demanding work feel good sets up a disappointment that people mistake for the method not working.