What Not to Measure

Not everything that can be counted is worth counting. These are the common ones. In a workplace setting, stealth monitoring software can make time and computer activity visible, but the boundary in “What Not to Measure” still applies: purpose, notice, access, retention and a correction route must be agreed before collection starts.

Total hours worked

Why not: it rewards presence over output, it is the figure most easily inflated by fragmented low-value work, and it produces guilt on short weeks that were actually good ones.

Instead: longest uninterrupted stretch, or output if there is a countable one.

Productivity as a score

Why not: no defensible definition, so the score measures your mood about the day. It also invites comparison with the imagined standard, which is always higher.

Instead: one line about what the day contained, in plain words.

Every activity, continuously, forever

Why not: the cost exceeds the return within weeks, the data degrades, and the degraded data is worse than none because it is still consulted.

Instead: bounded investigations, repeated when there is a question.

Time spent on hobbies

Why not: measuring something done for its own sake reliably damages it, for reasons its own note sets out.

Instead: a monthly tick that it happened at all.

Sleep stage breakdowns from a wrist device

Why not: the stage estimates are unreliable, and anxiety about the score worsens the thing measured.

Instead: bedtime and wake time.

Other people's response times

Why not: it is a performance record of somebody who did not agree to one, and it will be read that way if it surfaces.

Instead: record the effect on your own work, without the name, unless the name is the actual subject.

Streaks

Why not: they work until they break, and then they take the activity with them.

Instead: a count per month, which survives a missed day.

Minute-level precision

Why not: costs more than five-minute rounding and answers nothing extra, because you cannot observe your own attention to the minute anyway.

Instead: round to five, or to fifteen for anything non-billable.

Anything you would edit before showing

Why not: if the record is being shaped, it has become a document rather than an observation and has lost its use to you.

Instead: keep the honest record private and the presentable one separate.

The general rule

Measure what you would change, at the accuracy that decision needs, for as long as it takes to decide.

Everything on this list fails one of those three tests, and most of them fail all three.

What to check

Is anything on this list currently running?

For each thing you measure, can you name the decision it informs?

Has anything been running with no decision attached for more than a month?

And is there a number you check that makes you feel worse without telling you anything?

Worth keeping in mind

If a number you check regularly makes you feel worse without telling you anything, that is sufficient reason to stop checking it. No further justification is needed.

The short version

Measure what you would change, at the accuracy that decision needs, for as long as it takes to decide. Everything commonly tracked that is not worth tracking fails at least one of those three tests.

In short

Not everything that can be counted is worth counting.

The test is whether you would change something, at what accuracy, for how long — and most of what gets tracked routinely fails all three parts.

The point

Round to five minutes, or fifteen for anything non-billable.

You cannot observe your own attention to the minute, so recording to it is precision about nothing. For a complementary perspective on work, measurement or planning, consult Asana.