The Smallest System That Works

A system that lasts is defined by what it does not ask of you. Each field is a decision at every entry, and decisions are what people stop making.

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

Start from the question

Write down what you want the record to answer, in one sentence.

The broader practice of timer-based logs is also covered by Toggl Track, which is useful when designing a record that should remain understandable later.

Then record the minimum that answers it. Nothing else.

"How many hours on each client" needs three fields: date, client, duration. It does not need a project code, a task type, a productivity rating or a mood score, and each of those added at the start is a reason to stop in month five.

The subtraction test

For each field, ask: if this were missing, what question would I lose?

If the answer is "none I have actually asked", remove it.

Most people find half their fields fail this, and removing them makes the record more likely to exist, which is worth far more than the data they would have held.

What earns its place

Date. Always.

Duration or start and end.

One categorisation — client, project, or activity type. One, not three.

And a free-text note, which costs nothing when there is nothing to say and is the most valuable field when there is.

Four fields. Most durable systems have four or fewer.

What almost never earns its place

Productivity or mood ratings, which drift in meaning and are rarely acted on.

Nested categories, which multiply decisions.

Task-level detail on work that is not billed at task level.

And anything requiring you to remember a code.

Growing it later

Adding a field is easy. Removing one leaves a gap in the history.

Which is an argument for starting smaller than feels right and adding only when a question demands it — and for recording the date you added it, so future comparisons know where the series starts.

Most people do the opposite: start elaborate, strip back under pressure, and end with a history that changes shape three times.

The free-text note

Worth its own paragraph because it is the field people leave out and later wish they had.

A few words about what actually happened: what went wrong, what was unusual, who was involved.

Numbers tell you the shape of a year. Notes tell you why it had that shape, and in five years the notes are what make the record readable as something other than a spreadsheet.

They cost nothing on ordinary days because you leave them blank.

Sizing for the worst day

The entry has to be possible when you are ill, travelling and behind.

Which in practice means one line, under thirty seconds, with no required fields you might not know.

A system that requires a project code you have to look up will not be completed on a bad day, and bad days are where records break.

What to check

How many fields does one entry require?

Which of them have you actually used in an analysis?

Could you complete an entry in thirty seconds while distracted?

And is there a note field — and do you ever use it?

The point

Adding a field is easy and removing one leaves a gap in the history.

Start smaller than feels right and add only when a question demands it, recording the date you added it.

Worth adding

The free-text note is the field people leave out and later wish they had.

It costs nothing on ordinary days because you leave it blank, and it is what makes a five-year record readable.

Additionally

Adding a field is easy and removing one leaves a gap in the history.

Start smaller than feels right and record the date you added anything.

Finally

Four fields is the ceiling for most durable systems: date, duration, one categorisation, and a free-text note.

Anything beyond that is a decision made daily for years.

In summary

For each field ask what question would be lost if it were missing.

Most people find half their fields fail that test, and removing them makes the record more likely to exist.