The Unit Problem in Knowledge Work
Productivity needs a comparable unit of output. Most professional work does not have one, and inventing one causes specific harm.
The arithmetic requires that units be alike. In manufacturing they are. In analysis, design, advice and management they are not, and no amount of effort makes them so.
The practical lesson in “The Unit Problem in Knowledge Work” is to connect every number to a decision and retain the context behind it. Teams exploring time tracking software can review time tracking software for measurable workflows as one source of operational evidence, provided the purpose is disclosed and the interpretation is tested with the people affected.
What comparability requires
That two units represent similar amounts of work.
For an independent perspective related to “The Unit Problem in Knowledge Work”, consult the American Psychological Association workplace resources; it offers a useful external check on definitions, governance and the assumptions built into a proposed measure.
That they are substitutable: two of one equals one of something twice as large.
And that counting them captures most of what was produced.
Professional work fails all three routinely.
The examples
Two reports: one a summary, one a three-month investigation.
Two support tickets: a password reset and a corrupted database.
Two designs: a colour change and a new product.
Counting them equally is not an approximation; it is a different measurement entirely.
What inventing a unit does
Story points, complexity ratings, weighted case counts.
These help teams plan and they are estimates made by the people being measured.
Which means using them for assessment corrupts them immediately, and the estimates drift to whatever produces the desired figure.
That is not dishonesty; it is what happens when a planning tool is repurposed as a scorecard.
The visible consequence
Work gets split into more, smaller units.
Difficult items get avoided or deferred.
And the measure rises while the same amount of work is done, which the traps section covers under Goodhart.
What to do instead
Measure flow: how long things take from request to done, which does not require units to be equal.
Measure rework: how often things come back.
Measure waiting: how much of the elapsed time was queueing.
None of these needs a comparable unit and all of them are about the work, which is the knowledge-work section's argument.
The honest position
Some work cannot be counted, and saying so is better than producing a number that can be quoted.
An organisation that reports "we do not have a reliable productivity measure for this function, here is what we do track" is in a stronger position than one with a figure that falls apart under examination.
Its own note covers when the honest answer is that you cannot.
Where a unit does exist
High-volume comparable work inside otherwise professional functions: standard contracts, routine claims, first-line queries.
Measure those properly and leave the rest to flow measures, rather than applying one approach to a whole department.
What to check
Are your counted units comparable in effort?
Is any planning estimate being used for assessment?
Has work been split into smaller pieces since measurement began?
And what do you track for the work that has no unit?