Granularity You Can Sustain

Granularity is usually decided by what the tool makes easy rather than by what the question needs. Over a long record that choice compounds.

A concrete way to test the method is outlined in the planning resource. Use a short pilot and keep only the fields that still make sense when comparing the record months later.

The levels

Day-level: one entry per day, what it mainly was. Cheapest, and adequate for surprisingly many questions.

For a neutral companion reference on structured workspaces, see Notion; it helps separate the choice of format from the conclusions drawn from the record.

Block-level: a handful of entries a day, at the granularity of "morning on the report, afternoon on calls".

Task-level: every switch recorded. Expensive, and the only level that answers questions about fragmentation.

Continuous: start and stop timers on everything. Highest cost, required for billing, unsustainable for anything else over years.

Matching to the question

"How much time on each client" needs block-level.

"Is my work getting more fragmented" needs task-level, and probably a bounded investigation rather than a permanent system.

"What did my year actually contain" needs day-level, which is why some of the longest-running personal records are one line a day.

Most people choose task-level for a day-level question and stop within a year.

The cost of each switch recorded

Fifteen to thirty seconds per entry including the decision.

At four entries a day that is a couple of minutes; at twenty it is ten, every day, for years.

Ten minutes a day is sixty hours a year. That is a real price and it should buy something specific.

Mixed granularity

Legitimate and underused: fine granularity on billable work, coarse on everything else.

Or fine during a period when you have a question, coarse the rest of the time.

Record which mode you were in, because otherwise a period of dense entries looks like a busier period rather than a more thoroughly recorded one.

That confusion is common and it invalidates comparisons.

Coarsening deliberately

A system that started fine and is slipping should be coarsened officially rather than left to degrade.

Dropping from task-level to block-level, with the date recorded, produces a comparable series in two segments.

Letting it slip produces a single series that silently changes meaning, which is the drift problem in another form.

The floor

Below day-level there is nothing worth keeping.

A record of "this week I mostly did X" cannot answer anything and will not survive, because it never rewards the effort.

Day-level is the minimum that produces findings, and it takes fifteen seconds.

What granularity does to the note

At day level the note is the valuable part: the numbers say little and the sentence says what happened.

At task level the numbers carry the information and notes become impractical.

Which is worth knowing when choosing: coarse records are read as narrative, fine ones as data, and they age differently.

What to check

What question requires your current granularity?

Could you drop a level and still answer it?

How many entries do you make on a typical day, counted?

And if you have mixed granularity over time, is the change dated?

The point

Ten minutes a day is sixty hours a year.

Granularity is a real price and it should buy a specific answer rather than a feeling of thoroughness.

Worth adding

Mixed granularity is legitimate and underused: fine on billable work, coarse elsewhere.

Record which mode you were in, or dense periods will look busier rather than better recorded.

Additionally

Below day-level there is nothing worth keeping.

A record of what you mostly did this week cannot answer anything and will not survive, because it never rewards the effort.

Finally

Coarsen deliberately rather than letting a system slip.

A dated change produces two comparable segments; drift produces one series that silently changes meaning.

In summary

Match granularity to the question.

Day-level answers what a year contained, block-level answers where time went, and task-level answers questions about fragmentation at ten times the cost.

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.