The hidden tax of always-on tools

The number that started this study

The first time we plotted focus-session length against chat activity, the chart looked broken. On days when a knowledge worker’s messaging app was active more than 45 times, their median uninterrupted work block was 11 minutes. On quieter days it was 34 minutes. Same people, same jobs.

Eleven minutes is not enough time to load a hard problem into your head, let alone make progress on it. Most engineering, analysis, and writing tasks need 20 minutes just to reach the point where the real work begins.

We assumed we had a data error. We didn’t. We re-ran the analysis three times, restricted it to organisations with clean instrumentation, and controlled for role mix. The eleven-minute figure barely moved. What we had found was not an anomaly but the ordinary texture of the modern workday, rendered visible for the first time at scale.

How we measured it

Momentum’s agent records which application holds focus, in aggregate windows, without reading content. That gives us a clean behavioural signal: how long people stay in one tool before switching, and what pulls them away. For this analysis we sampled anonymised, aggregate usage patterns across the platform.

We defined an interruption as a switch into a communication tool lasting under three minutes, followed by a return to the prior application. Short, reactive, ping-shaped. We deliberately excluded scheduled meeting joins and lunch-hour browsing, because those are choices rather than interruptions.

One methodological note: we can’t see whether a switch was triggered by a notification or by habit. Behaviourally, it barely matters. The cost of the switch is identical either way, and the habit only exists because the notifications trained it.

The recovery cost is the real tax

The interruption itself averages 74 seconds. That sounds cheap. But the return journey is where the money goes: after a chat interruption, people took a median of nine additional minutes to re-enter a sustained focus state, measured as continuous engagement with their primary work application.

Multiply it out and the arithmetic gets ugly. A person interrupted 25 times a day loses roughly two hours to recovery alone, on top of the interruptions themselves. Across a 200-person organisation, that is the equivalent of hiring 45 people whose entire job is context reconstruction.

Worse, recovery cost isn’t linear. The third interruption in ten minutes costs more than the first, because at some point people stop trying to return to depth at all. Their afternoon becomes a queue of shallow tasks, which feels busy and produces very little.

Why it feels productive anyway

Answering a ping delivers a small, immediate, social reward. You helped someone. You were responsive. Deep work delivers nothing for hours and then everything at once, which is a terrible dopamine schedule for a human brain sitting next to a glowing sidebar.

There is also a visibility asymmetry. Fast replies are witnessed by colleagues; quiet concentration is witnessed by no one. In organisations where responsiveness has become a proxy for commitment, the rational move for any individual is to stay interruptible, even though it is collectively ruinous.

One product team we worked with in Berlin called this “the availability trap” and had genuinely believed their sub-two-minute response times were a cultural strength. Their focus data showed the tax they were paying for it: the lowest deep-work ratio of any team in the company.

The trap tightens with seniority. Senior people get pinged more because they know more, which means an organisation’s scarcest expertise ends up with its shallowest attention. Several customers only noticed this after plotting interruption load by tenure, and the slope of that chart is rarely comfortable viewing.

What the quietest teams do differently

We compared the top decile of teams by sustained-focus ratio against the median. The quiet teams were not less collaborative; they sent slightly more messages overall. The difference was timing. Their communication clustered into two or three daily bursts instead of an even drizzle across eight hours.

Most had a shared convention rather than a policy: mornings for depth, a midday window for anything requiring a reply, and a norm that same-day is fast enough for nearly everything. Where escalation was genuinely urgent, it went through one loud channel that people therefore trusted and never muted.

The Berlin team adopted the batching pattern in October. By January their median focus block had gone from 14 minutes to 29, and their internal survey scores on “I have time to do my job well” rose by a third. Response times slowed by an average of 38 minutes. Nobody noticed.

The tooling half of the fix

Behavioural change decays without structural support. The teams that held their gains changed defaults, not just intentions: notifications batched hourly, sidebars collapsed, keyword alerts reserved for incidents. Several disabled read receipts entirely, removing the social pressure to demonstrate instant attention.

Aggregate visibility helped here in a way we did not anticipate. When a team can see its own interruption pattern as a group, the problem stops being about individual discipline and becomes an engineering problem with a shared dashboard. Engineering problems get fixed.

One caveat we insist on with customers: never use interruption data to police individuals. The person switching contexts 60 times a day is usually the victim of the pattern, not its author. Fix the flow, not the person drowning in it.

Counting what silence is worth

Run the numbers for your own organisation. Take your average fully loaded salary, estimate two recovered hours per person per day at the current interruption rate, and the annual figure will be uncomfortable. For a 200-person company at typical rates it lands somewhere north of $3 million.

The estimate is conservative, if anything. It ignores the quality cost of work done in eleven-minute fragments, the errors that shallow attention produces, and the slow attrition of people who came to do deep work and found themselves employed as human routers instead. None of those appear on a dashboard, but every experienced manager has watched them happen.

That number is why we call it a tax. It is paid continuously, invisibly, by everyone, and it funds nothing.

The teams in our data who reclaimed even half of it did not work longer hours or adopt heroic discipline. They just stopped treating constant availability as free. It never was.

Writes for The Signal about research and the future of measurable, humane work — drawing on anonymised patterns from the teams and focus hours analysed on Momentum.