From the first sheet to a clean export
The same four steps run on every sheet, in the same order. What changes between one record and the next is how long each step takes, never whether it happens.
- 1
Set up the sheet
One sitting to decide what a row means, which columns are required, and which rules an entry has to pass. Almost everything else can be changed later; those three decisions are the ones worth making slowly.
- 2
Record as it happens
Entries go in on the day, in one place, from whichever device is nearest. A series kept in three places is three series, and at least two of them are wrong by the end of the quarter.
- 3
Check the batch
Duplicates, gaps and values outside the stated range surface while you type rather than at the end of the month, and the running total gives you a second opinion on the batch you have just entered.
- 4
Export and keep
A clean export whenever you want one, and a scheduled copy that lands where you keep your own archive. The archive is yours, and it is readable without the tool that produced it.
1. Setting up the sheet
Setting up is a working session, usually an hour or two, with the person who will actually be typing. The output is written down: what one row represents, which columns are required, what unit each figure is in, and which rules an entry has to pass before it is accepted.
The most valuable part of that note is the list of things the sheet deliberately will not hold. A record that never writes that list down discovers it anyway, two years later, as a column somebody filled in three different ways.
2. Recording as it happens
Entries go in on the day, from whichever device is nearest, into one sheet. Recording as it happens is not a preference; it is the difference between a series and a reconstruction. A figure written down within the hour is checked against the thing itself, while a figure entered at the end of the month is checked against somebody's memory of it.
Where the connection is missing, entry continues from a local copy and reconciles later, so a basement with no signal never becomes a reason to keep a second list on paper.
3. Checking the batch
Four checks run while you type: a duplicate of the row above, a gap where the sequence should be continuous, a value outside the range the column states, and a running total you can hold against the total on the sheet you were reading from. Each one is shown, none is enforced silently, and an override is written into the history with the reason you gave.
Two minutes of careful comparison at the end of a batch removes almost every error that would otherwise be found in a year's time, when the only remaining evidence is the figure itself.
4. Export and the archive
An export can be taken at any moment and carries the values, the column definitions and the revision history. A scheduled copy lands wherever you keep your own archive, on a rhythm you choose, and the printable sheet exists for the situations where paper is still the format that survives.
The archive is deliberately readable without us. If a file only makes sense inside the program that wrote it, then the program, and not you, is the one holding the record.
After the first month
The first month of real entry is when the setup decisions get tested, and it is planned for rather than improvised. Columns that nobody fills in get removed, a rule that fires on every second row gets rewritten, and the routine rhythm of entry settles into something that survives a busy week.
Columns nobody uses get removed. A sheet that only ever grows becomes slower to fill, harder to read and easier to abandon, one reasonable addition at a time.
What a sheet needs from you
- One decision about what a row means, made before the first hundred rows exist.
- A habit: entries on the day, not on the last day of the month.
- Two minutes at the end of each batch for the running total.
- One place where the archive lives, and a second copy of it somewhere else.
How the tool reports on itself
Each series carries a short status line: how many entries it holds, when the last one was added, which checks failed and were overridden, and when the last export was taken. Four figures, on purpose. A screen of charts about your own record keeping is a way of avoiding the question of whether the record is current.
When the tool is not the right fit
Sometimes the honest answer is that what you keep is prose rather than a series - case notes, minutes, a diary - and a document is the better container for it. Sometimes what is needed is a full accounting package with double entry and a statutory trail. Saying that in the first exchange costs a morning. Discovering it after eight months of entries costs the record itself, and it is the more common of the two.