Choosing a Format You Can Still Read in Ten Years

The data outlives the tool. Choosing the format is therefore a longer-term decision than choosing the application, and it is usually made without thought.

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

What a durable format looks like

Readable without the program that made it. Open it in a text editor and the content is there.

When this method depends on cloud backup, the material from Microsoft OneDrive offers a helpful independent point of comparison.

Line-based, so it can be processed by anything.

Self-describing: a header row, or a structure obvious from looking.

And stable: a format whose specification has not changed in twenty years is unlikely to change in the next ten.

Plain text, CSV and simple JSON meet all four. Almost nothing else does.

Why proprietary formats fail

Not because the company disappears, although it might.

Because the export is lossy. Applications export what their export function supports, which is frequently less than what they store, and the parts lost are usually the ones you added yourself.

Because pricing changes, and a record held hostage to a subscription is not yours.

And because the format encodes assumptions from the application's design, which will not match the next one's.

The practical arrangement

Use whatever application you like — they are genuinely better at daily entry than a text file.

Export monthly to a plain format, automatically if the tool allows.

Keep the exports somewhere you control, not only in the tool's own cloud.

This costs a few minutes a month and it is the whole of the insurance.

What the export should contain

One row per entry.

Date and time in an unambiguous format — the international one, year first — because every other convention is read differently somewhere.

Duration as a number of minutes, not as a formatted string.

Category and any tags as plain text.

And your free-text note, which is the field most often dropped by exports and the most valuable in five years.

Checking the export actually works

Open it. Read a row. Can you understand it without the application?

Count the rows against what the application says you have.

And try importing it somewhere else once, which is the only real test and which people do for the first time during a migration under pressure.

Time zones and the other traps

Record local time and, if you travel, the offset. A record of a year abroad with no offset is a record you cannot align with anything.

Avoid formats that store only elapsed time with no timestamp; they cannot answer any question about when.

And beware of tools that store duration as text like "1h 30m", which is readable by humans and painful for everything else.

If you prefer paper

Entirely viable and some of the longest-running records are paper.

The durability problem inverts: paper survives software changes and not floods, moves or fires.

Photograph or scan periodically if the record matters, which converts the main risk into the one you already know how to handle.

What to check

Could you read your record without the application you use?

When did you last export, and did you open the export?

Is the date format unambiguous?

And does the export include your notes, or only the numbers?

The point

The data outlives the tool, which makes the format a longer-term decision than the application.

Use whatever you like for daily entry and export monthly to something plain.

Worth adding

Paper inverts the risk entirely: no vendor, no subscription, no export problem, and one copy vulnerable to floods and moves.

For day-level narrative records the trade frequently favours it.

Additionally

Record local time and the offset if you travel.

A year abroad with no offset is a year you cannot align with anything else in the record.

Finally

Avoid formats that store only elapsed time with no timestamp, and beware of durations stored as text like one hour thirty.

Both are readable by humans and painful for everything else.

In summary

The data outlives the tool.

Use whatever application you like for daily entry, export monthly to something plain, and keep the exports where you control them.