Productivity Is Not Enough.

There is a question one can ask about almost any practice that, asked seriously, exposes a great deal: what is being measured?

The contemporary answer, for most kinds of intellectual work, is some version of throughput. Words written. Tickets closed. Tasks completed. Meetings conducted. Models trained. The metric is the rate at which output leaves the worker.

This metric has obvious uses. It is observable, comparable, schedulable, and incentivisable. An organisation in which throughput is measured is an organisation in which throughput improves. This is not nothing.

It is also not the same as capability. The two come apart in ways that the throughput metric is structurally unable to see, and the gap between them — between what a person can produce in a sprint and what a person can do, unaided, a year from now — is the gap this essay is about.

§ 01The two metrics

Throughput measures the rate at which one finishes things. Capability measures the depth and breadth of what one could finish, given a problem one has not seen.

The two are correlated, but loosely. A person of high throughput and low capability is a person who closes many tickets that look the same, and stalls on the next slightly-different one. A person of high capability and modest throughput is a person who, when handed an unfamiliar problem, produces a workable solution where others would have asked for help. Most senior technical careers are built, in the long run, on the second.

The interesting point is not that capability matters more than throughput. (It does, but the claim is unsubtle.) The interesting point is that the activities which produce throughput frequently degrade capability, and the activities which produce capability frequently cost throughput. The metrics actively pull against each other.

The activities which produce throughput frequently degrade capability. The activities which produce capability frequently cost throughput.

§ 02How throughput degrades capability

Three mechanisms are recurrent.

Substitution. Tools that increase throughput often do so by performing the cognitive operation in place of the user — drafting, summarising, deciding. Each substitution improves the moment-to-moment metric and atrophies the substituted capability. The throughput goes up; the capability the throughput was supposed to be expressing goes down. The two diverge silently.

Avoidance of slack. The activities that produce capability — sustained reading, deliberate practice, problem-finding, idle reflection — produce nothing measurable in the moment. They look like slack. An organisation, or a person, optimising against throughput will reliably squeeze them out. The squeeze is invisible until the next non-routine problem arrives, at which point the absence of capability becomes legible all at once.

Granularity drift. Throughput is easier to measure on small, similar units. Cumulative throughput pressure pushes work toward smaller and more uniform units, because they are easier to count. The deeper, less granular work — the rare, the difficult, the sui generis — becomes administratively hard to do, and so is done less. The capability for that kind of work, having had less practice, declines.

§ 03The metric problem

One might imagine that the solution is simply to measure capability instead of throughput. Unfortunately, capability is hard to measure in any time window short enough to be useful for incentives. It is most clearly visible in the events one has been spared from — the problems that did not become crises, the projects that did not fail, the colleagues who did not have to be hired. None of this lights up on a dashboard.

The asymmetry is structural. Throughput is observable in the unit of time the modern organisation operates in. Capability is observable in the unit of time the modern organisation does not. The result is that, even when both are valued, the observable one wins by default.

This is why the discipline is, finally, personal. The metric problem cannot be solved by the institution. It can only be solved by the individual who has decided that, regardless of what is being counted, they intend to remain capable, and who therefore protects — actively, against the local incentive gradient — the slower work that produces the capability the institution will need but is not currently in a position to ask for.

§ 04What this implies in practice

Concretely:

• Some fraction of every working week is reserved for activities whose output is not visible in the throughput metric. Reading. Practice. Problem-finding. Reconstruction. The fraction is small enough to be sustainable and large enough to be real.
• Tools are evaluated against capability, not only throughput. A tool that doubles output and halves the user's ability to operate without it is not, by the relevant metric, a good tool.
• Self-evaluation occurs at the unit of capability, not at the unit of output. The annual question is not how much did I produce? but what could I do this year that I could not do last year?
• Slowness, where it produces capability, is defended as a value rather than apologised for as a cost.

§ 05A note on the cult of throughput

It is worth saying explicitly: this is not an argument against productivity. It is an argument against productivity as the only metric. Producing things is one of the more meaningful uses of a working life. The error is to conclude that, because production is good, only the rate of production matters.

The slower work is not a failure to be productive. It is the activity that makes future productivity possible. It is also, in many cases, the activity that makes the productivity of others possible — the load-bearing thinking on which the visible output downstream eventually depends.

An hour spent reconstructing a proof from memory looks, in any throughput-shaped measurement system, indistinguishable from an hour spent on nothing. From a sufficiently long view, it is the opposite: it is the hour the rest of the year quietly rests on. The argument of this essay is that the long view, increasingly rare, is increasingly worth keeping.