Categories That Age Well

Categories are the analysis, fixed in advance. Over a long record this matters more than any other structural decision, because changing them retroactively is either impossible or a substantial project.

For a practical application of this principle, see project time tracking. It provides a supporting reference for testing a small real-world sample without adding unnecessary friction to the record.

What ages badly

Project names. Projects end. A record with forty dead project categories is unreadable, and the successor categories are not comparable with them.

The broader practice of statistical review is also covered by the R Project, which is useful when designing a record that should remain understandable later.

Employer or client names, as the primary division. Useful for billing, useless for asking what your working life looked like across two jobs.

Anything tied to a tool or a platform, which will be replaced.

Value judgements — productive, wasted, deep — whose meaning drifts with your mood and your circumstances.

And anything with more than about eight options, because consistency collapses beyond that.

What ages well

Kinds of activity rather than instances: writing, meetings, admin, learning, maintenance.

Broad enough that a change of job does not empty them.

Defined by what you are doing rather than by who it is for.

And mutually exclusive, with the overlaps decided once and written down.

The two-axis approach

One axis for the kind of activity, stable over years.

A second for the specific instance — project, client — which can change freely.

This is the arrangement that survives, because the stable axis carries the long comparison and the volatile one carries the detail.

Recording both costs one extra field and solves the main structural problem of long records.

How many

Five to eight on the stable axis.

Fewer than five and the categories stop distinguishing anything. More than eight and the decision cost rises, consistency falls, and by year two you are guessing.

The volatile axis can have as many values as you like, because you are not comparing across them.

Defining them

One line each, written down beside the data.

Not for anybody else: for you in four years, wondering whether "admin" included invoicing or only email.

Include the overlap rules: a client email counts as client work, a meeting about admin counts as admin.

This file is what makes year one comparable with year five. Without it the older data is present and not safely usable.

The catch-all

Every scheme needs one, and it should be watched.

A general bucket prevents the deliberation that kills entries on bad days.

If it grows past about a tenth of your time, it is hiding a category you have not created — which is a finding rather than a failure, and it appears in the quarterly review.

Testing a scheme before committing

Take one past week you remember well and categorise it.

If more than a couple of entries are ambiguous, the scheme has an overlap problem that will produce inconsistent data for years.

Ten minutes now against a structural problem later.

What to check

How many of your categories are named after things that will end?

Could a change of job empty one of them?

Have you written down what each means?

And what proportion is going into the catch-all?

The point

Two axes: one stable for the kind of activity, one volatile for the specific instance.

The stable axis carries the long comparison and the volatile one carries the detail.

Worth adding

Every scheme needs a catch-all, and it should be watched.

Past about a tenth of your time it is hiding a category you have not created, which is a finding rather than a failure.

Additionally

Test a scheme before committing: categorise one past week you remember well.

More than a couple of ambiguous entries means an overlap problem that will produce inconsistent data for years.

Finally

Categories named after projects, clients or tools age badly because those things end.

Kinds of activity survive a change of job; instances do not.

In summary

Five to eight categories on a stable axis, defined in writing, with the overlap rules decided once.

Fewer stops distinguishing anything and more collapses into guessing.

Last word

The practice this points at is unglamorous: a small daily act, a short weekly correction, an annual read, and a written page of conventions beside the data. None of it is difficult and all of it is easy to skip, which is why most records do not reach the point where they start paying.