Trust versus monitoring software: what you lose when you install it

Activity monitoring buys one number and sells the habit that made the team work. We installed it on forty laptops in January and lost two engineers by April.

The month the screenshots started

The company had forty employees and one director who could not see whether anyone was working. A tool promised a dashboard that answered the question at a glance.

It captured active windows, idle minutes, and a screenshot every ten minutes. Installation took an afternoon, and nobody in the team was asked to agree to it.

By the second week the dashboard was open on a screen in every leadership meeting. The number was always visible, and it was always slightly lower than expected.

What the tool actually measures

It measures input events on one machine. A person thinking for twenty minutes looks identical to a person asleep, because both leave the keyboard alone.

Dario, a senior engineer, spent most of his week reading and deciding. His idle time was the highest on the team and his output was the highest too.

The dashboard showed the first figure and could not show the second. We had built a measurement of typing and then called it a measurement of work.

The behaviour it reliably produces

Within three weeks people learned to move the mouse during phone calls. One support agent admitted keeping a document open to look busy while reading it.

That is not laziness, and calling it laziness is the expensive mistake. It is rational behaviour inside a system that pays for visible motion, and the system was ours.

Every minute spent performing activity is a minute not spent on the queue. The tool did not measure that loss, so nobody counted it.

Idle time is not absence

The highest idle score in the company belonged to the person who wrote our payment integration. He thought for long stretches and typed in bursts.

Deep work looks like nothing from the outside. A dashboard calibrated for support tickets will always misread a person solving a hard problem.

We knew this before installing. What we underestimated was how quickly a number on a screen replaces the judgement we had already made.

The senior people notice first

Dario asked whether the screenshots included his lunch break. When told they did, he said nothing and stopped joining the optional calls.

He resigned eleven weeks later and gave no reason beyond a better offer. Two others followed within a month, both from the same team.

The cost was not the licence fee. It was three people who knew where the fragile parts of the system were, replaced over five months.

The team that stops writing things down

Written work becomes risk once it can be screenshotted. People started keeping drafts local and moving decisions into calls where nothing was recorded.

Our documentation thinned out without anyone deciding to thin it. The monitoring tool had made written evidence slightly dangerous, so written evidence became scarce.

For a distributed team this is close to fatal. The written record was the only mechanism holding six time zones together, and we had made it risky.

A cheaper alternative that already existed

We already had most of the data we needed. Closed tickets, merged changes, resolved incidents, and the queue depth at the end of each week told us more.

Those signals cost nothing and describe outcomes instead of keystrokes. They also survive a person changing laptop or working from a train.

The reason we ignored them was that they needed interpretation. A dashboard is attractive because it removes the need to think about context.

Measure the handover, not the keystroke

One metric proved more useful than the entire dashboard. Did the work arrive in a state the next person could pick up without asking a question?

That is measurable through review comments and the number of clarifying messages after a handover. Both are already in the systems we use.

It also measures the thing a distributed team actually needs. Activity is local, and handover quality is what crosses a time zone.

The suspicion outlives the software

We removed the tool in June. The habit of assuming surveillance lasted longer, and one team kept a shared note of who was online at nine.

Trust does not return when a process is deleted. It returns when someone with authority says plainly that the deletion was a correction.

That sentence took two minutes to say and should have been said in January. Saying it in June cost us the benefit of the doubt.

When monitoring is defensible

There are cases where it holds up. One is a regulated environment with audit obligations; another is a specific incident, investigated with a stated end date.

Even then the scope should be narrow, disclosed in advance, and given an end date. A permanent, silent, company-wide feed is a different instrument entirely.

The distinction is whether it answers a question you can name. If nobody can state the question, the tool is collecting on spec.

Rolling it back without a speech

We deleted the agent, published the removal date, and explained what would replace it. The explanation was three paragraphs and named no one.

Two managers objected, both of whom had used the dashboard in one-to-ones. We offered them the handover measure instead and they used it.

What we did not do was apologise at length. A long apology keeps the topic alive, and the team wanted the topic to end.

What trust actually buys

Trust buys early warning. People tell you about a slipping deadline while it can still be fixed, because they are not protecting themselves from a number.

That information is worth more than any dashboard. It arrives in a sentence, from a person who believes the sentence will not be used against them.

By autumn our incident reports were longer and our outages were shorter. Neither change shows up in a productivity tool, and both were the point.

An activity dashboard is a confession that outcomes are not being read. If the only evidence of work is a screenshot, what would you change before switching it off?