Activity is the count of things a device observed: keys, clicks, windows, minutes with input present. Productivity is whether valuable work happened. The two are connected by a relationship that varies enormously between roles and is weakest precisely where the work is most valuable.
The device is what is being measured
An agent observes a machine. It cannot observe a conversation that solved the problem, a meeting, a whiteboard, a phone call, a printed document, or the twenty minutes somebody spent staring at a wall and then wrote the thing correctly the first time.
This is the limit described in the article on what cannot be measured, and here it has a consequence with somebody's job attached: the further a role's value sits from a keyboard, the more badly an activity score misrepresents it.
The classification is a configuration, not a fact
Every one of these tools sorts applications and sites into productive and unproductive. That sorting was done by a person, usually during onboarding, for roles that have since changed.
A browser is both. A technical question-and-answer site is research for one person and avoidance for another, and nothing observable distinguishes them. Reading documentation looks identical to reading anything else. A spreadsheet open with nobody typing into it is either analysis or an abandoned tab.
The productivity score is therefore a measurement of an opinion, expressed in numbers, and the numbers carry an authority the opinion never had.
Good work is frequently quiet
Concentrated work produces long periods of low input followed by a burst. Averaged over a day it looks like a poor day. The same average is produced by someone who spent the day in a chat window doing nothing in particular, and by a device moving a pointer with nobody present.
An activity score cannot distinguish those three, and of the three it is the last that it scores most favourably, because that one is optimising directly for the measure.
The metric selects against integrity
This is the part that deserves stating without hedging. Once activity is scored and the score has consequences, the person who works honestly in long quiet stretches is ranked below the person who keeps the pointer moving.
The organisation is not merely getting a bad measurement. It is applying a penalty to the behaviour it wanted and a reward to the behaviour it did not, continuously, and the article on measures becoming targets explains why the effect grows rather than settling.
Output measures are harder than they look
The obvious response is to measure output instead, and every candidate has a known failure. Tickets closed rewards closing rather than resolving. Story points are a team's own estimate and move when they are watched. Lines of code have been discredited for decades and still appear. Revenue is lagging, collective and mostly outside any individual's control.
The honest position is that the value of most knowledge work is assessable by judgement and not by counting, and that the appropriate role for data is to inform a conversation rather than to replace one. An organisation unwilling to accept that will keep buying numbers that answer the wrong question confidently.
Where activity data is genuinely useful
Aggregated across a team and used to examine a process, it earns its place. A department spending a large share of its hours in one internal tool is describing that tool. Time fragmented into short blocks across a whole group is describing a meeting culture. Neither conclusion needs anybody's name attached, and the article on who sees the dashboard argues that attaching one makes the conclusion worse rather than sharper.
It is also the clearest available signal of overwork. Consistent late-night and weekend activity across a team is a staffing finding. It is worth noticing that using it that way is still surveillance, and that the reasonable version involves telling people it is being looked at and what will be done about it.
Time theft is a framing, not a finding
The phrase appears throughout this market's marketing and it carries an assumption: that what an employee owes is hours, and that a shortfall in observed hours is a taking.
For hourly work with a contractual clock, that framing is coherent. For salaried knowledge work it presupposes the conclusion, since the obligation there is generally the work rather than the time. Adopting the vocabulary adopts the argument, which is worth noticing before it appears in a policy.
What we cannot verify
Correlations between activity measures and any independent measure of value are not published in any form we would rely on, and the parties who could produce them are the ones selling the measures. The bimodal pattern of concentrated work is well described by practitioners and is not something we have measured. Nothing here is advice on how to evaluate an individual, which is a question for the organisation and, where consequences follow, for its legal counsel.
The short version
- An agent observes a machine, not the work, and value often sits away from the keyboard.
- The productive and unproductive lists are somebody's opinion expressed as numbers.
- Concentrated work is quiet, and an average cannot tell it from doing nothing.
- Scored activity ranks the honest worker below the one keeping a pointer moving.
- Every output measure has a known failure; judgement is not replaceable by counting.
- Aggregated, it diagnoses processes well; attached to a name it gets worse, not sharper.