“One row, one thing” means that every row in a table represents the same kind of record. That might be a customer, an invoice or an invoice line. Choosing this level of detail prevents accidental double counting and makes the table easier to combine with other records.
Put facts about that record in separate columns. Avoid mixing transaction rows with monthly totals, section headings or handwritten explanations inside the data area. Keep explanations nearby, where people can read them without treating them as transactions.
A small-business example
An invoice contains three products. An export repeats the full invoice total on each of its three product rows. Adding the total column now counts the invoice three times.
The team separates invoice-level facts from line-level facts. Each line has its own quantity and line value; a separate invoice table stores the invoice total. Both tables share an invoice identifier, making their relationship explicit and their totals easier to check.
Try this
Finish the sentence “Each row in this table represents…” without using “and”. Check ten rows, including a return, a blank-looking row and any total row. If different rows need different sentences, split the table or document a clear record type before calculating anything.
Unique identifiers connects related records. Data dictionary explains each field, and Duplicates helps distinguish repeated records from legitimate repeated events.